This is the nineteenth part of The Ouroboros Sequence, a series on puzzle games.
Back in the day when I was a beautiful youngling, skipping levels in arcade games like Pac-Man or Space Invaders was out of the question. You played from level one, every time. And early puzzle games did not escape the gravity of tradition; you played from puzzle one, every time. At least when I was making Sokobanlike The Citadel, I assigned each level a password so you didn’t have to play through the whole game whenever you wanted to have a go. I still clung to three lives and a time limit, though.
In time, the Age of Player Punishment gave way to the Age of Contemplation. We’ve gone so much further than just skipping levels. Today, unless you’re playing something like a Puzzlescript game, your fancy puzzler is unlikely to force you through a list of non-negotiable challenges. I’ve previously discussed the level select as a tool to review the past; what we’re going to do today is discuss it as a road into the future.
Let’s start out with the simplest approach, no choice at all. There is a sound design principle behind the merciless single track methodology. The developer can be confident about what the player has learned and the level number is a measure of the player’s mental maturity. This makes level sequencing a less complex task.
But it has a problem which is obvious to the modern player: if the player gets stuck, they cannot do anything else. This increases the possibility they will quit entirely either consciously or unconsciously. As a result, it has fallen from favour in these modern times but there are still games that follow this model. Machine at the Heart of the World (Evidently Cube, 2018) refuses access to the next puzzle until the current one is complete. This is because Machine is about deciphering symbols and jumping around would undermine that process. There are plenty of other titles which cajole the player through a linear sequence, such as Linelight (My Dog Zorro, 2017) and Portal (Valve, 2007).
Some games offer token choice. The options are so limited, it is no choice at all. Archaica: The Path of Light (Two Mammoths, 2017) looks like it offers multiple options but in fact is linear for most of the game. It requires the player to follow a strict sequence partly because there just isn’t enough puzzles to provide alternatives in each zone. Adding choice is not just about adding a pretty level select GUI, but making sure you’ve got the levels to make it viable. Note that optional challenges do not constitute genuine choice as they are advanced options for the advanced player.
RYB (FLEB, 2016) is another one that implies choice but is actually a straight line. The Android version of the devilish Jelly no Puzzle (Qrostar, 2013) offers some choice but there are stretches of desert where the player has no alternate route, easily leaving players stranded at an unassailable challenge.
Some titles offer a limited form of choice that I’ll label partitioning. Here levels are grouped into “worlds” in which players are free to explore but they can only advance to a new partition after every available level has been solved. Stephen’s Sausage Roll (Increpare Games, 2016) demands every puzzle to be solved before progressing to the next group of puzzles. This is done to manage the “release” of new information about the sausage mechanics. Cityglitch offers a twist on partitioning – you don’t need to complete every level, you just have to complete enough to open the boss level but you must defeat the boss level to proceed to the next partition.
A superior alternative, from a player perspective, is rolling choice in which a game offers a set of levels to choose from, but each success unlocks further levels. This ensures the player has not just one brick wall to bang their head against but three or four. You can see this implemented in the excellent Minesweeperlike Tametsi (Grip Top Games, 2017), Sokobond (Hazelden & Lee, 2013), Cosmic Express (Hazelden & Davis, 2017) and Tricone Lab (Partickhill Games, 2017).
Let’s go further. The Witness (Thekla Inc, 2016) looks partitioned but is actually a hodge-podge of different structures. There are partitioned areas and single track sequences which act as grooves to roll the player’s brain down, control their understanding. But there are also sections, such as the village ruins in the centre of the island, where the player can choose puzzles at will. There’s a certain Metroidvania feel to this, because these puzzles can only be unlocked with the right key of knowledge. If there was no gating, however, the purposeful chaos of the village would infect the whole game. See also Full Bore (Whole Hog Games, 2013) which is extremely forgiving aside from a few bottlenecks. I tend to think about this as managed freedom. The Witness feels generous. Go anywhere. Solve whatever you find.
