This is the final part of The Ouroboros Sequence, a series on puzzle games.

Do I hate Snakebird? I don’t love it.

It was October 2016 and I had been playing Snakebird (Noumenon Games, 2016) for about a month. Finding it impossible to make progress around twenty levels in, I strapped the game to a concrete block and threw it into the choppy waters of the Thames. And while I did this, others continued to carry the Snakebird banner high, proclaiming its brilliance. I didn’t get it. I really didn’t get it.

I wanted to know why. I scribbled into my book of half-baked ideas an article about why I don’t like Snakebird. But the exercise would be pointless without an answer to the question.

Time passed. I had many notes that sketched out areas of potential interest and I found myself more interested in the player side of the puzzle game equation rather than the design side. I wanted to think about how puzzle games might be failing their players instead of players failing their puzzle games. Failure can become very personal, hardening players against continued experimentation in the puzzle genre.

I grew frustrated with the whole concept, though. I could never weave these ideas into a complete whole. I was looking at an infinite number of darts thrown at an infinite number of dartboards by an infinite number of monkeys… but not one bullseye.

Screw it, I thought, I’m going to write the series anyway, even if I don’t have any conclusions. I’ll call it The Ouroboros Sequence because Snakebird is behind the whole series and I knew I wasn’t going to be any wiser about my disaffection for Snakebird after reaching the end. The truth was in the title, up there in every episode. No revelations would be coming. We’d end in the same place as we started. Disaffected with Snakebird, reasons unknown.

And yet. And yet.

This journey has not been a waste of time.

We have returned to the beginning, to Snakebird. But things are most definitely not the same.

A Theory, A Plan

Over the course of the series, I developed my pet Snakebird theory which went like this:

Snakebird was a puzzle game for the hardcore, thus its tight design did not encourage players to review their solutions (Dead in the Water). I suspected this meant I was not learning the lessons I was supposed to. Therefore my progress was happening entirely through guesswork (The Monte Carlo Player) and I eventually got mired in levels where blind guesswork was not enough. At that point, because I was hitting the same puzzles again and again (All Roads), I was going around in circles and felt trapped (Claustrophobia). Eventually, I used a walkthrough to make further progress and that was the death knell for feeling invested (Hole In My Chest).

But how to test the theory? To figure out if there was truth in this, I decided to have a second pass at Snakebird. I admit I wasn’t really looking forward to it; playing the game had left such a horrible taste in my mouth. A really bad taste, like something had crawled in there, had babies, and the whole family had died from a grotesque skin infection.

The plan was to proceed through the levels in precise order and never leave levels I was “stuck” on. I would carefully analyse each solution and attempt to intuit heuristics. What was a puzzle trying to tell me? Did I miss the clues? Would I need a walkthrough?

In the middle of last year, I kicked off a fresh game of Snakebird and wrote notes on every single bloody level. These notes ran to 4,000 words. The things I do for this site. The things. I promise no swears. Cross my heart.

Let me quickly say the following analysis reflects my particular strengths and weaknesses which will be different for every player. I also do not think every Snakebird level sports a unique solution; I’m pretty confident some levels have multiple solutions.

The Student

The first few levels comprise the first year of your Bachelor of Snakebird Arts at Sokoban University. Get out the magnifying glass because we are going to need to look at detail.

Level 00 / Baby Steps

The economy of Snakebird level design is marvellous. Level “0” doesn’t just introduce how a snakebird moves and that fruit must be eaten before the level completes… no, it also conveys that eating fruit extends the range of a snakebird. Simple and important. But this last lesson is incomplete, as we shall see.

Level 01 / Negative Fruit

Level 1 is more focused on how fruit extends a snakebird’s range. The level is also trying to get you to notice how a snakebird moves, as orientation and trajectory can make or break a solution. This is an absolutely essential Snakebird skill but fiendish to master. I remember not getting a feel for it this early.

It is time to revisit our understanding of fruit. While fruit is a positive for most of the game, there are examples where consuming fruit will lead you into the valley of the shadow of snakebird death. In this level, if the snakebird stands up directly and snaps up the fruit, it will become trapped and perish.

Level 02 / Dark Cave

Two lessons here. The obvious one is tight spaces are dangerous because this is the key problem here – how do you get your snakebird into that small space, eat the fruit and get out alive? Like the previous level, eating the fruit causes the snakebird becoming trapped.

I said two lessons. In this level, if the snakebird falls on its face, which is a natural consequence after climbing up to get the fruit, the snakebird can trap itself. This highlights that a snakebird’s body can be just as much a hinderance as any other obstacle.

Level 03 / Bend of Doom

Everyone’s favourite obstacle: the spike. We’re here to learn that snakebirds must remain grounded in the presence of spikes.

A mistake I am sure even veterans make is to bend the snakebird when over a spike. Unless you can get the head supported, this is a countdown to death, for that bend will fall into the spikes once the tail leaves the ground. The snakebird in the above shot is doomed. The bend of doom, I call it.

Level 04 / Spike Valley

Level 4 throws you a red herring right away: the level opens with the snakebird pointing the wrong way, but it won’t be clear until the player has dabbled with the puzzle. Snakebird is now guiding players towards the fundamental tetris drop maneuver. When a snakebird falls, particularly in the vicinity of spikes, it must often fall in a particular shape to solve the puzzle. In this level, the snakebird must fall in a precise shape to make it down from the far left platform onto the middle section safely.

There is yet another subtle fruit lesson. Eating fruit makes the snakebird grow, this effectively means the snakebird’s body does not move. That is, growth is a stationary move. By this point, it’ll be baked into the player’s brain that each move results in actual movement, but this is not true when eating fruit. Without that understanding, eating the fruit at the bottom and surviving will seem impossible.

Level 05 / Tetris Drop

The tetris drop is explicit here. To get the fruit on the right, you must curve the snake to fall safely on top of the spikes. The fruit on the left is impossible to collect unless the snakebird is long enough (bend of doom).

Level 06 / Straight Up

Level 6 is actually a form of tetris drop: the snakebird must fall in a vertical straight line to be able to eat safely. However, I do not find this level satisfying; the confined space means it is not clear whether the snakebird is safe until after several moves which are difficult to predict in your mind’s eye. This feels more dependent on Monte Carlo for success.

Level 07 / Snakebird Squared

The seventh level introduces the wonderful wacky world of multiple snakebirds.

As a result, it’s a bit of trial by fire. This is where my Dead in the Water feelings come to the fore and I am staggered there is not a gentler introduction to snakebird-on-snakebird physics. Without prior experience, it is fiercely difficult to figure out how to shape the two snakebirds so as to help each other. (Notice also this is the first level without fruit.)

Level 08 / Reverse Park

Level 8 is diabolical. It requires the player to grasp a new mechanic – that snakebirds can push each other – and a discover another fundamental maneuver, the reverse park.

The green snakebird is needed to bridge the spikes for the red one, so it can tetris drop safely onto the island in the middle of the spikes. Doing this the normal way means the green snakebird charges out two steps onto the spikes; it will then be impaled on the next move, whether attempting to move forward or turn around.

If only the green snakebird could move backwards… and there is your solution. Use the red fella to push the green snakebird backwards over the spikes so the green snakebird can then safely move left, off the spikes, once the red snakebird has used it to reach the island.

This new maneuver, using a snakebird to reverse park another into position, is hard-earned but crucial.


At the start of my analytical run through Snakebird, I was terrified that I would become as stuck as before and end up abandoning the game. This anxiety transformed my approach into one of deliberate practice as discussed in The Monte Carlo Player.

Gradually, it became clear I had a much stronger understanding of Snakebird this time around and began to relax. There are less concrete lessons to learn once you hit level 9 and levels are more about cultivating an organic approach to each new situation. Where are the problems? What must a solution look like?

Level 11 / Glutton

I’ll call out Level 11 as a design misstep. You must eat the fruit in correct order so that the snakebird does not get trapped. After several attempts, players will realise the cherry is the problem. Because of the position of the cherry, it is not possible to eat all the fruit in a single loop around the cave. However, it should become obvious that the cherry must be eaten on the first loop.

Thing is, this is not an enticing challenge. Neither does it impart a useful skill to the player. To the developer’s credit, this style of puzzle only ever turns up again once. But much, much later.

Now if you thought Snakebird was pretty difficult so far, well… you really had not seen anything yet.

Level 14 / Baby Snakebird

Level 14 is a minimalist nightmare. So small, so impossible. And now we’re confronted with the “baby” two-block snakebird which has substantial mobility and mortality problems – this little thing dies very, very easily. It’s no problem getting one snakebird to the exit – but two? The apparently superfluous platform to the right is a hint to the trick move: reverse parking baby blue onto it solves everything.

And the challenges continue. Level 15 stumped me for a day, but then I hit level 17 and I began wondering, maybe this was the end. This was where it stops.

Level 17 / The Descent

The problem is to find a way to get these snakebirds both down to the bottom alive. Endless snakebird deaths, endless rewinds… this is one of those levels where you must work backwards. What must the solution look like?

It’s impossible to coil the giant snakebird down there, so baby must go first. And… baby has to fall facing up otherwise baby will be stuck facing away from the exit and be dead. Even if you arrive at this realisation, the stench of impossibility remains. If I drape the long green snakebird over the spikes then how on earth is baby blue going to be standing up… wait.

Slowly, the structure assembled itself in my mind, step by step. My thoughts were a Rube Goldberg machine of effect rewinding into cause until I arrived at the full sequence, the impossible solution realised. It’s actually a damn sight longer than I thought and the baby is running back and forth. It’s all reverse parking and tetris drops. And my God it is a beautiful solution.

Now, a move that should be obvious and thus never earns itself a set piece puzzle is something I’ll term the stretch. You’re going to wonder why I bother mentioning it all when I tell you what this magical move is. The stretch: lie a snakebird out flat. Ta da! Thanks for watching, that’s all folks.

The stretch is important because it maximises a snakebird’s horizontal reach and many levels are built around these reflexive stretch moves. The stretch has been key in several levels already: in level 17 the long snakebird needed to stretch across a chasm. But level 19, oh boy, is all stretch.

Level 19 / Elastic Limit

This is the level that probably encouraged me to quit. My reaction upon seeing it was something like: you are joking, mate.

I had no understanding of how this could be done first time around. What shocked me at the time was the distance between the lonely island and the fruit implied the snakebirds having to be stretched out. There was no slack. Jesus, I thought, how the hell am I going to turn them around?

I gave up trying to solve it and reviewed a video walkthrough… which actually made things worse. I’d switch back to the game and couldn’t remember the steps. My brain hadn’t understood exactly what was going on so unless I had the walkthrough out beside the game, I wouldn’t be able to get through it! This was the moment I understood. I understood that I hadn’t understood anything. The beginning of the end.

In 2018, however, there was no mystery at all. It was bloody obvious. If I need snakebirds to stretch out to the right but be able to return safely to the exit on the left… then it’s reverse parking! I solved it in minutes.

I tried it again last week, cold. I successfully solved it commuting between Southwark and Green Park on a crowded Jubilee Line train.

“You have vanquished the past, Joel,” you might now decree. “You have escaped from beneath the grave shadow of Snakebird failure.”

Well. You would be wrong there. You would be wrong because we have not yet discussed level 21.

Level 21 / The Impossible Cherry

I remember losing my mind over this level during the original run, seemingly going around and around in circles with the same moves. Nothing seemed to make any difference. Trying to pick off any of the fruit seemed impossible. I never solved it. I just left the game and never went back.

In 2018, I attacked this level with renewed optimism. I was not the same person I was two years ago. Oh no! I knew about reverse parking and tetris drops and stretches! I can take your punches, Snakebird!

This knowledge and enthusiasm had precisely zero impact because I was unable to read the level at all. I could not see what type of solution I was looking for and eventually my play collapsed into desperate Monte Carlo, praying that inspiration would jump out at me.

Inspiration never struck but Monte Carlo revealed the solution: particularly tricky tetris drops were needed to survive the fall from the upper fruit onto spikes; and once the top fruit were consumed, the snakebird could reach the cherry safely. Although I analysed the level, I never understood the lesson of the level. I did not know what I had missed or what I was supposed to learn.

The fight was still on.

