A Good Snowman Is Hard To Build

A few weeks ago I published The Developers Who Won’t Hold Your Hand which discussed design considerations around a growing subgenre of games that leave the player to figure out the mechanics. I made use of droqen’s term “discoverable systems” because, frankly, we didn’t have one.

This was based on a number of interviews comprising over 5,000 words in total. That meant I had to cut a lot of words, even interesting stuff.

But recycling is good for us, so I’m presenting some of these lost words in a separate post.

Managing Player Learning

Alan Hazelden had a lot more to say on differentiating tutorials for controls vs learning mechanics through the game. I think the important bit here is how players appropriate tutorials as mindless clickfests for the purpose of progress instead of using them to learn. Yow, you got me, Alan.

All my games have very limited controls, and even then they still need a “this is how you control the game” tutorial! It’s very brief, but it’s there in all my games. You’re told how to control the game (press the arrow keys, or use the mouse over here), and you’re also prompted to use undo/reset the first few times the game detects that you’ve failed a level. The reason my games feel like they don’t have a tutorial is because this stage of teaching is so brief, but the only way I can get away with that is because I never change the controls or add new UI elements.

Once you have players understanding their possible range of inputs, you can deal with teaching them how to understand the consequences those inputs will have. Here there’s two parts – introducing the concept and then reinforcing it until it’s something the player actually remembers.

There’s consensus that just telling someone how something works is broadly insufficient – lots of people will skip past text without reading it, or even if they read it there’s a good chance they won’t have internalised it (i.e. they’ll have forgotten what it says 10 seconds later). So the traditional “intrusive tutorial” approach is to tell the player what to do, and keep that text onscreen until they’ve done it. The non-reading player either stumbles onto what they need to do, or runs around the level for five minutes until they slow down to read the flashing text saying “hey you can use your sword attack on those crates over there”.

Even then, however, it’s possible that the player hasn’t internalised the lesson you want them to learn because they’ve only done it once and you don’t know if that’s because they understand or because you told them what to do. Players go into a different mindset when following instructions. If you’re being told to press button X, you might not need to pay all that much attention to what happens when you press it. All that matters is that now you can walk further down the corridor, right? So the connection that forms in some people’s heads is between “there was a button prompt/tutorial message” and “I pressed this button to progress” rather than between “there’s a crate in the way” and “my sword can break crates”.

So if there’s any delay between when you introduce a concept and when you start requiring the player to use it, you’ll have to repeat it a few times to make sure it’s remembered. In a bad tutorial, this feels repetitive because you’re being asked to jump through hoops (sometimes literally) to prove you can do something. In a good tutorial, there is no delay – you immediately start using the mechanic that you’ve just been introduced to, in a non-trivial capacity.

This is essentially moving part of the game into the tutorial, replacing the reinforcement stage of the tutorial with a section of the “real” game that serves the same purpose. But what if you could do the same thing for the first stage of the tutorial too? What I do is to design levels which make it easy to discover gameplay mechanics as a natural consequence of experimentation. I don’t need to explicitly tell you “if you move into a crate you’ll push it” if I put you in a room filled with crates and the only controls you know about are for moving. Even if you’ve never played a Sokoban game in your life, you’ll work that out yourself (or get very bored).

I then gate your progress, and make sure that to get to some sections of the game, you need to have gone through A) at least one level which makes a specific mechanic discoverable, and then B) several levels which require the use of that mechanic.


Prior Experience

A player without any experience of first-person shooters is going to struggle with DOOM yet an aficionado is going to yawn through a shooting tutorial. Most players do not happen upon discoverable systems as game virgins. Who are your players? What do they know? What can you assume about them? First, droqen.

Some games rely on existing knowledge. Some games rely on players figuring things out. Likewise, some players consume things in different ways… I have clear memories of my sister not really grasping the concept of moving and jumping at the same time. Maybe it was an input dexterity thing, whatever, but when was the last time a game described “move and jump at the same time to do a moving jump”?

Games are built on skipping instructions. Show me any game and I’ll tell you here’s where they didn’t tell you what to do and you had to deduce, decide, risk, try — something.

A shooter might not tell you to shoot things, but most do, right? Or at the very least they teach you “here’s how to shoot, and here’s something that will hurt you if you don’t do something about it.” You put two and two together. It’s only existing knowledge because it’s something you’ve seen so much you forgot why you did it in the first place.

If a hypothetical shooter does tells you to shoot things the instructions will be skipped down the line, eventually. “Shoot this thing,” it says, but it won’t say “shoot this thing first.”

So, you know, Starseed Pilgrim is, I hope, not built on existing knowledge. Most of its rules are weird and alien and things I haven’t seen before. I’d like to say that was intentional: I tell you explicitly how to do all the basic platforming stuff, but the weird stuff? That stuff is the stuff I think and hope is interesting to learn.

Santa Ragione exploited what the player thinks they know for Mirrormoon EP:

We are aware of genre clichés and we are aware that certain objects or situations are self-explanatory. That is also an important characteristic to research when you are designing an object: the more self explanatory the object you design is, the less you’ll have to teach the customer and so the more natural will feel the experience with it.

MirrorMoon EP plays with this concept, but it turns it upside down to confuse the player. We use common videogame interactions but we design unexpected outcomes so the players will often find themselves out of their comfort zone. What is really hard to balance is to create curiosity without falling into frustration. On the other hand we refused to have any “gaming knowledge” required for the game, that is why there are no skill/reflexes based interactions, or mouselook controls.