But are there any games which just let the player have at it? The closest to this open world ideal I can think of is A Good Snowman Is Hard to Build (Hazelden & Davis, 2015). There is some early gating through a small tutorial section but after that the entire puzzle universe is open to you. The reason this works is because A Good Snowman is a flat game; there are no extra crazy mechanics like ice blocks to uncover. There is no educational imperative to gate the levels. (Update 31 Jan: Matt W cautions me in the comments that this isn’t true, that Snowman gates heavily but I just didn’t notice it. In exchange, he mentioned the online version of Jelly no Puzzle lets you select any level you want.)
The purpose of choice is to give players space to move. When the player is thwarted by a puzzle for too long, they end up in a Groundhog Day of performing the same failed solution attempt again and again hoping a revelation will emerge from the void. Instead of pleasure, the player gets claustrophobia.
But all roads lead to Rome.
No matter how much choice you throw in, the choices inevitably converge to a single level, the final challenge between you and victory. While puzzle games can offer multiple springboards to completion, many only offer a single “end of game boss” puzzle. In fact, a partitioned game will deliver many of these moments, because you cannot progress from a partition without clearing every level and you’ll do the hardest one last. How many people remember being stuck on “The Great Tower” of Stephen’s Sausage Roll? Look, I’m not embarrassed to put my hand up here. Uh, surely someone else can put their hand up as well?
Nova 7 was my Rome.
Although you do not need to solve every level to complete Cosmic Express, you do need to solve the final boss level, Nova 7. I had been bouncing between a few difficult levels but, gradually, they all succumbed… except for Nova 7. It became a terrible curse. Instead of a wonderful morning commute on the train pitting my mighty wits against the fearsome brain of Alan Hazelden, I now faced daily frustration, drawing out failed solution after failed solution each suspiciously resembling the previous failed solution. I had run out of choice. I had run out of road. I took a break for a few days but upon returning the brutal truth quickly snuffed out the rekindled embers of optimism.
After channelling countless hours into this one level, I had made no progress at all. The level made no sense, I saw no order beneath its surface, I couldn’t break it down into smaller problems, Monte Carlo had failed and I understood nothing. Overcome with claustrophobia, I could not do this anymore. A whisker away from finishing, Cosmic Express had broken my mental spine.
Only one option remained. A dark magic that we dare not acknowledge in polite company, spoken about only in whisper.
Cheat.
Next: Hole in My Chest
Note on A Good Snowman
A Good Snowman is a Sokobanlike where you have to roll snow around to build a snowman. The park of the game is effectively open and you almost roam at leisure to tackle the puzzles as you please. It is not that difficult but there was a gauntlet of optional challenges buried in the game.
Without spoiling too much, to solve these challenges, you have to redo many of the puzzles that you had considered done. It turns out there are alternate solutions for many of these snowmen that you might not have realised existed. Some of them are downright evil.
The trouble is you’re often not sure where to move your snowmen to be victorious; perhaps the solution you’re after is actually impossible and you’ve got the wrong target after all? This is why I abandoned the challenge because this was some galaxy-brain level shit. I was not confident about some of my snowman rebuild attempts. I couldn’t face the frustration of trying to square the circle when I was supposed to be circling the circle.
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!
Isn’t Snowman more of a rolling unlock puzzle with pretty generous unlocks? Solving a puzzle opens nearby hedges and I think it takes several solves at least to unlock the whole park.
Also my hand is up. Is there anyone who didn’t get stuck for a long time on The Great Tower? It took me weeks.
The online version of Jelly No Puzzle is absolutely open world; you can select any level at any time. Though that’s more the programmer not wanting to bother with a level select. The Swapper winds up being somewhat open world in that way–you have to solve a certain number of levels to gate new areas, but after a certain point more or less everything has opened up, and there’s definitely no Final Boss Puzzle (or even a Final Boss Puzzle for a given area). The last puzzle I had to solve was one that I’d just rushed past in my eagerness to unlock new things (and that was a bear to get back to, because it was at the other end of a teleporter and teleporter paths aren’t marked on the map).
Then there’s this Free Flow game I played a fair amount because it happened to be on the iPad when my computer was broken, which lets you pick any level. Because it’s casual as heck. If you want to have a complete open world with any sort of texture, you need some sort of overworld movement (as the Swapper has, in fact the big set pieces tend to be about moving between areas of the map rather than the puzzles), so you’re not just picking interchangeable items from a menu.