I tried 21 again this week and had the same problem reading the level. I did eventually manage to figure out the difficult tetris drop needed to get the fruit on the upper left, but it felt far from a confident performance.


I was now into all-new territory, levels I had never seen before, and Snakebird and I then fell into a sort of rhythm. I would battle through several levels then hit a challenge in which I would experience life-threatening mental constipation.

Let’s talk about one of these, which had a solution which was so preposterous, it called for a brand new maneuver.

The next set of Snakebird levels introduced frames which were either there to get in the way or facilitate solution… or both. They were your basic Sokoban blocks but I found “frame” was more useful to describe them as they had these thin sections which could pass through solid parts of the environment.

Level 25 / The Scaffold

In the above image, four blocks are held together to form a cross, but only the four blocks interact with the environment. The thin lines can pass through rock and snakebird. I’m not exactly sure how I feel about that, because it made for a weird headspace where you could push frames “through” solid objects. This lead to frustration when frames fell when you least expected because you didn’t have a handle on whether the block parts were adequately supported. I guess all players get there in the end, but it was hard work perverting existing mental models of how Sokoban blocks should work.

And this brings us to brand new manuever I had to invent when solving level 26.

Level 26 / Engine

The snakebirds need to cross a long sea of spikes. And the solution is incredible, preposterous and highly unintuitive. You put the red snakebird in an ‘S’ shape over a block then use the blue snakebird as an “engine” to push them across. It doesn’t break the physics rules of the game, but it feels like it breaks everything in your brain. And I found this move almost by accident!

Level 29 / Stairway to Heaven

Once you get to level 29, you have a much better understand of the frame mechanics. In this level, you have to create a staircase with boxes to get two snakes out – laborious more than anything. There is one trick at the start to lift both boxes up, but it was constantly frustrating how a single wrong move would wreck the entire sequence – and on a mobile screen, wrong moves were my jam. Undo, undo, undo. Snakebird is an intolerant game, every move is vital. And I think this doesn’t help anyone love it.

Level 30 / Now You’re Snakebirding With Portals

After the frame levels, we reach portal mechanics. Oh, I hate portal mechanics in puzzles and I feel they are a cheap way of generating complexity by corrupting the normally Euclidean nature of game space. I loved Portal but I realised I’d already had enough when I quit playing portals-in-2D game Gateways (Smudged Cat Games, 2012).

Still, the portal mechanics in Snakebird have lovely foibles. As soon as a snakebird touches a portal, they will transport to the connecting portal, in exactly the same shape. If their shape and position is incompatible with the destination, they won’t teleport. And after teleport, the snakebird won’t teleport back again while the snakebird maintains contact with the destination portal.

This leads to two types of puzzle: tetris portalling, where you must find the right shape to use the portal, and what you might consider anti-portalling, where you don’t want to trigger the portal because it is stopping your snakebirds getting somewhere. This can be done using a shape that doesn’t fit the portal properly or blocking the portal somehow.

It was slow going, figuring out how portals worked… and then I hit Level 35.

Level 35 / The Echo

This looked absolutely impossible.

The pear at the top requires a relatively simple tetris drop into the portal at the bottom. But to reach the bananas, it appears the snakebird must emerge out of the bottom portal in a reverse ‘C’ shape with the head at the bottom. Like… where the hell can you construct that portal move? This was a real impasse.

Turns out there is only place that particular shape can be formed and if you do that it leads to a double portal drop that lands you in precisely the right spot. Genius clever, but I didn’t know if it was possible to think that deeply into the puzzle. If you focus on the snakebird shape, the puzzle funnels you towards the correct action, but this was another one of those Rube Goldberg machines: push in the right place and it all naturally falls out. I solved it through Monte Carlo and I understood how it worked – but I do not think I could solve this at the time without guesswork.

After the portal levels there were a couple of frame levels before I entered the final snow zone.

Disaster Approaching

At this point, you’ve either got with the program or kicked the game to the kerb. Puzzles now varied between “most vexing” and “you what?”.

Level 40 / Head of a Pin

So here’s a level I decided I never wanted to do ever again. There are quite a few of these “dancing on a head of a pin” levels and they’re all bloody intolerant, one wrong move and it falls to shit. Fine, sue me, I broke my swears rule. Look, these levels are asking for it.

Now I’m not sure I can say I “solved” 40 but rather Monte Carlo’d it. I just kept throwing stuff at the wall to see what stuck, trying to rule out sequences which were obviously dangerous. Eventually, I escaped into… oh man…

Level 41 / Eat Fruit and Die

At some point in the game, the player realises that snakebirds can use fruit as a platform. It’s not really needed on the early levels, but it’s a pleasant revelation to discover this is a thing. Here’s a solid example of where it becomes key. Eat a fruit and your snakebird will fall to a watery death.

Contrary to Level 40, this one really felt like I solved it. I drilled down into the possibilities hard and understood the do’s and don’t’s of each move. And there was this divine tinkle of the penny dropping when I realised the solution had to be asymmetric. An earned victory. Come on, we can do this.

Level 42 stumped me for ages but resulted in punch-the-air euphoria. Level 43 was a “you what” on first sight, but the solution gradually came to me.

Level 44 / Thread the Needle

And then 44.


On the verge of disaster. This was it. The Rome of Snakebird had arrived to assassinate the game. I would report back to my readers that 44 killed me and it was all over. God damn this level.

There was just so much to unpack in the intricately placed details. I had several different stages of discovery. Once I had figured out how to get the bananas on the left, I was baffled how to get the strawberry. Everything I tried failed; I just couldn’t see how to do it – the portal means there’s a Rube Goldbergesque sequencing to the level which I found difficult to predict.

I hit upon a highly specific tetris drop through the portal that required a reverse park to engineer. But then, the problem with fruit, which doesn’t really bother you during most of the game, is that, well, you have to go in mouth first. And reverse parking means the snakebird is going in arse-first. Boy, did it take some reconfiguration to pull this off.

The next problem, after getting both fruit, was saving one of the snakebirds. Yes, it was one of those levels where, whenever you reach that moment of YAHOOOO I’VE GOT IT I AM GENIUS, you open the Matryoshka doll of mindfuck to find another Matryoshka doll of mindfuck. Because even though I figured out how to save that snakebird… neither of them could get to the exit.


The portal I’d been exploiting this entire time to get the fruit was blocking the exit.

Mummy, I want to go home now.

The worst thing about a level like this is that it involves such intricate precision to get to a particular state that pressing restart is painful. Rewinding moves too far can also be painful. And now 44 was really getting on my nerves.

I knew I had to block the portal somehow after getting the second fruit but how? Monte Carlo found nothing and I was wandering in circles. Claustrophobia began to encroach and the commutes were feeling increasingly grim.

Then I had a batshit thought because I had run out of reasonable ones. Perhaps… perhaps it’s possible to do the fruit the other way around? Who knows?

It was just about possible. Just about. And I spotted during the process that the snakebirds could reach the floating portal from the ground. This means instead of having to get through the exit-blocking portal, you can portal into it.

Christ. Oh Christ, my heart.

I did it.

44 is Snakebird in a bottle. If you don’t see the right trick then you, my friend, are destined for your own private torture chamber in Hell. Puzzle-solving as a masochistic act.

Disaster Imminent

Level 45 was challenging but nowhere near as difficult as 44. After that, it was time to take on the six star levels which had been haunting me throughout. The star levels gradually unlock as you make progress through the game. As a result, you can take them on at any point, but they are mainly there to dazzle you like the village ruins in The Witness. Most of these puzzles require a black belt in Snakebird-fu. They can seem utterly impossible if you take them on before that point.

I mean, the state of this.

Level 1* / Crossing the Void

However, this time around I was, indeed, a fully-certified black belt in Snakebird-fu. I knew I was going to have to hang the frame on the spikes at the top and I’d have to swing the frame forward somehow. Oh… oh, of course. We’ve already done this. The snakebird engine move from level 26!

It was not easy and there was a lot of pushing and shoving and rewinding before I got the three snakebirds across. It felt particularly laborious which meant the anxiety level was through the roof. Imagine if this was an intolerant level where squandering one early move meant ten pushes later your snakebirds are out of alignment and there’s no fix without rewinding back the ten moves. But that never happened and every problem I faced was dealt with as it arose, not six months in the past.

As I said, a bit laborious but I was well pleased with myself.

Level 2* was another glutton puzzle like level 11, but much larger, and I hit upon the solution quickly. Level 3* another crossing-the-void-with-a-frame puzzle but the first few moves are largely dictated by the initial configuration; definitely challenging, echoing 44’s Matroyshka shit, and requiring a lot of messy work to achieve victory.

Level 4* / Steeplechase

But I guess I have a problem with the number 4. The Japanese consider 4 to be unlucky because it sounds like death. 44 almost ended me. 4* was here to finish the job.

It looks remarkably structured but this one really tests your mental ability to deal with frames. These dumbbell shapes will fall if both square ends are unsupported so I spent most of my time mouthing “fuck” every time one dropped.

This became another Monte Carlo level because I could never quite get a mental handle on complex frames like this. Sure, I knew to avoid certain silly moves, but when I hit upon the solution it was completely unsatisfying. It did not make me feel smart. It made me feel lucky.

And after I defeated 4* with luck, I dropped my phone with unluck.

And my entire world was destroyed.


On impact, the battery was jolted out of the phone and the screen went black. No biggie, I thought. All of this has happened before and all of this will happen again. I plugged the battery back in, restarted the phone and reached for Snakebird. Huh, I thought it was a bit odd I didn’t get the cloud-themed title screen associated with the star levels. No matter. I wanted to see level 5*. I started the game and up popped the first tutorial level. Huh, okay. Back to the map and–

Fuck. Oh fuck me. Oh you have got to be fucking of fuck fucking fuck kidding – what IN THE ACTUAL FUCK

Deep Breath




The Third Game

So it came to pass on the twenty-fifth day of July in the year of Our Lord, 2018, that I began my third fucking game of Snakebird.

Certain levels were easier on this pass but a lot of came down to recalling the trick without having to probe thoroughly for a solution. It was interesting that a lot of the answers still did not jump out at me. During the third game, the only level I gave up on was level 21, the one involving what felt like unintuitive tetris drops. It took just one week of commuting to reach Level 5* again. While level 5* was tricky, opening with an apparently impossible problem, I was able to solve it in good time. And level 6* was, to my surprise, a piece of cake. That’s it, then? I’m done?

Not so fast, my little cucumbers. Completing all the star levels unlocks one more level. The finale. The big one. The boss.

Level Finale

Do not be fooled, reader. This surprisingly small level has an enormous amount going on – just think about how easily you could lose a snakebird eating those cherries into a dead end. I could take you on a blow-by-blow account of how this level plays out but let’s skip to the end. I had no love for it because it was a laborious labyrinthine Matryoshka level. I worked my balls off for each inch of progress within that confined space – and every problem solved revealed a new problem. And, guess what, you realised you needed to rewind eight moves and set up an entirely different configuration. The three snakes often need to be in exactly the right position at each stage. Again and again, rewind, redo.

Without doubt, clever puzzling but it’s the most Snakebird of levels. It left me with a feeling that I wanted to smash my phone on the floor again.

Nevertheless… nevertheless… I solved it. Snakebird was done. And it left me with the biggest mindfuck of all. You see, I fucking enjoyed it.

Good Luck On The Conclusion, Mate

Whoa, okay. What happened there?

There’s the obvious. We can argue that research for The Ouroboros Sequence had improved my puzzle skills but, please, there’s more going on here. It’s not about the skills, it’s about the attitude.

I reckon I was right about why Snakebird failed for me initially and I had a completely different attitude going in for the second run. I had committed myself to not just solving levels but deliberate practice: I made sure I understood how each level had been solved. As I’ve pointed out in this series, puzzle games fail to encourage review, so this was a new approach for me.

Through Ouroboros, I had developed puzzle mindfulness, the ability to concentrate completely on the constraints and consequences in logic games. Before I had approached them casually and hoped to acquire wisdom through osmosis rather than analysis and deliberation. Some puzzles work well with osmosis, but Snakebird is Sokobanlike which needs your undivided attention.

I do not assert that this is a good thing or a bad thing. Dead in the Water suggested players are being left behind through such indifference – and perhaps the release of Snakebird Primer (Noumenon Games, 2019) is an admission of this. Contrast with Recursed (Portponky, 2016) which develops complexity gradually and I found the rule set fascinating: it sucked me in, rather than feeling I had to work up enthusiasm. And Cosmic Express, as you know, got me through tough times.

Yet, as I articulated in Why We Do This, the experience of forging a brand new mental model in the heat of challenge is a glorious thing. You encode a new puzzle language directly onto your neurons, so that you understand the tetris drop, the stretch, the reverse park, even if you do not have names for them. Once I had committed to Snakebird, I was hooked.

Jelly no Puzzle (Qrostar, 2013) is infamous as one of the hardest puzzle games out there and, naturally, when I was reminded of it in the comments, I took it on, to test my mindful approach. And they are not wrong, you know, Jelly no Puzzle is a real fierce bastard, a thug of a puzzle game, and I found it much harder to identify heuristics compared to Snakebird. My jelly journey was chronicled on Twitter and it took me about a month. As with Snakebird, I did not enjoy all of it, but I did enjoy most of it.

I now understand I need to do the same thing with Stephen’s Sausage Roll (Increpare Games, 2016) which I had become frustrated with to the point of quitting. I should wipe all my progress, start afresh, and work through it mindfully and consistently. No taking off a month here or there. Solve, review, understand, repeat.

And so, we have the answer to why I hated Snakebird in 2016.

It was because I had not yet learned how to love it.

Download my FREE eBook on the collapse of indie game prices an accessible and comprehensive explanation of what has happened to the market.

Sign up for the monthly Electron Dance Newsletter and follow on Twitter!

56 thoughts on “Tail Meets Head

  1. So what you’re saying is, scrub nubs need to git gud?

    Just kidding. As much personal satifcation as I would have gained from the series ending in a stirring fatwa of condemnation against the whole genre, which I could have scribbled down like a doctor’s note to excuse myself from PE (Puzzle Exertion), I’m genuinely glad that this series has led you to a personal breakthrough – or several, as the case may be. And of course it’s also produced a truly awesome body of writing, and game design-rabbit warren-spelunking. From which you came out alive! I can’t stress enough – this is some feat.

    Looking forward to whatever you set your sights on next!

  2. I may not be able to read the whole post right now–looking forward to it!–but another thing about Level 01 is it encourages you to screw yourself, and so to learn its lesson, by making the other piece of fruit the pear. The pear blends into the background a little, so I went straight for the strawberry and got stuck (thus both learning “You can get stuck” and “Here’s the undo button,” which it highlights at that point) before going “Oh yeah, there’s another piece of fruit over there.”

    You also learn the key falling-in-a-way-that-lets-you-hang-off-the-side dynamic after getting the pear.

  3. “At some point in the game, the player realises that snakebirds can use fruit as a platform. It’s not really needed on the early levels, but it’s a pleasant revelation to discover this is a thing.”

    This is the game playing fair with you: On level 3 the snakebird starts on the fruit. It might be difficult to figure the fruit-as-a-platform trick without that, since it’s not obvious from the physics such as they are, but this lets you know about it early on.

    A couple of levels you highlighted were among my favorites. Level 17 because there’s so much trickery, and also because there was this detail about how to turn around at the edge of a platform which I’d used before without mastering but which this level made me really git gud at, because you have to do it about eight times. There’s that nice feeling of “Now I have this tool in my toolbox and can move to a higher level of puzzle-solving”–instead of working things out step by step you can just say something like “I can do this thing with Redbird as long as Greenbird is over there” and boom, Greenbird is over there.

    (This is something that’s represented in Andrew Plotkin’s Hadean Lands, where once you do a thing you can then just type “DO THING” and the game will take all the intermediate steps for you–but in this case it’s happening in your mind, not the mind of the PC.)

    I also particularly liked level 26 because it looks like something like that–you’ve got this long row of spikes, you obviously need to use the block to cross it, surely you’ll have to figure out some complicated sequence that moves the block one square right and then repeat it. But no, once you get your setup you can just hold down the right arrow! And when you think about the rules you’ve learned it makes perfect sense that it works that way.

    Level 35 is just, well, if anyone worked out what happened in advance God bless. I just desperately tried everything and I was like “Well, I’ll work it into this shape and see what happens, maybe that will put me in a position where I can thrash around for new stuff,” and then it solved the level in a way that left me making the Wee-Bey face. Because it makes sense once you see it! But dang.

    I don’t know if it helped that Level 08 is, IIRC, not in the demo. The demo skips some levels which sent me off into a nonlinear set of explorations when got the full game. I wound up learning the reverse park in 20 which buffaloed me for a while, and I first encountered frames in Level 43 I think, which was absolutely a Matryoshka.

    Another thing here is I like Matryoshkas. I see an obvious impossibility, I work out a way to get past it, and then there’s a new obvious impossibility–but I’ve made progress! You crest the mountain to see the bigger mountain, but it’s there. Even if you have to retrace your steps to do the first mountain a better way, that can be fine. The terrible thing is when you realize the first mountain was a dead end and you need to go back the other way (I mentioned the second frame level with the long bird where I went to a lot of trouble to do something that jammed me up completely).

    Got to go now, but maybe I’ll explain *4 later.

  4. Aha! The readers have arrived and waded through 5,000 words which could have been compressed to “git gud” yeah cheers CA. 🙂

    CA – I’d rather stand back from that take. That’s what happened to me but I stand by previous pronouncements that these games don’t support players outside of the hardcore. I’ve been playing Snakebird Primer today and, geeeezuz, I have been ripping through it. I can’t imagine what Snakebird Proper would have been like if I’d played Primer first. It takes time over each special move and even pointed out rare moves I’d forgotten about (there’s an extraction hook move which turns up once or twice).

    I still have ideas for puzzle posts and, if relevant, I might graft them onto Ouroboros as appendices. (I do this occasionally for the Where We Came From series.)

    As for what I’ll play next… I’ve got Stephen’s Sausage Roll to play!!!


    Yes indeed, there are many tiny design nuances which I’ve completely glossed over. That’s not to say I just “forgot” your examples, I don’t think I even notice! I was focused on broader aspects. In fact, I found my notes – which I originally planned to post as a bonus post – turned out to be in error several times. I assumed X was how something was done but when I revisited it, boom, X was optional. It made writing this really slow as I had to play many of the levels *again*.

    On favorite levels: Level 17 was probably a moment when I realised I was starting to have fun. Stepping backwards from final state backwards was amazing. Level 19 was when I realised this wasn’t the same experience at all.

    I find the engine move in Level 26 remarkable. It’s totally an artefact of the crazy rule set which makes it feel so preposterous. I don’t dislike it, per se. It’s a real moment of discovery. OMG I CAN DO THAT??

    I’m not against the Matryoshka. But they are dangerous, because you need to dip backwards and undo the previous progress that can be frustrating. You’re constantly learning your progress is illusory. That’s why I really disliked the final level. But 44, I tend to have positive memories even though at the time I really was quite anxious about it.

  5. Hi Joel,

    I really wanted to thank you for this series, which I’ve been following for quite some time. I’m a puzzle and game designer myself and I’ve actually been working on a book and lecture series on the topic of puzzle architecture and design (whether for games, physical puzzles, crosswords, logic puzzles, escape rooms, mystery hunts, or really anything). So it’s always very useful to read detailed player feedback or gameplay reports concerning their experiences solving puzzles. Your writings are very informative, well-written, and entertaining.

    Puzzle game design is in a weird spot right now. A lot of the best resources on good puzzle design come from somewhat insular communities (crossword creators, chess composers, etc.). These folks have highly evolved standards as far as evaluations of aesthetic, thematic, and logical quality are concerned; they’ve got amazingly developed vocabularies for describing different aspects of a puzzle; and they have editors and publishers that demand high-quality submissions, provide critical feedback, and allow the best creations to rise to the top. On the other hand, puzzle games are still in a maturing phase. Creators and publishers are still figuring out what it takes to satisfy a diverse set of audiences.

    Game designers actually have a special handicap in designing puzzles… they are quite used to motivating and rewarding players through challenge and success—delivering pleasure via the “fiero” or “yatta” emotions—but they often have a hard time prioritizing “eureka”. It’s not difficult to create puzzles that are hard, but difficulty and progression alone are not enough to motivate most players. The most memorable and enjoyable levels in snakebird (and a lot of other games you’ve discussed) tend to be the ones that lead to the most interesting player insights, or that reveal interesting emergent consequences of the rulesets. Think like the fling/double fling in Portal.

    Unfortunately, a lot of levels in puzzle games—even really good ones—sometimes contain no real insights to be extracted. They’re just tree-search grinds. Mazes with a single exit. Challenges of statespace exploration and navigation rather than analytical deduction. You can’t “know the trick” and solve them faster next time. Short of memorizing the solution, you can only get faster at searching and predicting ahead to explore and cut off the dead-ends more quickly.

    To me, it’s a sign of immaturity in the genre that these types of levels continue to be created and shipped in games. Compare, e.g. the best publishers of Japanese logic puzzles, who only print hand-crafted puzzles that can be solved with logic alone. No guessing or monte carlo search required.

    There’s clearly a distinction between analytical puzzle-solving behaviours—like establishing structural lemmas, identifying metaconstraints, pruning the solution space, performing logical deductions, stepping backwards from the solution—and non-analytical behaviours: experimentation, nishio, bifurcation, fiddling, intuitive solves, haystacking, etc.. Puzzles with unique analytical solution have the property that you can *prove* the uniqueness of the solution simply by solving the puzzle in the intended logical way, thereby deducing the one and only solution. But puzzles with a single answer and no analytical solution can only have their uniqueness established by exhaustively cutting off everything that doesn’t work. These are the “I never understood the lesson” levels. There is simply no lesson to understand. Just try everything until you find the one thing that works. Jelly no Puzzle has a lot of these types of levels.

    I love the fact that you invent language to describe the different solution techniques… the reverse park, the tetris drop, and so on. It sorta reminds me of the skewers and forks of chess puzzles, the naked triples and x-wings of soduku, or the glue and mercury of Picross. Giving names to different lemmas, templates, and sequences helps to concretize the abstract and facilitates deeper thinking about mechanics and solutions. I kinda think it’s a shame that puzzle game developers don’t include more language to help players understand the solutions and lessons of the puzzles. Everyone is so obsessed with being minimalist, even to the point of demanding wordless tutorials. But puzzles are about leading solvers to new insights, and words can help with that.

    Really, I think many of the failures of puzzle games to be enjoyable boil down to failure of the designer to lead the player to understanding. You shouldn’t have to play the puzzles a second time with a deliberate, concerted effort to learn the lessons. The lessons should come naturally as part of playing, or they should be tutorialed and tested the same way that important game mechanics are reinforced in other genres of games. It’s not simply a question of difficulty… it’s about facilitating mastery instead of learned helplessness. A lot of games, even really good ones, are terrible at that aspect.

    I feel like puzzle games need a bit of an overhaul in their design thinking… like how Don Norman introduced human-centric thinking into industrial design: it’s not the customer’s fault that they don’t know how to use a product, it’s the product designer’s fault that they didn’t make their design easy to use. Of course, it’s trickier with puzzles, because intransparency and difficulty are part of the challenge! But I think a lot of games are needlessly cruel to the player to the point of being impenetrable by much of their audience. Games should tell you what you were supposed to have learned after you beat each level, when it’s not obvious. They should name the solution steps and templates. They should provide references and guides to the ideas contained in the puzzles. Chess puzzle solutions even contain annotations that show how other (incorrect) approaches can be refuted. All of these ideas help deepen player understanding and appreciation without sacrificing difficulty. And they’re not hard to implement. We just need to improve our thinking as designers.

    Anyways, bit of a long rant but thank you again for your posts.

  6. OK, let me see if I can summarize the secrets of *4 quickly. (And not to ignore your very interesting comment, Elyot!)

    Constraint: Letting a frame fall to the ground is an irreversible failure state. Every frame must always be supported by a snakebird, a spike, or another frame.