Ah, Matt, that’s a shame and A Good Snowman. I’ve just reinstalled it on my phone and given it a twirl for the last 15 minutes and you’re absolutely right. There is a boatload of gating. I think I never noticed because I just solved as I progressed. I’m going to have to make an update to the article! Annoying as I was having to play through a number of titles to verify my memory last night but Snowman I felt was a slam dunk.
As if to fire back at you, my memories of The Swapper diverge. I have this strong feeling like you kept having to earn “security orbs” to open locked doors and my memory is of significant gating.
Thanks for the Jelly no Puzzle online example!
I can’t think of many open world puzzle games in terms of the ‘clever, single bearded-designer with a vision’ type. I’ve got quite a few logic puzzle games of the sudoku/picross/slitherlink style that might qualify, though: they usually just dump all N-hundred levels out in a grid and let you pick whatever. Although to really be an open world do you have to have an avatar that traverses it?
Rolling unlocks always feel generous; Art of Balance had a system where you had multiple (1-3) tracks through a world, but rather than unlocking the next world at the end of the current one each was unlocked after a set number of completions globally. It was almost too generous in fact. You could end up in a situation where you unlocked world D before you’d even been in world C. I started feeling a mild tug of anxiety that, ‘hey, maybe I should be checking out those other worlds I unlocked?’ even though I was perfectly happy progressing along the world B track that was right in front of me.
There wasn’t a tremendous amount of difference between one world and the next; the overall difficulty of each successive world was higher but at each track started out pretty easy so there was no strong incentive to stick or twist (unless you were mondo stuck). The most noticeable difference was that the art and music changed.
In the end I found myself strategising my trips through the worlds based on a criteria of aesthetic diversity, parcelling out my attention between worlds so that I wasn’t subjected to long blocks of the same theme, even though I hadn’t noticed any repetition until I’d been provided with the agency of choice. The game introduced the tools to solve a problem I didn’t recognise I had even as I was implementing a self-made solution to it, and none of it had anything to do with the puzzles themselves.
I don’t know if this is a design problem so much as a freak emergence that can happen with certain types of player (slightly mad ones).
Hi CA! I just used the moniker “open world” because it mapped better to the wonderful implementation we see in The Witness and, I thought, A Good Snowman – I still think it blows wide open later, but probably once you’re halfway through? But “open world” really just means you can select any level you want. This is, after all, just about player choice rather than representation of the choice.
I wouldn’t talk down your point about the impact of too much choice. I think arguments for agency and choice in games are rather overblown and sometimes people want to chill out and follow instructions. I don’t find an RPG with armies of statistics particularly restful or zen, although I’m sure some find comfort under the heel of numeric order. So there is an important point to make here about offering too much choice and the impact that has on the player. Without a guide, it can feel overwhelming. When Metroidvania games keep offering you options and you KNOW some of them are eventually dead ends, it can feel frustrating being forced to make a choice.
Full Bore massively opened up – but it saved itself with a map which highlighted secrets and roads untravelled, which was absolutely brilliant. The Witness shepherds you around a small walled tutorial area and then guides you into a tutorial field, effectively, but you can run away as soon as you’re out of the walled area. How many do? Sometimes breaking free of the convention in front of you can be liberating but this kind of liberation can have consequences from importing confusion (the reason designers opt for control) and the paradox of choice (where do I go? what is the “optimum” choice?).
Some of these issues did tug at the back of my mind when writing All Roads but I felt it was too off-piste for the focus of this article – particularly for where I wanted to end up on the final line. It’s definitely a design consideration.
Whatever Deadly Rooms Of Death is
DROD is the same thing as Stephen’s Sausage Roll, I think–at least King Dugan’s Dungeon is. Open world within levels, but you have to clear one level before you can access the next at all. (Excepting DROD’s secret challenges.) There’s some within-level gating I guess and on Level 13 of DROD there’s a lot of stuff where you have to solve one level to be able to do another in non-obvious ways, but so far that seems to be the exception? From reading Carl Muckenhoupt on DROD I get the impression that some of the later games break this pattern a bit, but I think you still have self-contained chunks you have to solve–they’re just formally more than one level, with stairs you have to go up and down? (Looking through his archives, this seems to be something about Journey to Rooted Hold levels 13 through 15.)