    Goal: Getting frames onto the bottom two spikes is fairly easy. Getting a frame onto the top spike is not. The top spike is five squares off the ground, which means that in order to get a frame onto it (from the side) the frame has to be six squares off the ground, and a snakebird’s head has to be six squares off the ground to be able to push it.
    Supporting the frame on a snakebird on the ground only gets it four squares up. The lowermost frame is two squares up, so a snakebird that rests on it (level 3) could support a frame at level 6 if it stood straight up–but because of the horizontal alignment, the topmost frame wouldn’t be where it could be pushed onto the spike. The middle frame is at level 4 and Greenbird can get on top if by climbing the bottom frame, so this gets Greenbird high enough to be able to push the topmost frame in place. But Redbird has no way to elevate a frame to the level where Greenbird can push it.

    Useful strategies: The snakebirds can turn around at the edge of the clouds or on frames without messing things up.
    If you need to reposition one snakebird, you can use the other one to support the frames while you do it.

    Big insight: We need both Greenbird’s head and a frame to be on level 6. The way to do this is to find a way to lift Greenbird and the frame at once. So we put a frame on the middle spike, let the frame that is not on any spike fall to level 2 with Redbird’s head under it, and wedge Greenbird in between those two frames in an L shape.
    So level 1 has Redbird’s head, level 2 has a frame supported by Redbird’s head (also the frame that’s on the lowest spike), level 3 has Greenbird’s tail supported by the lower frame and the middle spike, level 4 has Greenbird’s head and the top frame, which was supported by the middle spike while you assembled this and is now also supported by Greenbird’s tail.
    When Redbird straightens to its full height, it lifts the frame it was supporting to level 4. This lifts Greenbird’s tail to level 5, which lifts the top frame and Greenbird’s head to level 6. Greenbird can now push the frame onto the top spike, and also push the frame from Redbird’s head onto the middle spike.

    The rest is technique (or anyway maybe seeing that you can now shift a frame from one side to the other by stretching out a snakebird and sliding the frame on the top).

    I’m not going to pretend that I reasoned this through when I solved it–there was a lot of thrashing around, including the discovery that you can get both snakebirds on the same side–but once I realized what I’d done, I could make sense of it.

    …now if you have an explanation of how you pulled of *2 so quickly, I’m all ears. I wonder if all the qrth-phyl time you put in helped you here. I never did much Snake (I did play some of Snakes on a Cartesian Plane).

  7. Level 21 makes use of a pretty interesting technique that I believe never gets used again, at least not to it’s full potential: The Double Tetris Drop.

    You can’t collect the right fruit directly from the platform and make it back without falling, however by moving one tile up while on the platform you can make it to the fruit and then fall just a single tile, and now you can make it back to the platform. That one tile you fell made it act as if the fruit you were going for was a tile lower, as your body that was in the fruit’s tile is now in the tile below.

    The left fruit its the same distance away, and the one extra length isn’t enough to reach an extra tile, so you’ll have to do the same double-drop to reach it. Unfortunately the shape you have to make to survive the fall forces you to block off your route back. The route opens when you fall the first time, but you’d have to waste moves for that first fall to happen.

    The trick here is to form a backwards C on the platform, so that the two drops are 3 steps away instead of 1. This lets you activate the first drop in the crucial point between collecting the fruit and returning home without needing to waste any moves.

    I understand why this isn’t used again: first it requires a fairly long bird to set up two separate drops like this, and second it requires this specific combination of spikes and ceiling for the timing of the first drop to matter.

    (I guess the right setup with two adjacent drops is technically a double drop, and I’m pretty sure that’s used a bunch, but I’m referring to the left setup as the unique technique here. Maybe “delayed double tetris drop” would be better but now the names are getting silly)

  8. Great read Joel!

    I think you touch on an important aspect here, which is that the players do have to express a certain amount of effort in order to keep up with the game, and thus have the best experience. I’m not really sure where the line is drawn from this being a designer-problem to a player-issue.

    For example, having a story that is complex and hard to follow might have the blame rightfully put on the writer when a reader gets lost. But let’s say the story is simple and very easy to understand, yet a certain subset of readers still get lost. Is the blame on them? Their reading comprehension? Their interests? Did they fall asleep on the ride home and use the book as a pillow? Where is that line drawn?

    Personally, I’ve always considered any form of medium a conversation. It doesn’t really matter that the author of the medium isn’t there in front of you. And like any good conversation, there needs to be effort on both parties to listen and understand each other.

    Perhaps the first time you played through Snakebird you weren’t really vested in the conversation it was trying to have. Sure, you were *there* but you weren’t *really* listening… Snakebird was telling you about it’s day and you were just nodding while daydreaming about drawing lines in The Witness, or thinking about porting your old game to PuzzleScript, when suddenly it asks you a question and you quickly sit up, blink, and say “uh… sorry, what? Yeah, that one I guess..?”

    “No!” Snakebird snaps back at you, crossing its arms. “Of course not that one! Haven’t you been paying attention at all?” And then guilt begins to wash over you… a negative feeling. You look to the door longingly. Perhaps it’s best to walk away now and deal with making amends later. Yeah, let the dust settle.

    I’ve forgotten what I was writing about. Anyway, really enjoyed your films on The Witness and No Man’s Sky! And have plenty of your writing to catch up on too. I’m glad you and Snakebird made friends in the end. Cheers!

  9. An apt finale for a great series. Thanks for it 🙂
    For what it’s worth, this and Andromeda 14’s Cosmic Express meets house purchase were my highlights.

  10. Right, first proper response!

    Welcome to the comments, Elyot

    With the sentence “experimentation, nishio, bifurcation, fiddling, intuitive solves, haystacking, etc” I feel out of my depth already. This is the kind of knowledge I wished I had when I started out on Ouroboros.

    I think that’s great insight where you identify the difference the problem of tree-search grinds vs deduction. I’ve been circling that issue for a while. I like levels where I feel smart and I’ve figured things out. That doesn’t mean I’m negative on tree-search levels; having used smarts to eliminate vast swathes of false solutions and leave with a smaller tree is amazing.

    I quite like the Minesweeper-like Tametsi but it’s clearly about scanning for combinations of logical rules that result in certainty; it’s the “intellectual” equivalent of a wordsearch. Tametsi is good at forcing you to look for more complex combinations but at the same time I find precious few insights that accelerate your descent through the levels. Snakebird has tools which I can reuse; Tametsi has few. But people like Wordsearch. Minesweeper – and all it’s offshoots – is a pleasant a time killer. It’s not about proving smarts.

    There’s an interesting sweet spot between analysis and tree-search, where some experimentation is required to feel how the level operates. It’s difficult to see what the problem is until you start humping stuff around like in Jelly no Puzzle. The level looks straightforward, so where is the problem? But you’re right that Jelly no Puzzle has a lot fewer tools to offer than Snakebird; there are a few handy rubrics but generally it’s always about the move you didn’t think of, rather than the right tool for the job.

    And perhaps this is the heart of how some bounce off puzzles, because they feel they aren’t learning enough to take on challenges. The main thrust of Ouroboros is that puzzles aren’t doing enough to help players craft the right tools. But there’s also the other, much more serious accusation: that the tools aren’t even there in some designs.

    This one is going to keep me up at night.

  11. Yeah, on that point… I’ve learned a lot about this “are the tools even there” question while interacting with other puzzle designers (especially on Discord servers like puzzler’s club and draknek’s thinky puzzle games). The answer depends super heavily on the designer and the puzzle type.

    A lot of designers of puzzlescript games or other types of stateful puzzle games (block-pushing and similar things) just make levels by putting some game objects on the board in one formation or another, seeing if it’s possible to win, and playing around until they find something hard but not impossible (and hopefully interesting). Then they sort the levels by difficulty and ship the game. This is very different than the “design-by-solution” or “forward design” methods that a lot of paper logic puzzle designers use where they first start by thinking of steps or logical break-ins that they want the player to make, and then proceed by engineering clues to facilitate those break-ins, adding more clues and restrictions as they go until they are left with a puzzle that has a provably unique solution. Puzzles that are fully forward-designed can always be solved without guessing, while puzzles in the former category don’t have that guarantee.

    I think the majority of puzzle games on your list are the types that lie in the middle. The designers do try to include a core idea in each puzzle, but there’s often some fiddliness required to craft a challenge that implements the idea, requires it for a solution, but simultaneously isn’t trivial. This is a hard challenge, and we don’t even have really good language to express which games succeed at it (I think in general most of the highly regarded games like Snakebird, SSR, Pipe Push Paradise, Recursed, etc. are good at this).

    I agree that Tametsi, Hexcells, and similar games can seem a bit dry and shallow, mostly because of the nature of those puzzle types. There are way way better puzzle types out there in terms of stateless logic puzzles and WPC-style puzzles, many by absolutely amazing designers (but it seems like a lot of video game folks don’t really know about them). I know you already get enough recommendations but if I could suggest one thing to try, my go-to recommendation is Palmer Mebane’s Statue Park pack which I would easily include in my top 5 puzzle games of all time even though it’s only a pdf file: (you can still play it on the subway with any pdf annotating app). Palmer is the inventor of Statue Park and he’s a freakin genius, if you like Tametsi then it will blow your mind. 🙂

  12. Matt

    I’m going to have to come back to your solution of 4* a little bit later. Right now I can’t face going back into that level. I’m afraid I do not have particular insight into 2*; it just felt natural to try and keep the out of its own path for as long as possible and I solved it pretty much first try on Game Two; took a little bit more time on Game Three. Not sure why I don’t have too much of a problem with it but, saying that, I don’t really like the level. I don’t think qrth-phyl would have helped as that game is about panic 🙂


    Hello and welcome! Ah, Level 21. Much like Matt’s attempt to explain Level 4* above, I found that I was floundering a bit when trying to parse your explanation of the solution. I’ve just been fighting through it again and with my more advanced Snakebird head, find that I can extract a solution with a little bit of jiggling around but it feels like clever guesswork.

    But I think I can see where you’re coming from.

    The right fruit means you need to just have to assume the right shape and the step in the spikes means you can afford to “drop” a whole level when getting the snakebird back to the safety. The left fruit doesn’t give you that step which means you cannot afford to drop a level. The snakebird will just die.

    So I can see the snakebird ends up “dropping” twice and this is why I had a problem. Two drops means it is particularly difficult to predict the solution; I had the same problem with Level 35 with the double portal drop. While it became possible to sound out a solution in the end, I never felt I reached it through intuition or reason; I just found it. Were you able to “predict” your moves or was it, like me, a sort of inspired guesswork after lots of pushing and shoving?

    As Level 35 is also double drop puzzle, we might consider Level 35 the spiritual sequel to Level 21. My guess is this type of puzzle is very difficult for many. Perhaps if this technique had been fleshed out more, these double drop puzzles would become easier. But when I completed Level 21 for the first time, I couldn’t fathom what I had actually done.


    Hi! I did try to expand on this tension between “player problem? designer problem?” in a Twitter thread recently. In a nutshell, while there are players who can rise to the challenge, should we restrict a game’s player base to a handful of superbrains? From an artistic point of view, it probably doesn’t matter. It is what it is. But from a commercial point of view, would the players who bought Snakebird after a Foddy recommendation and then felt burned even bother with another puzzle game from Noumenon?

    Thanks for enjoying some of my films! I probably should make another one soon…


    Thanks! I’m still surprised that Andromeda 14 hit such a chord with some of you.

    Elyot, again

    I can’t really add much to what you’ve said here. Like, yes, yes, I agree, yes. To make a game which is a presentation of “tools” and then puzzles that rely on careful use of those tools without seeming trivial or unfair – Christ, let’s just make 2D shooter, okay?

    I also want to be clear I don’t necessarily mean to imply the Minesweeper class is dry and shallow: while they’re on the zen side of the puzzle equation, the player still needs to be smart in these games because solutions simply do not bubble up from pure guesswork (unless you’re cheating). However, within the context of tool-based puzzle design, they tend to be simplistic and relying on the player as a smart search algorithm.

    Your comments at the end of the series here have really got me thinking. Got me thinking that I missed out a chapter of Ouroboros about tools 🙂

    Can you really do those Statue Park puzzles on a screen without pen and paper…?

  13. I see many cases of “idea pareidolia” in games (in both players and designers): situations where there is really no “big idea” required (or in some cases, not even present) in a puzzle, but the designer or the player believe there is.

    Some designers can make a level thinking that it requires a certain idea to be decoded and enjoyed by the player, but the player can solve the level without even imagining the idea nor internalizing it as a tool to solve other levels. It happens even in carefully designed games like The Witness.

    On the other hand, levels can be produced without the guidance of high-level ideas, and even so some players can extract general principles and tools that help them to solve the next ones. I myself have observed that situation in my procedural games, with testers or streamers.

    Obviously I enjoy those situations and they are the main reason I keep using procgen in my games. Indeed one of the parameters I use when refining my level curators is the number of “idea-sensations” that they produce when I try to solve the levels.

  14. Marcos: A nice example of what I think you’re talking about (or is it pareidolia?) is the final level of Top Hero Arena. It’s a devilishly difficult puzzle whose solution, at least for me, involved seeing through what looks like a particularly clever bit of Schmuck Bait. It’s a nice little game, I recommend playing it through before revealing the final spoiler (I just used some walkthroughs from the JayIsGames thread to get to the final level though):
    Gur nynez cneebg, juvpu ernpgvingrf nyy gur gencf/perngherf, frrzf yvxr fbzrguvat lbh fubhyq chg va yngr gb znkvzvmr gur gbgny nzbhag bs qnzntr qrnyg va n eha. Ohg va snpg lbh unir gb chg vg va irel rneyl ng n cbvag jurer vg jvyy whfg or ernpgvingvat fbzrguvat gung qrnyf n fvatyr cbvag bs qnzntr, orpnhfr ernpgvingvat gung pybfrf bss n cngu gung zrnaf gur ureb zhfg jnfgr n urny juvyr ergevrivat n fhvg bs nezbe. Fb gur ebyr vf abg gb znkvzvmr qnzntr, ohg gb phg qbja gur rssvpvrapl bs gur bccbarag’f urnyvat fb lbh qba’g arrq gb znkvzvmr qnzntr. Naq guvf vfa’g fbzrguvat gung’f vzzrqvngryl boivbhf rira bapr lbh’ir tbg gur fbyhgvba; V unq gb qb fbzr rkcrevzragvat nsgre fbyivat gur yriry gb pbzr gb ernyvmr ubj vg jbexrq.

    OK, so that’s a clever little trap that force you to d something counterintuitive. But it wasn’t laid intentionally at all. In some notes concerning the game, which now seem to be lost off the internet (JayIsGames is the only place I can find this game), the designer mentioned that the last level was computer-generated as a reward for the winner of some contest (that’s who “Jan” is) to be a level with that amount of complexity that had a unique solution. There was no contest of wits with the designer at all–just cold circuitry generating a challenge whose difficulty can be put in strictly mathematical terms

  15. > Can you really do those Statue Park puzzles on a screen without pen and paper…?

    I just use ms paint but if you’re on mobile then there are other free options (pdf reader?) that let you annotate (i.e. draw on) the pdf. You’ll definitely want one that has an eraser. It might help to have a stylus. I know a lot of people still prefer to print the puzzles out and do them on paper but it’s definitely not necessary.

    If you like handcrafted minesweeper-type puzzles then I think the minesweeper chapter in The Art of Puzzles 2 is the best collection there is (disclaimer: I authored a few of the puzzles in the book!)

    Thanks for the twitter mention btw!

  16. (Sorry, above comment came out as anonymous but it was me, I think because I typed in my work email instead of the one that gravatar knows about.)

    Marcos: I’ve experienced so many examples of this! But I think “pareidolia” is sorta the wrong word. Puzzle creators are not perfect logical machines, and they often miss deductions that their players spot. I remember once creating a polyomino tiling puzzle that I thought had a really interesting and fascinating break-in. I gave it to some test-solvers and two of them found completely different break-ins. So now I know of 3 completely distinct logical paths that lead to the single solution. One of them was actually even more fascinating than my original intended break-in, and the solver told me he thought it was the best puzzle I had ever created and the most interesting step he had ever seen in a puzzle of that type! Before giving it to him, I had no idea that the break-in he found even existed!!

    For reference, that puzzle is here: (the goal is to perfectly tile the shape on top with copies of the two tetrominoes below it, with flips and rotations allowed). Warning: it has a single unique solution, and it’s hard!

  17. Perhaps a better word instead of pareidolia would be “paraschedia” (from the word for design). Surely a Greek speaker can enhance that.

  18. (I’m Tim)

    re: level 21

    on both sides of the level you have to drop a level to get the snakebird back to safety

    on both sides of the level you end up dropping twice

    the only difference between the two is a matter of timing. the lack of a step in the spikes forces your shape to be higher, which traps you between your body and the ceiling; by separating the drops (making the first one earlier) you get that dropping a level exactly when you need it.

    I can’t remember if I found it by accident or not my first time, but I know I knew what I did because I only played it once and when I saw the level in this article I went “oh, that’s the double drop level”

    I also don’t think this level is anything like 35; they both involve something you could call a double drop, but the techniques aren’t related at all.

  19. Hi Tim! I’d defend my suggestion of commonality between the two double drops not because of technique but because they involve a consequence that is hard to predict, where a second drop occurs. This was also present in level 2 which I went through in the stream: there are two drops involved in the process of exiting the cave. This makes it hard to “see the future” and turns the puzzle more towards Monte Carlo where it’s difficult to learn a particular lesson. That is, I was more focused on the Rube Goldbergness of a solution that contains two drops rather than the type of technique – although I was mistaken in suggesting there was such a thing in both levels (I wasn’t wedded to this idea as you can see with my meek declaration of a “spiritual” sequel).

  20. I can’t believe the tiling puzzle is more impressive than the Protoplasm story. 😛 I tried to solve it but I’m bad at analytical puzzles :/ Tell me if this is one of the break-ins?

    There’s a lot of levels in puzzle games where I feel like they only have *one* idea. Right now I’m playing through Baba Is You and I’m quite disappointed every time I solve a level and the idea is one where I’m just like “Well, I was *waiting* for the game to do that…”.

    I’m not the biggest fan of Recursed. Based on its level design, I was able to make a lot of ‘tools’, but I think the game just got really tedious. The “green chest chain” is reused over so many levels and it takes a long time every time. I’ve only seen two games that don’t have the complexity balloon in size as the size of the ideas goes up: Jahooma’s LogicBox and Euclidea, and those have specific game mechanics that support higher level ideas. It’s why I actually prefer some games with a maximum level size, because at least there’s the guarantee the designer won’t design something massive and complicated.

  21. Hi main_gi! Now you’ve got me having a go at Elyot’s puzzle. I’ve only made a little progress so far…

    It’s interesting that you get that feeling of “agoraphobia” from Recursed. I realised much later that I liked Recursed a lot because it never usually had many elements in a level (although that had nothing to do with the actual sequence length of a solution). Some of the bonus levels in Pipe Push Paradise are actually big but carry few pipes, which I really like, because those puzzles feel “smart” rather than cavernous.

  22. One thing that I think merits more discussion is is the number of puzzles you solve as you play. You mentioned that you got good at Disassembler when you were playing the daily puzzles—6 puzzles you could beat every day. Whereas SnakeBird has what, 52 puzzles in the entire game? 9 days of Disassembler and you’d have more than that. Solving levels teaches people techniques, but people don’t always learn them immediately. You can play a level repeatedly, taking notes on what works and what doesn’t—or you could just play a bunch of levels that teach you the same ideas, and hopefully people will eventually get it.

    Puzzle designers seem to want every puzzle to have a unique idea, but as puzzle player, I don’t necessarily want that. I know you bounced off of Trainyard, but the slow difficulty ramp helped me learn the ins and outs of the game, so that when I got to harder levels, I felt more prepared than I would have with a short, steep curve. I didn’t just know the techniques; I had used them all repeatedly, and seen subtleties that having only done a thing once, I wouldn’t be aware of.

    Or thinking about games like Sudoku or Picross—games where there’s a rich terminology of solution techniques. Would those techniques have been developed or named if there were only 10, 20, 50 puzzles of each type? It seems like they came out of the experience of solving dozens or hundreds of puzzles. Even vanilla Sokoban has had tens of thousands of puzzles created, giving players a rich bed of experiences to draw from if they’re interested.

    And I have no evidence of this, but I wonder if you get nearly as much from not solving a puzzle, as you do from solving it. If you spend a week on a puzzle, and finally solve it, what have you learned? You learned the technique that the puzzle requires (hopefully), and you learned a lot of things that don’t work. I’m curious how much that helps—how much better you get at the puzzle from all of those false starts. You come up with techniques to try, but then they don’t work; will that help you solve a puzzle where it would work? Would it help you identify a situation where it would help, or make it easier to identify a situation where it wouldn’t?
    Again, going back to your experience with Disassembly, you were fed a steady diet of successes; do you think you would have gotten as much out of fighting against 20 or 30 puzzles that each took hours?

  23. Hello Josef! I think this dials back to Dead in the Water, which I feel is the most important episode of the series. It’s there I grapple with the problem of efficient design, each puzzle crafted perfectly as a stepping stone to another, with no overlap.

    I think this is not a problem for the experienced hand, so may be good at picking up lessons or aware that review is crucial to progress. But for the average player I think it creates a passive aggressive git gud barrier. Like I said in the Snakebird stream recently, I think Primer would be better integrated into Snakebird somehow – a slow mode with more levels?

    I think procedural generation can solve some of this problem of “excess” levels (see Dissembler, of course) but there’s also something to be said for the hand crafted level. Maybe a halfway house is advisable?

    On the duration comparison, I tend towards the theory that the time spent on a puzzle is only valuable when the player is engaged and not just pushing things around randomly. I don’t mind a little bit of random exploration but a week of Monte Carlo (see my Nova 7 story) can easily destroy enthusiasm and focus. You’re not learning, you’re throwing dice.

    Thanks for stopping by!

  24. Jesus, Matt, that actually sounded like your final comment on Electron Dance.

    Truth is I’m wondering whether to respond to the comment in an article rather than a comment.

    But tomorrow is Rezzed which definitely means no regular stuff for a week or two.

  25. No no that was “fine, I’m going to start leaving the roguelike comments.” I would never leave!

  26. Hey there 😛

    I prioritize puzzle games with a low “sequence length” nowadays. I think my most valued design principle is just ‘stop wasting time’, so I really don’t like games that show the same concepts over and over either. I’m actually quite ambivalent about time spent in the sort of block-pushing grid game, because I feel like so much time is spent walking. That’s a bit of a tangent though.

    I’m at the point where most puzzle games force me to go through much of the “Boredom end” as you call it. However, I remember being such a bad puzzle game player I practically wasn’t one (looking up walkthroughs and feeling satisfied for cheating). I can’t really remember the point where I went from cheaty non-player to one willing to take up the challenge and become really good at it.

    Which is to say many players have this ‘lower skill’ modifier and you could say a bunch of easy levels showing the same concept is actually justified, because then they ‘get it’. And giving the worse players something to do is still kind of like a ‘git gud’ barrier, it just didn’t jump them to the next level.

    I don’t know how well LogicBox succeeded at this because the early third of the game is kind of slow to me, and it only really ramps up near the middle with many difficult levels that all use a cool different trick; instead of tutorials you’re actually making the tools yourself. At that point, if there was a new level that repeated a previous concept, the solution would… just be an instance of the previous level and be really short. This doesn’t quite fit for all the techniques but it reduces a lot of filler at least.

    It’s worth noting that some of Witness’s panel puzzles are just as ‘drawn out’ (pun) – at some points I got somewhat bored of how many puzzles there are just as a tutorial to one mechanic (like the quarry shack) that didn’t add something different.

    Oh, and speaking of Snakebird, I wonder if they’ll do a Snakebird-but-harder. I think it was a somewhat strange development decision to go release an *easier* game when part of the word of mouth marketing was “Looks like a kid’s game, but gwahahaha it’s hard!”. So I was sort of expecting Snakebird Primer to troll the player and it to be Snakebird-but-harder in reality.

  27. Hi Joel. I love the Ouroboros series. It was also very nice to see a fellow human evolve in a meaningful way, like you did while writing this series. Especially this last article, resonated with me very strongly.

    It is common for humans faced with difficult tasks to quickly dismiss them as boring and go on with other, less difficult, or more familiar tasks — I’m not sure what the evolutionary basis for this is, if there is any. Your article, clearly demonstrates, that struggling, and working hard to overcome difficulties can, potentially, lead to gaining deep knowledge or expertise, or being rewarded in other ways. As is well known, “There is no such thing as a free lunch”; you have to work hard!