Joel, you’re right about the gating in the Swapper. I do remember going through a cluster of similarly-themed puzzles (at one point I had forgotten there were ever crates in the game) to get up to enough orbs to unlock the next area. The thing that distinguishes it from the SSR/DROD model is that often you get enough orbs to get to the next area before you’ve finished everything in one area. What I’m really influenced by is the end when the whole map is open, which is to say I had to hunt everywhere to figure out how to reach the last puzzle I’d skipped, which seems like a quintessential open world experience.
Are there any more games like CA mentioned – ones which don’t require you to bottleneck through one puzzle, but require unlock points? It seems to me like a good solution to some bottleneck issues would be a partitioned level space, but instead of beating a “boss” level to get to the next partition you just need to complete 80% of the puzzles in the partition.
I feel like I need to pay DROD with more sincerity one of these days, just because it has this status attached to it.
Rolling choice is the most common because it is the most generous within the constraints of managing player learning. I tend to see anything with a boss level barring progress as partitioning but if the worlds are more permeaous then it’s rolling choice.
From this perspective, The Swapper is rolling choice because there are multiple routes to progress forward – gather enough orbs and move on. These routes don’t have to be pre-assigned by the developer, a dynamic system has the same impact. Player always has options.
I can’t recall the exact mechanism behind Sokobond but it was always showering me with me levels and choices. Look how many worlds I’ve dipped into in the above screenshot. I think it’s seven unfinished partitions?
And JAMES HAVEN’T YOU GOT THE SPINNORTALITY RELEASE TO BE LOOKING AFTER WHAT ARE YOU DOING HERE
PANICKING AND ALSO WAITING FOR THINGS TO COMPILE
I bring up the Chromatrons way too much here, but I’m gonna do it again because it’s an interesting evolution I think. Chromatron and its sequels all have just a flat list of puzzles, but they have different unlocking rules.
Here’s the level select for all the games in their start state: http://nothings.org/remote/chromatron_level_select.png
At any time there are at least N puzzles unlocked.
In Chromatron 1, N starts at 1. After you solve level 10, it increases to 2. Later 3, then 4 at the halfway mark, and by the end it’s 6. (There’s 50 levels total.) So initially it’s a forced linear sequence, but then later it becomes the simplest form of rolling choice, so you can start skipping some if you get stuck, but not many. I definitely wanted to encourage people to go through linearly so they wouldn’t miss the introduction of new pieces, since there’s a bit of helper text for them on the levels they’re introduced. I believe the fact that levels 11 and 12 unlock when you solve 10 is because level 11 is particularly tricky, and I didn’t want people to get stuck there.
In Chromatron 2, N starts at 5. It grows similarly; by the end it’s 11. I believe I did this because I figured you’d already played all of Chromatron 1 so it could start with more spread.
In Chromatron 3 and 4, N is 25 for the whole game. In part this is because I figured you’d played Chromatron 1 & 2 already, but I also remember thinking “good grief, you paid for the game, why am I not letting you have access to everything in it?”
A weird complication I left out happens at the beginning of Chromatron 2. Although it starts with five levels unlocked, when you solve one of the first five puzzles, it doesn’t unlock anything new (effectively, N shrinks to 4). Onle when you solve a second puzzle, does a new one unlock (N is still 4). And when you solve a third puzzle, two levels unlock, growing N back to 5. I have no idea what the purpose of this was.
There’s an online version of Jelly no Puzzle?
A Good Snowman is functionally the same as Sokobond (completing a level unlocks all adjacent levels), but walking around the open world makes it feel very different. There’s only two chokepoints – the “snowballs remove snow from the ground even if they’re not growing” intro and the “you can push the top snowball off a stack” intro.
Hand up for The Great Tower.
What happened to Matt in The Swapper happened to me too: I missed one level in the first section and I wasn’t able to find it again… so I decided to just reset progress and restart as it was faster than finding a way back! I enjoyed it a lot nonetheless.
Online Jelly no Puzzle: http://avorobey.github.io/jelly/
It is really no frills at all. For instance: no undos! Which can be particularly painful because I find it’s easy to misclick. But I don’t really feel like trying to get the official version running under Wine.
So I guess there are two different kind of rolling choice models here–the Swapper/Chromatron 1 & 2/Cogs (?) model where you have to finish a certain number of puzzles/get a certain number of points to unlock later levels, but there’s no strict order within that; and the Hazelden/Snakebird model where solving a puzzle unlocks adjacent puzzles, and there are usually several different adjacent puzzles, so the map gradually opens until it starts closing again because you’ve finished certain areas of said map. Where Cosmic Express innovates by having the final domain start with many unlocks that point you toward the center boss level, and Snowman mixes that with the SSR thing where the puzzles are on the map.
The puzzles-on-the-map thing really does seem to make a difference for the open world feel and pacing. It’s often nice not to have things end right after the Boss Puzzle but to have a little denouement where you can reflect on your triumph. Which could be an easier puzzle or a screen where you collect a reward or fail to do so… maybe like the castles in Braid… but where that works even better if you’re doing something on a map instead of a level select screen.
ps Alan did you ever see this comment I hope it might amuse you
I think I probably saw that at the time.
Thinky Puzzle Games discord links this: http://jackkutilek.com/puzzlescript/jelly-no-puzzle.html which has undo and more of the levels
Re-reading the thread matt w linked above, this passage by matt about “schmuck bait” got me thinking:
“Another thing about Cosmic Express is that there are plenty of red herrings in the levels, like crosses that you wind up not using. In Stephen’s Sausage Roll this usually doesn’t happen and when it does it feels like a violation of the social contract, but in Cosmic Express it’s just part of it–that the levels sometimes have space in them for symmetry’s sake, and that part of the challenge is figuring out what you need to not use.”
One of the things that happens to me in Jelly no Puzzle, and happens in certain Puzzlescript games, is that you know you’re playing a no-schmuck-bait game and there’s all this extra space on the playfield that it doesn’t seem like you need. I.e. you’re trying to mentally work through how you think you should solve the level, but even before fully working it out you’re like “but that wouldn’t use this open space on the map over here, and that must be there for a reason”.
It feels mentally weird to me–you’re bringing in this meta assumption about authorial intent–and maybe it’s not actually a significant factor in solving it–you still have to work through these flawed ideas anyway to understand their consequences. But it does feel like a thing that happens a lot in SOME games and is worthy of note. (Maybe this was talked about at some point in the Ouroboros Sequence and I just don’t remember.)
I was just solving this Jelly no Puzzle level and I don’t think it’s true to say that there’s no red herrings in that game – that area on the left is not used in the solution.
The original version of the level is even more red-herring-y: https://avorobey.github.io/jelly/?20
You can work out that this part of the level isn’t relevant, and it feels fine to do so, but I don’t think it quite matches what you’re describing. Maybe it’s rare enough and minor enough that this level doesn’t stand out among the others which largely don’t do this? It’s been too long since I played the game to remember.
Currently stuck on level 2 of that Jelly game. I am extremely cross with all of you for linking it!
Level 2 solved! I will reward myself by closing the browser and NEVER PLAYING AGAIN EVER.
…
..
I’m not going to get anything done today am I
Maybe Jelly no Puzzle isn’t as much as I remember, but I gues part of it is about prominent features. Like, if there’s a little weird awkward niche that doesn’t match the natural shape of the rest of the level, it’s probably there for a reason.
The “natural shape” thing is important. In Jelly no Puzzle all the levels are the same size (I think), and in Cosmic Express they’re all the same shape, and that leads to unnecessary space. Which could be worked around by filling in the unneeded parts with walls/obstacles, but that doesn’t seem necessary given the overall aesthetic, and in fact would be a major hint in many CE levels at least.
The real red-herringy thing about about level 20 is the isolated block on the left. It won’t work, and it’s not that hard to work out that it can’t possibly work, but you get so stuck that you want to experiment with things that can’t possibly work. In SSR the things I’m remembering are features like ladders that don’t get used, which seems particularly rude because he had to make a special effort to put them in, but then again in those levels it’s such an effort to figure out what you can do that once I figured that out I wasn’t that tempted to use the extra ladder.