    What I have been struggling with lately is, when do you actually know something is worth pursuing. When do you know something is actually bullshit and you should just give up, when have you struggled enough? Is the result even important, or is the struggle itself what is transforming you and making you a better, improved human being? It feels a bit as if we are treading a landscape and we are driven by a hill-climbing algorithm. We cannot see what is at the top nor around the hill.

    I guess these questions are also at the core of the struggle between competing philosophies. To live in the moment or to carefully plan the future. To take risks, or be risk averse.

  28. Hi Nick – glad you enjoyed Ouroboros and done interesting thoughts. Part of me is attracted to the suggestion that much of our life had been simplified (overly so) and this had pushed us towards simplifying more and more and then spending the gains on doing more instead of enjoying the extra mental space these gains could have provided.

    However, this is a simplification in itself – the lot of many in the development world was harder a hundred years ago and people weren’t necessarily inclined towards learning and challenging themselves in these ways when daily life was challenge enough thank you very much. So… shrug.

    I think the issue you’re talking about towards the end of your comment is faith. Faith that there’s value in persistence. It might turn out to be false in any individual instance but we must have faith that seeing things through – on average – will reward us. It will easily pay for the times we’ve made the gamble and lost.

  29. This is especially true in the age of smartphones and mental snacking. It sure took me a while to realise this.

    Although the lot of many in the development world was indeed harder a hundred years ago, it was so in absolute terms. Happiness is instead relative, and the overal wealth inequality has increased the last 50 years.

    Taking your use of the word ‘average’ literally, I wonder if there’s model we could simulate. If we compare our pursuit of goals to a stochastic gradient descent algorithm, a persistent person could be compared as someone that uses a big batch size, and someone that easily gives up as using a small batch size. The rate of convergence of course depends on the landscape, but it is clear that in most cases a too large or too small batch size are not optimal. Anyways, food for thought 🙂

  30. Just finished Ouroboros and thoroughly enjoyed it. Although actually I skipped most of this last article because my own Snakebird journey stalled on level 17 and I want to give it another go.

    Your comments, Joel, about mindfulness and really absorbing lessons are very useful; I think I’ll give Snakebird another go from the beginning. This is what I always tell myself is just “giving it a good run-up” after some time away, and which worked for me finally this year with The Talos Principle — which I nearly finished a few years ago, then forgot all the lessons of, so I gave it another go from the beginning and finally cracked it. (Although it surely helped that I rocketed through the first three quarters as I presumably half-remembered solutions from before, so I had momentum and confidence when I got to the previously-unseen hard stuff near the end and in Gehenna.)

    In fact reading this series is sending me back to a few unfinished games, and I’m bound to get overwhelmed by them (I’m tempted to start a todo list of puzzle games but I don’t know if that’s taking it too far…). I finally beat level 3 of The Flames, for example — so just need to try again with Enigmash, and Cosmic Express, and the Snowman bonus levels, and Jelly No Puzzle (which has me blocked at two literally impossible levels, one of which is the one you highlighted and claimed was actually solvable which I don’t believe for a second).

    I’m firmly anti-walkthrough when it comes to (my own solving of) puzzles, partly because I feel I need to learn the lessons of each level, and partly because that feeling when getting the insight that leads to a solution is largely why I play them in the first place (with a dash of demonstrating to myself that I’m very clever). I mentally beat myself up if I accidentally read something which gives away an insight or partial solution. (My word, the weeks I spent on two levels in Baba Is You that resisted all attempts to be solved, and I refused to ease my burden with a hint. Thank goodness for the ability to take screenshots and review them on one’s phone in idle moments.)

    And yet I’ve relaxed those rules in other genres, and find I’m happy enough to look up the last few collectables or secrets in a platformer if I’ve stopped having fun (not always though, and I guess it depends on how fair I think the developer is being with me).

    Lots of interesting stuff in the comments here too!


  31. Steve! Thanks for working through the series.

    I’ve been finding some games don’t have an issue with mindful puzzling – the universe of Baba Is You is so broad that I don’t feel that I’m missing out on lessons if I finish quickly. Every puzzle takes the time it takes. On the other hand, I’ve got bogged down in Kine – very hard to internalize the special rule set, so I may have to do a reboot somewhere along the line. Ha ha, Christ, I feel somewhat criminal for sending you back into Jelly no Puzzle. I still have mixed feelings about the game yet immensely proud for conquering it.

    I’m still working on a book version of Ouroboros. It has no obvious name at the moment – it won’t be built around Snakebird in the same way – so I don’t think the name is appropriate. The book will include older essays and newer ones (e.g. an essay dedicated Stephen’s Sausage Roll).

  32. Oh and I should add that, by and large, mastering Snakebird is worth it. What you won’t read in the above article you are skipping is that there are a lot of techniques and methods of observing the levels which only come through deliberate practice.

  33. Another thing about Baba Is You is that in many cases–not all–the main moves are conceptual. So even if it’s possible to klutz your way into a solution, it’s pretty transparent what you did after you did it. I think. The levels that require the most fine spatial manipulation are often the worst, though there were a couple where there was an insight that if you do (elementary Sokoban trick 1) you can accomplish something and that was nice. (Thinking of qrrc cbby and also gevpx qbbe where V gevrq gb qb n ohapu bs fghss gb jbex nebhaq gur qbbe gheavat vagb n jnyy oruvaq zr naq gura ernyvmrq gung gur cbvag jnf gb qrivfr n ehyrfrg gung lbh jbhyq yrnir oruvaq ohg gung jbhyq yrg lbh fhpprrq jvgu n ovg bs gur nsberzragvbarq fcngvny vafvtug. Oh dear it’s rot13 again.) But there are others… well I think I hate shegure svryqf.

    Also what is up with zrna srapr? I can’t even get started with it!

    OK, the real thing about Baba Is You is that it is crack for me and dangerous. Like I hate to see when I actually bought it because I may have spent 10% of the time since then playing. OTOH it beats reading people arguing about the US primaries (this is not an invitation for anyone to talk about the US primaries, I really mean it).

  34. I won’t say that I’m already regretting returning to Jelly No Puzzle as such, but I am reminded why I quit it. Still, I’m making a decent fist of the first few levels (again) in Puzzlescript and am regaining that experience (but when I look at the two I’m stuck on in Android I quickly exhaust all possible moves (I mean, not really I guess but it feels that way) and turn it off again). I’ll give it a week or two before I curse you. (And then try Snakebird again.)

    Lessons learned from Baba’s earlier and middle levels do come back into play later, so I suppose its broadness could be a problem: one might see every level’s trick being unique and marvel at the variety, but then hit a wall a hundred levels later when an earlier trick must be remembered and reused as a basic building block (with the new level’s real trick still undiscovered). With such a richness of interactions and tricks, it’s hard for a player to know which ones need to be internalised and relied upon later, and which were fancy one-offs — so one might remember them all and get overwhelmed, or remember none and be missing that information later.

    The book version sounds interesting — I was pleased to see you’d revisited and beaten SSR, btw. I must have played through it four times — it felt very “pure” and yet it didn’t frustrate me in the same way that ordinary sokoban does.

    Matt: [Shegure Svryqf] is definitely a hard one, and not only that but it actually looks complicated too so it’s hard to know where to start — as well as being hard to know how to finish (in that working back from the end it still seems impossible). I’ll rot13 this observation as it will likely give away the key insight that leads to the solution: [V guvax znal cynlref trg fghpx ol uvggvat na nccnerag snvyher fgngr juvpu vf abg shaqnzragnyyl n snvyher, ohg va rirel yriry hagvy abj unf orra rdhvinyrag gb snvyher. Guvf vf n yriry va juvpu gung nccnerag snvyher pna or birepbzr (jvgubhg fvzcyl uvggvat “haqb”), ohg V guvax cynlref graq abg gb jnvg nebhaq gb frr gung unccra. V qba’g guvax vg’f cbffvoyr gb cevzr cynlref jvgubhg tvivat gur tnzr njnl… V fhccbfr znlor n pelcgvp yriry gvgyr zvtug fhopbafpvbhfyl yrnq bar gb vg.] Or maybe you already know the magic trick and just hate the faffing that surrounds it, which is fair enough.

    As for [Zrna Srapr], I’ll put a hint here since I think you may actually be asking for one (but ignore it if not): [“snpvat” vf abg gur fnzr nf “arkg gb”, naq orvat ba gur fnzr gvyr nf K vf abg gur fnzr nf snpvat K rvgure].

  35. Steve, about Shegure Svryqf (I haven’t revealed the whole hint): erbevrag gur Xrxr Vf Lbh ehyr ubevmbagnyyl, ol univat gur Ononf chfu gur Xrxr gvyr yrsg naq gura qbja gubhtu nf V glcr guvf V thrff V pbhyq nyfb gel chfuvat gur Lbh gvyr evtug gura hc.–bbu, fbzrguvat ryfr V whfg ernyvmrq vf gung V qba’g unir gb qb guvf va gur cynpr vg hygvzngryl arrqf gb tb, orpnhfr V pna nyjnlf hfr n Onon cunynak gb chfu gur guerr gvyrf bapr V unir vg va gur evtug bevragngvba. V nffhzr gur snvyher fgngr lbh’er gnyxvat nobhg vf univat abguvat or lbh?

    There are two things that bug me about this, if my presumption is correct:

    Bar vf whfg gung vg frrzf yvxr vg erdhverf n pbzcyrk frg hc gung gnxrf znal gheaf gb frr vs vg’f tbvat gb fhpprrq, fb gung’f n uvtu pbfg bs rkcrevzragngvba naq vaibyirf n ybg bs pbhagvat bs oynax gvyrf (gubhtu gurer vf na bcgvba gb ghea ba gur tevq). Gur bgure vf gung vg vaibyirf fbzr rfbgrevp zrpunavfzf nobhg ghea beqre gung unira’g orra pevgvpny gvyy abj naq nera’g ernyyl gnhtug VZB; V unq na nggrzcg gung frrzrq gb snvy orpnhfr gurer jrer gjb Ononf va cbfvgvba gb chfu gur Xrxr gvyr ng gur fnzr gvzr naq gur bar V qvqa’g jnag gb zbir jrag svefg. N srj chmmyr tnzrf unir guvf fbeg bs pbeare-pnfr fghss; serk, va Pbfzvp Rkcerff gurer pna or n fvghngvba jurer na nyvra pna whzc vagb rvgure bs gjb obkrf naq vg znggref juvpu bar gurl pubbfr. QEBQ rkcbfrf gurfr zrpunavpf n ovg naq vg riraghnyyl yrnqf gb chmmyrf juvpu qrsrngrq fbzr bs gur uneqpber cynlref:

    But probably I’m overlooking a simpler solution, that has happened before with levels that I thought required some fiddling. Also AFAICT I don’t “need” to solve this puzzle, so I can let it stew.

    There’s something in what you said reminiscent of Stephen’s Sausage Roll here: Va FFE fgvpxvat gur sbex va gur raq bs gur fnhfntr frrzf yvxr n snvyher fgngr orpnhfr vg’f nofbyhgryl varkgevpnoyr. Jura V qvfpbirerq guvf V jnf yvxr “un un org gurer jvyy or n yriry jurer lbh unir gb qb vg naljnl” ohg gung yriry jnf sne rabhtu nybat gung vg gbbx zr n ybat gvzr gb ernyvmr gung abj jnf jura V unq gb qb vg.

    Joel I’m sorry about all the rot13 but when you think about it it’s your fault for mentioning a puzzle game that I’m currently playing. Oh wait, Steve mentioned it first.

    Steve, re your last point: I think I know that but I don’t see how it

    *sees how it*


  36. Hey Matt. Yes, that’s the failure state I was talking about so you’ve spotted the trick to the level. As to your gripes: I agree with the first one, and it really can reduce the fun depending on how long the impasse lasts. It’s especially a problem with games that have an avatar for moving things into place (as opposed to just being able to omnipotently construct a solution). Of course, with sokoban that’s an intrinsic part of the game. I suppose we want a way of saying “I’ve proved my understanding and that I can trivially solve certain micro-puzzles — now let me play with a setup that I assume will be achievable using those micro-puzzles”. A sort of higher-level macro language within the puzzle framework: if you’ve shown that you can (say) reverse three tiles using one actor in a certain space, pull that up from a menu to run it again maybe. (I don’t know, the more I think about that the less playable it seems here, but in some other game it would be interesting — like the way you fight for pennies in a clicker, but later abstract them away and fight for pounds instead. Arrange smaller puzzles early on, and allow them to be abstracted away when encountered later (eg SSR’s ubiquitous 2×2 grills). But then, that takes away some agency and also could too easily mask a puzzle twist where your macro implementation is unsuitable.)

    I will have to rot13 my reply to the other gripe, and Matt, you may want to not read this yet: [univat gb xabj nobhg gur beqre va juvpu frrzvatyl-vqragvpny bowrpgf chfu bgure bowrpgf qbrf vaqrrq fhpx, naq gur tnzr qbrfa’g grnpu lbh vg ng nyy. V erzrzore ernqvat nanylfrf bs fhpu guvatf naq zneiryyvat gung crbcyr unq tbar gb fhpu gebhoyr nf vg fbhaqrq yvxr ab sha ng nyy, naq V jbhyq abg rawbl cynlvat n yriry gung erdhverq vg. Shegure Svryqf qbrf abg erdhver vg. (V’z fghpx ba bar irel yngr yriry juvpu znl be znl abg erdhver haqrefgnaqvat gur vagrenpgvbaf orgjrra gjb fvzvyne ireof — naq hagvy V fbyir vg V jba’g xabj vs vg jnf snve be abg!)]. I’d finished the game until the Switch was updated with one final secret, and now I’m out of practice…

    I will add two more rot13’d items here so that you have the possibility of getting unstuck if you want them. First, a vague observation: [lbh ner svaqvat guvf pbzcyvpngrq orpnhfr lbh ner bire-pbzcyvpngvat vg]. Second, a more obvious hint (but I think the insights you’ve already had are the more important ones): [lbh qba’g arrq gb punatr gur ehyr va dhrfgvba sebz iregvpny gb ubevmbagny].

    Finally, I thought of exactly that SSR level in this context of [vagreanyvfvat na nffhzcgvba gung ubyqf guebhtubhg gur tnzr qhr gb shaqnzragny zrpunavpf, naq gura univat gung nffhzcgvba or vainyvq va bar fcrpvsvp pnfr juvpu vf pehpvny gb fbyivat n chmmyr] (for the level [Pengre], of course). Although unlike you I didn’t figure it out ahead of time so spent a long time being completely stuck (and then utterly delighted when the penny dropped). I’m trying to thing of examples from other games… Talos Principle has it I think, especially in Road To Gehenna: univat yrnearq gung pbaarpgbef pna qrny jvgu bar pbybhe ng n gvzr, ng yrnfg gjb Truraan yriryf erdhver gung bar pbaarpgbe or ubbxrq hc gb rzvggref naq erprcgbef bs obgu pbybhef, jvgu gur rzvggref fryrpgviryl oybpxrq va fbzr jnlf (rt yriry Gur Fjnccre). Naq univat pnershyyl yrnearq gb arire, rire yrg ornzf pebff orpnhfr gurl’yy oybpx rnpu bgure, gurer ner yriryf juvpu erdhver rknpgyl gung oybpxvat (rt Phg Vg Bhg). Although Gehenna then goes on to eke a few more puzzles out of the newly-violated assumption, which makes you feel powerful and clever for a bit longer.

    (Well done on [Zrna Srapr]!)

  37. Hey Steve, about the Baba level in question, I’m probably not going to de-rot13 it for a while, because that’s not a level that’s bottlenecking progress unlike the other one I was stuck on. It’s good to know that I have the right basic idea (I think), someday I’ll block a few hours out to work out the details.

    One comment about it, I guess: vg frrzf n ovg qrprcgvir gung gur erfgneg/haqb cebzcg cbcf hc rira jura lbh unir frg guvatf hc va n jnl gung lbh ner nobhg gb or erfgberq (V purpxrq guvf ba n yriry jurer V fhpprrqrq va chyyvat bss gur gevpx). Ng svefg V gubhtug gung vg bayl unccrarq jura lbh jrer va n ubcryrff fvghngvba, ohg vg’f gung vg’f ba n qrynl, fb vg qbrfa’g unccra vs lbh xrrc gnccvat gur fcnpr one. V thrff gung’f snve.

    I signed up for The Talos Principle when it was free at the EGS, but it turns out the EGS only has the Windows version for some reason. Oh well.

    Figuring the SSR thing out ahead of time didn’t stop me from getting stuck for a long time when I actually had to apply the lesson! It’s pretty clever that there’s such a long delay between the time when you might notice The Thing and the time you have to use it. Any memories you have of it will be wiped out by The Great Tower.

    About the micro-puzzles: Upthread I mention Andrew Plotkin’s text adventure Hadean Lands, where you’re constantly restarting, but once you’ve done a thing you can generally type “DO THING” on a restart and the game will automatically do the thing for you. Like, if you have to solve six puzzles to get to the airlock, when you restart you can type “GO TO AIRLOCK” and the game will respond “(doing thing 1)(doing thing 2)(doing thing 3)… You’re in the airlock,” with all the ingredients that you have to use to solve the puzzles destroyed/retrieved/whatever.

    This is easier to do in a text adventure with its command format. Good practice in Sokolikes, I guess, is that once the player has demonstrated that they can do a certain twelve-move sequence to accomplish something, don’t give them problems that make them do that very twelve-move sequence over and over. I think DROD probably violates this rule a lot.

    Another thing about Hadean Lands is I was stuck for a long time, someone recently posted a very accessible hint file and I found out why I was stuck (I had thought something was a puzzle that wasn’t a puzzle, kind of). Now its been so long since I’ve played it I think I’ll have to restart the whole game. Kind of in line with the theme of the post.

  38. Here’s a Baba is You level that really frosted me off, and it was a bottleneck:

    Cnenqr. Nsgre fbzr gvzr V jbexrq bhg n jnl gb trg n Xrxr zbivat npebff gur ybjre unys fb gung vg jbhyq riraghnyyl perngr n fxhyy vf Xrxr vf fxhyy cnggrea, eha hc gb gur gbc, vagreprcg gur Xrxr/fxhyy naq qrynl vg gvyy gur bgure bar znxrf synt vf jva, eha gb gur synt… naq V pnzr hc bar zbir fubeg. Fcrag cebonoyl ubhef gelvat gb svther bhg vs gurer jnf n jnl gb funir bss n grzcb, frr vs V pbhyq zbir gur synt ol hfr bs n fxhyy orsber fgnegvat, ohg ab. Riraghnyyl V ybbxrq sbe n uvag naq gur fbyhgvba jnf pbzcyrgryl qvssrerag.

    Two reasons I found this unfair and am not ashamed to have used a hint:

    One is related to what Joel described in I hate playing with myself and I guess vague enough that I can leave it spoilered if you don’t know the name–it’s not fun to have to count huge numbers of steps, and to realize that your solution doesn’t work because one number of steps that is impossible to visually estimate is one greater than another number of steps that is impossible to estimate visually. In fact Joel did mention that in the comments: Counting the squares is boring. It’s a common experience to try an elaborate solution and find it’s just one bit off at the end and doesn’t work at all–you’re left with a round peg and a square hole–but it feels much fairer if it’s something you could’ve figured out in principle form the beginning.

    The other thing is that I found the mechanics of this impenetrable in a way that forced some trial and error (even after I’d cheated). Things seem to rely on the deep mechanics of the game in a way that I haven’t internalized even very far in. Nobhg zl nggrzcg ng zl fbyhgvba: Ng gur raq, Onon vf arkg gb gur synt, Xrxr vf arkg gb gur “Onon” grkg, Onon zbirf bagb gur synt, Xrxr chfurf gur grkg, naq lbh trg gur ehzoyr. Jul qba’g lbh jva nf fbba nf Onon zbirf bagb gur synt? V thrff lbh unir gb or va n jva fgngr ng gur raq bs gur ghea, naq gurer ner bgure cynprf gung zvtug nccyl, ohg vg frrzf havaghvgvir. Fvzvyneyl, gur fgngr gung vf arprffnel sbe gur fbyhgvba V tbg jvgu Xrxr naq fxhyy fjnccvat cynprf naq gheavat vagb rnpu bgure vfa’g fbzrguvat gung *boivbhfyl* vf tbvat gb jbex gung jnl orsber lbh frg vg hc. V thrff V fubhyq’ir svtherq vg bhg–gur jubyr yriry jbhyqa’g jbex vs guvatf qvqa’g zbir orsber gurl genafsbezrq–ohg ntnva vg frrzf svqqyl.

  39. Adding, and sorry for the triple post: I think my issue here is that the level has to be designed precisely so the solution I tried doesn’t work. But this isn’t “Your solution doesn’t work because you misunderstood the level,” it’s “Your solution doesn’t work because of this circumstance that’s not foreseeable in advance without an unreasonable amount of calculation.” Where other levels often block unintended solutions but do it with a rule like “wall is wall.” This block felt cheap. Some previous levels allowed a solution and then set up an extra level where that solution was blocked and you had to find a trickier one, and I would’ve enjoyed things much more if that was what was done here–the current level as an extra, the main level with gur synt bar fdhner uvture.

  40. My experience with [Cnenqr] is that I solved it in some way that was apparently unintended. I know that it’s had a lot of reworking since release to distill it down to the trick that the dev wanted the player to use. I too found that level hard to reason about and to solve, and even thinking back I can’t remember exactly how it worked. (I’m pretty sure I’m going to have to get a hint for accessing this final level. Or at least get confirmation or otherwise that the broad steps I’m executing are right. I’m running out of fun.)

    Hadean Lands sounds good, although it’s been years since I played a text adventure — I wonder if I’ll still enjoy it. (Not sure I ever completed one.) Missed your reference before as I was skipping comments that looked like they might have specific Snakebird references in. 🙂

    Finally made some progress in Jelly No Puzzle, on that one level that was mentioned in the previous Ouroboros post (Joel having accidentally seen the trick in a YouTube thumbnail). That was very satisfying. It was well set up to [yrq lbh qbja n pregnva cngu gung frrzrq fb angheny lbh qvqa’g rira ernyvfr lbh jrer orvat yrq qbja vg, fb vg ybbxrq yvxr lbh’q rkunhfgrq rirel cbffvoyr zbir ng gur fgneg naq tbg abjurer. Va snpg vg jnf cebcreyl yvxr n zntvp gevpx, jurer gur erny jbex vf qbar fbzrjurer lbh qba’g rira guvax gb ybbx; ol gur gvzr lbh’er svthevat bhg gur boivbhf chmmyr cneg, lbh’ir nyernql tbar jebat. Guvf vf abg n Ghea 1 Qvpx Zbir gubhtu V guvax, orpnhfr vg’f nyjnlf snveyl boivbhf gung fbzrguvat’f abg evtug. Gur Ghea 1-vfu zbir punatrf rirelguvat, engure guna whfg frggvat hc sbe n fync va gur snpr bar fgrc orsber gur raq].

  41. A thought I meant to add regarding the Turn 1 Not A Dick Move (leading one down a garden path in such a way that the player doesn’t realise it… so probably the Garden Path situation to avoid terminological confusion): I’ve been beta-testing Puddle Knights (which is very good!) and suggested a couple of puzzle layouts to the devs. One was rejected because it did this Garden Path, and that’s not the philosophy they’re following for their puzzles: they didn’t want something which was set up to have the appearance of a long but plausible solution that didn’t quite work, where the player spent a while working through what was presented to them only to fail at the end and have to seek out the clever-clever idea that was hiding. (I think that sort of puzzle has its place but respect the fact that Puddle Knights is not it: their puzzles are less deceptive and cleaner for it.)

  42. Welp I spoiled myself on Yriry Svanyr and it was my fault; I got hung up on something relatively clever I could do that left me in an obviously unwinnable position, and missed the key to the level which was much simpler.