(For reference, the SSR levels I’m thinking of are onolebpxnaqpbyqtngr, rot13ed for spoilage. I may have spent some time trying to deal with the red herring in one of them–also it has multiple solutions though I don’t think either of the ones I know of use the red herring.)
(Sorry CA. I have been thankful that the one Alan linked, which contains MORE LEVELS after I already solved all the ones in the version I’d found, seems not to be working on my browser.)
I’ve been meaning to pop a short response on this thread but I was busy doing other things like saving the universe and completing Jelly no Puzzle.
On Alan’s point on Snowman that “the open world makes it feel very different” I’d have to second that. My memory was overwhelmingly one of unfettered travel which was actually not true.
(Incidentally, Fede, I feel a bit silly for talking about The Great Tower all the time because I *know* there are other levels players get stuck on.)
Matt, I really don’t think you can play some of the later Jelly levels without an undo. You really need to explore the interactions and consequences of certain moves. Without an undo, god that is gonna be painful. (And, additionally, on a mobile device, mis-swiping is dreadfully easy.)
Sean: I’ve referenced somewhere in this series that “obfuscation” through unnecessary noise has fallen out of vogue but it does rear its head here and there. The problem with noise is if the game has already convinced you the structure is always optimal – depending on the circumstances in which this “contract” is broken, it can really hurt.
Jelly no Puzzle initally convinces you that the levels are like this – but there are definitely some which have noise/red herrings in there, parts of the level that do not matter. (I also got confused on level 50 in the Android version, a seriously tough gauntlet, because part of the space looked like I was meant to deposit a piece… but I misread. No, no, no, it was an an obstacle!)
(And another sorry to CA from over here. Somehow, Jelly is a real fierce beast, such complexity from such small levels!)
I’m the Tametsi developer; I appreciate that you’ve given my game some thought, and I’m really happy that you have a positive impression of it.
Tametsi’s level select method is one of the first decisions I made while developing the game, and it never really changed. If I could amend anything about it, I would try to find a way to encourage players to take things a little more out of order — save the more daunting puzzles for a day when you’re feeling especially sharp, that sort of thing. I ironed out a lot of the difficulty spikes during testing, but (for example) puzzles 64 and 65 are really rough to tackle back to back, and a player who feels compelled to do those two next could end up very frustrated.
This series on puzzle games is great. You’ve made a lot of my own nebulous ideas about game design concrete, and you’ve also brought up some topics I hadn’t given much thought. Thanks for sharing.
Hi Tudwell! It was @edderiofer that referred your game to me in the first place and he’s reliable when it comes to recommendations 🙂 I also streamed it last year.
I find a single Tametsi level takes a good 30 minutes to solve, at least, as solvable Minesweeperlikes are always about constructing logical arguments that lead to decisive conclusions. And identifying what construction that is… well, that can take some time. I only tend to sit down with it to do a single level and then come back to it another time and do another. I think the game will last me awhile yet 🙂
I haven’t found that I want an “easier” level simply because I get stuck all the time, and moments when I get reallllly stuck tend to be deep in a level and I don’t want to quit! You’re carrying so much baggage in your head about how the puzzle structure works. I actually didn’t pay much attention to the level select until I was writing this article, because I was just ploughing though it level by level 🙂
This is, of course, not a universal experience, just mine. I find it difficult to play puzzle games at home – I find they fit me better on a commute. So I have to make sure I have some “Tametsi time” set aside for it. (And nice work; the damn game is addictive.)
Thanks for dropping by and commenting as I had gone looking for you online when it came to talking about Tametsi here and there in the last few months!
I’m looking forward to your final article in this series. I have a very similar background to you (math), and I’ve spent a lot of time thinking about how to measure puzzle difficulty. For example, I think (but cannot formalize or prove) that minesweeper on a bipartite graph is easy.
After you’re finished with Tametsi, if you feel like having a discussion of the puzzle design process, feel free to send me an email.