    Ol qbvat n Fbxbona pyrne-gur-guerr-obk-jnyy znarhire ba “oryg vf fuvsg,” trggvat Xrxr vagb cynpr, naq chfuvat gur “chfu” oybpx bire gurer, V jnf noyr gb frg hc “oryg vf chfu” juvpu jbhyq ranoyr zr gb ercbfvgvba gur orygf–ohg gura V unir abguvat gb qb orpnhfr nyy gur “vf” oybpxf ner genccrq. Nyfb vg’f abg urycshy. Naq V zvffrq gur boivbhf vafvtug, juvpu vf gung jura gur orygf ner bss lbh pna serryl ragre naq rkvg gur hccre-evtug nern ng gur gbc.

    I do find that the overall rolling unlock approach where you don’t have to complete every level in an area, you don’t have to complete every area even that much to progress further, etc., while it really helped cut down the frustration early on, may not be leaving me as much patience with the late game bottlenecks. With Stephen’s Sausage Roll you know you have to put in the work on every level.

    I am a little surprised by your description of Cnenqr with the designer tweaking it until you’re forced to use his trick. That seems at odds with the relatively permissive approach of a lot of the game! And there are so many pairs of levels where the first one lets you use a trick and the second tweaks things to block that trick, it’s surprising to me that the designer didn’t do that here when the tweak would’ve been so simple!

    Text adventures! Text adventures are cool! Hadean Lands is biiiig though, but there are many many smaller text adventures that are being made for free and that are much smaller and more completable. I’m going to do some egregious self-promotion and suggest you try my Faithful Companion, which is very short but has some mechanics and approximately three puzzles, but I think you might enjoy it. (I’ve tried to sell some of the other people here on it and they all said “TYPING in COMMANDS?”) Also this is a good list of IF games to start out with.

  43. Just to let you know Steve/Matt that I have been casually reading these comments but not too deeply. I’m still working through Baba Is You – I’m not sure I really love it, and it hasn’t drawn out the obsession that Stephen’s Sausage Roll did. I’m trying to avoid spoilers for now…

  44. Matt: thanks for the IF recommendations. I will endeavour to actually play one, and your short one sounds like a good place to start. (Let’s see if I manage a bit during this lunch break…) I think the last one I tried was The Hobbit on the Spectrum (or did I maybe have a go at Adventure when I finally got a PC?) so I’ll be rusty…

    Joel: Baba obsessed me just as SSR did before it, although I think the sausages got to me at a deeper level. Not just the level of a person that really likes sausages, but it was one of those games where I found myself replaying satisfying solutions in my head when I was going to sleep, just for the enjoyment of it (like thinking of a good song when I can’t hear it). I played completely through SSR about four times, and would likely enjoy a fifth. (Aside: I abbreviate it to SSR and then half-think it’s maybe called Sausage Sausage Revolution, which maybe should have been the real name.)

    With Baba I did think back over satisfying levels, but I haven’t had the urge to replay it yet. (But then, SSR is done and feels like it was released complete, whereas I have one bonus level left since the Switch update of Baba, and there’s a level editor to go so I know there’s a vast number of levels out there for me…)

    SSR was a very thorough exploration of just a few core mechanics; Baba goes very deep in terms of its major mechanic (rebuilding rules) but then has a LOT of sub-mechanics to play with, so perhaps doesn’t feel as “pure” a puzzle experience in that sense? And of course you may change your mind as you get further — from your most recent Baba post on Twitter, I think you still have a way to go and some interesting things to see yet.

  45. Steve! If you’re still here!

    I have, I think, completely finished Baba Is You, though I spoiled myself on a few levels. One thing is that the progression can be a bit weird. I already mentioned that the rolling unlocks may have left me less patient, but also, I failed to make progress so thoroughly on one of the worlds that I went and did a bunch of other stuff before I went back and did it and discovered there was a whole other world left to unlock behind it. Which is fine, but things can become such a mess that it’s hard to keep track of where you are. (Like, according to some sources I may be short a couple of levels, but also apparently the game keeps getting revised so the level count isn’t steady?) Also I wound up spoiling myself on one of those fairly basic levels because the later levels had so completely broken my brain.

    I did have a bit of the fun running out feeling for a while–like I was doing a series of Nova 7s–but then I got my wind back, and I can explain why, in rot13:

    Gur Qrcguf vf xvaq bs onq, VZB. Vg unf svir znva yriryf, gjb bs juvpu ner onfvpnyyl vagresnpr fperjf–pnfgvat na rfcrpvnyyl wnhaqvprq rlr ng Yvivat Ynaqf–naq gura Pehfuref’ erny fbyhgvba vf cerggl vashevngvat. Lbh unir gb qvivar gung n pregnva guvat vf qbnoyr naq sehvgshy, ohg vg gnxrf fb zhpu frghc gung vg’f abg whfg fbzrguvat lbh pna rkcrevzrag jvgu. N uvag fvgr V ybbxrq ng cbvagrq bhg gung vg jnf yvxr n pregnva gevpx hfrq va gur Sberfg yriry, ohg jub’f tbvat gb erzrzore gung? Gurer ner bgure yriryf jurer lbhe cbffvovyvgvrf ner fb erfgevpgrq gung lbh pna’g uryc ohg gel guvatf bhg, naq V jbhyq’ir rawblrq guvf yriry zhpu zber vs vgf gevpx unq svefg orra vagebqhprq va fhpu n frggvat. V’q fnl gur fnzr nobhg Cnenqr–lbh unir gb qb n ybg gb frg hc gur guvat lbh arrq gb qb, naq vg’f abg pyrne gung vg jvyy jbex–naq gura onfvpnyyl gur fnzr guvat vf arprffnel sbe Svercynpr, naq zhpu rnfvre gb qvfpbire gurer, fb vg’q unir orra avpr gb trg Svercynpr svefg.

    Nyfb gb cebterff ba sebz gur Qrcguf lbh unir gb fbyir na rkgen yriry, juvpu frrzf yvxr n ivbyngvba bs gur fbpvny pbagenpg.

    V rawblrq Zrgn zhpu zber, gubhtu. Gur birejbeyq znavchyngvba tbg grqvbhf ng gvzrf ohg gurer jrer zber sha yriryf, naq V tbg zber va gur fjvat bs bireynccvat grkg. Gur svany yriry jnf n ovg bs na bbu yriry ohg vg jnf nyfb snve va gung gurer jnf fb yvggyr gb jbex jvgu gung lbh unq gb fnl “Uzzz, V pna qb guvf… naq jung vs….”

    Joel, with all the grumps, I did sink a huge number of hours into Baba is You with enjoyment. (60, I think.) I do think it’s a bit more scattershot than SSR because as Steve said there are a lot of different mechanics, and sometimes the details of how they work can be obscure. The main moves are usually conceptual and when there’s the occasional fiddly bit of manipulation that can throw me off my game (there was at least one level where I was so excited to learn I could do a basic Sokomanipulation that I didn’t think about whether I should). But this is just an explanation about why it’s not in my all-time top 5 like Sausages.

  46. Congrats, Matt! Final win-count should be 226 I believe, [cyhf guerr obahf beof, ohg lbh pna’g trg gb gur shyy jva pbhag jvgubhg gurz naljnl V guvax]. On your grumbles, [V sbhaq Qrcguf uneq ohg fgvyy sha; V yvxrq gur zrgn nfcrpg gung jnf vagebqhprq gurer, juvpu jnf yvtugre naq yrff vaibyirq guna gur frevbhf zrgn jbex va Zrgn. V nyfb sbhaq Pehfuref vafnaryl uneq (vg jnf gur yriry V fcrag gur frpbaq-zbfg gvzr ba, jvgu bayl Obbol Genc pnhfvat zr zber cnva), ohg V qvq srry V fubhyq unir jbexrq vg bhg fbbare. Lbh’er evtug gung fbzrgvzrf Gur Gevpx vf uneq gb svther bhg ol rkcrevzragvat, naq Onon vf fb oebnq va vgf zrpunavpf gung lbh fbzrgvzrf sbetrg gur cnegvphyne zntvp lbh qvfpbirerq n uhaqerq yriryf ntb. Ohg V jnf sbetvivat bs gung, hfhnyyl chggvat n sbetbggra zrpunavp qbja gb zl bja vzcresrpg haqrefgnaqvat — V jnf npprcgvat bs gur vqrn gung, ng guvf fgntr va gur tnzr, V unq orggre jryy or na rkcreg. Juvpu vf jul gur vqrn bs erivfvgvat Wryyl Ab Chmmyr jnl onpx va gur rneyvre pbzzragf vf fb nccrnyvat, nf vg srryf yvxr gur yngre chmmyrf jvyy or fbyinoyr ol cebcre rkcregf naq V whfg arrq gb yrnea gur evtug guvatf sebz gur rneyvre yriryf gb orpbzr bar…]

    SSR is definitely more focused.

  47. Well I solved the problem of the missing two levels. (Did I mention that I was missing two levels?)

    V arire obgurerq gb *jva* yriryf 9 naq 10 ba gur znva znc.

    For the rest, I do feel like maybe I’m more wiling to blame the designer for my failings.

    V unq n ybg bs sha jvgu Obbol Genc naq rira qvfpbirerq n gevpx sbe gur svefg erny fbyhgvba gung vf V thrff abg arprffnel–lbh pna hfr grkg gb fjrrc gur ynin vagb guerr yvggyr cvyrf orsber oybpxvat vg bss. Ohg vg jnf ragveryl qrcraqrag ba zl univat fcbvyrq zlfrys jvgu gur zrpunavp sebz Pehfuref. V thrff vg frrzf yvxr n pbagenfg jvgu znal zrpunavpf juvpu ner cerfragrq va n yriry jurer gurer’f syng abguvat gb qb hagvy lbh frr gur gevpx, naq gura yngre yriryf gung rkcybvg gur gevpx. Yvxr Gbhe (juvpu V tbg fghpx ba n ybat gvzr) jurer lbh unir gb frg hc n Onon vf Xrxr vf Onon punva gung trgf hfrq va Cnenqr, be Grafr Ngzbfcurer juvpu vagebqhprf k vf abg k juvpu trgf hfrq gjb yriryf yngre va Frphevgl Purpx. Ohg va Pehfuref gurer’f fb zhpu lbh unir gb qb gb trg gb gur gevpx, naq V guvax vg’f cbffvoyr gb ghea bar qvpx zbir lbhefrys bhg bs vg ng gur ortvaavat ol oernxvat “Oryg Vf Fuvsg” gur jebat jnl.

    Then there was a bit where the usual thing worked really well:

    V gubhtug Gur Obk jbexrq terng nf n Ynfg Erny Yriry. Vg vagebqhprq vgf gevpx, yrsg lbh va n cbfvgvba jurer gurer jnfa’g zhpu gb qb orfvqrf gung gevpx, naq znqr lbh guvax yngrenyyl gb svther bhg ubj lbh pbhyq npprff gur gevpx–naq gura vg raqrq, orpnhfr n ohapu bs yriryf gung znqr lbh hfr gung gevpx va n punyyratvat jnl jbhyq’ir orra n avtugzner. V pna’g guvax bs nal chmmyr tnzr gung fb fhpprffshyyl vagebqhprf n arj zrpunavp va gur ynfg yriry. Rira Fgrcura’f Fnhfntr Ebyy, juvpu raqf jvgu n yriry jurer gur punyyratr vf ragveryl ohvyqvat gur cngu gb gur arkg yriry, cerivrjf gung jvgu gur arkg-gb-ynfg yriry; juvpu vf cbffvoyl zl yrnfg snibevgr va gur tnzr, orpnhfr bs gur gevny naq reebe va znxvat fher lbh raq va gur evtug cbfvgvba, naq nyfb orpnhfr VZB vg znef gur chevgl bs Tbq Cvyyne.

  48. Agreed about the last level — it was audacious and a great way to end. And yes, that one SSR level was really annoying and fiddly (which is a shame as the overall concept was clever).

  49. Hey Joel, just wanted to leave a comment to let you know that I’ve just finished meandering through the Ouroboros Sequence over several months, and very much enjoyed your thoughts. As an aspiring puzzle designer I love reading thoughtful criticism and analysis of this type. In fact, reading your blog gets me thinking about how uncommon such good deep games writing really is, and encourages me in the direction of wanting to do some similar writing of my own. Thought it might be nice to know there are still people making use of this stuff all the way out in 2021.

Comments are closed.