Andreas Zecher related a similar approach for Future Unfolding:

Beyond basic controls to move the avatar, we use existing game conventions to break them and surprise the player. For example, in other games forest make up invisible walls, while in Future Unfolding you can dash past each tree.

Future Unfolding

Living Without

I asked developers if they thought we could live without tutorials or instructions, because I was not convinced this was universally achievable or even desirable. Consider the complicated interface you learn to play an RTS or the clarity of an RPG with their armies of stats. Yacine Salmi from the Ellipsis team:

Yeah, I think you can’t avoid for certain game types. But I think it’s critical to think about how you will teach the player. The tutorial (whether explicit or implicit) is the first interaction the player has with your game and will form a lasting impression. I don’t think it’s something you can simply throw in at the last minute. You also can’t teach everything in one shot, especially for a more complex game. Zelda games do a good job of reinforcing various concepts throughout the first few hours. The Super Mario games do an amazing job of this. Super Mario 64 didn’t force you to read through any texts. You could experiment and learn all the various jump mechanics through trial and error, but you still had the optional (and clunky) signposts to teach various specific mechanics.

My favorite types of tutorials are the ones that don’t feel like tutorials at all. You mentioned Stephen’s Sausage Roll. There’s no explicit tutorial but the whole game itself feels like one giant tutorial. You’re constantly learning and re-adapting various strategies. It’s the same with The Witness. No explicit tutorial (or text!) but the whole puzzle progression is a tutorial (or a set of tutorials).

And Eliott Johnson suggests this should not embraced as an one-size-fits-all ideology, more of a useful approach that appeals in the right circumstances.

I’m not entirely convinced that all games should take a step back – though it might help normalize a more methodical and patient approach to games. I guess I lean more towards a hands off approach where it might create more meaning within the game, rather than as this blanket, slightly puritanical notion that working it out for yourself is always better, or “good for you”.

On the other hand, Santa Ragione suggested the usability of a discoverable systems game might be superior to one where the designer’s mission was simply to reduce instruction. All because of intention.

MirrorMoon EP is a game designed around the absence of knowledge and so the absence of a tutorial is part of its monastic design. The game is like a book designed to be hard to read: you have to provide your own translation and compare it to others to find a shared meaning.

On the other end, games that don’t purposely go down this path face a common usability problem from lack of affordance. The more the affordance the less the struggle to understand and use a new object. In this regard tutorials in games should not be an excuse for poor affordance. Explicit tutorials should in any case be there only if any other solutions are not possible. More often than not games are complex because of bad design practice rather than by choice. Tutorials are then used as a way to address these problems, but often they make it even worse from an expressive/artistic standpoint.

There wasn’t really much argument about whether all games should become instructionless artefacts. However, Andreas Zecher disagreed about the limits of such a design practice:

I don’t think those limits exist. Tooth and Tail is an interesting example of making an RTS game more accessible and less arcane.

Probability 0


Eliott Johnson went into heaps of detail about how the design of A Light in Chorus had developed; perhaps the most interesting detail was about the origins of the idea. To set the scene: A Light in Chorus asks what an alien might make of the Golden Records on the Voyager spacecraft.

The project of the Voyager Golden Records was to make something that could be understood and operated without language. Affordances (and some big assumptions) were all they had to work with.

We’re really only running with the same brief and trying to give players a first-hand experience of how something like that might play out. We’re asking how an alien (the player) might interpret an object (our version of the Golden Record) that they have no experience of, and how that might compare to the thing it purports to represent. It’s a specific scenario but one that could be generalised across a lot of games (our take on the classic player as amnesiac trope, learning the rules of a strange new world).

This style of design had been there from before we’d settled on the Voyager narrative though. I’d always admired and enjoyed the types of games that left you to put the pieces together more than those that explained it all, it left more to the imagination. Starseed Pilgrim, Antichamber and Proteus are all important touchstones.

What took time was realising just how powerful the narrative wrapping/imagery of a game could be as its own form of instruction- priming players for a certain type of experience. Earlier designs (pre-Voyager story) were very heavily focused on animal imagery, the deer especially, and I think there was a conflict between players’ expectation for a more story-based experience and what we were actually making, which at the time was much more puzzle based.

A Light in Chorus


They Just Don’t Get It

A Good Snowman is Hard to Build is a remarkable little Sokobanlike; it isn’t ridiculously hard and the levels, in the main, are snug with just a handful of moving parts. However, it does have an important secret: a secondary, fearsome challenge that forces you to reconsider everything you’ve done. This hidden game was partially why I invited Alan to chat about instructionless games.

I found this second stage but failed to figure it out and, reluctantly, turned to the internet for advice. Does Alan consider this a success or failure?

The dream world in A Good Snowman definitely has problems, but I think they’re mostly a failure to bridge the gap between “I’ve completed all the levels and I think I’ve seen the whole game” and “I’ve completed all the levels but I’m aware that there’s a second stage to the game and this is where I go to start it”.

Once you’ve found where the meta-puzzle starts, I feel okay about the lack of instruction about what to do – it’s part of the puzzle really, and everything there that’s slightly obtuse is satisfying when you realise how it works. Plus, if someone has got to that point, they’re probably fairly invested which helps.

I’m always surprised when players don’t get what appears to be simple, but everyone comes at games in their own way. I thought Ellipsis was pretty well done, but Yacine explained they had their own demons to deal with:

In earlier versions, there were some issues. Some players would forever go back into the first level until we prevented that. There were some more subtle mechanics that some players didn’t pick up on, like how the little dots spread out based on your accuracy or speed, or how the score system works. But these are less important as you don’t need to understand them in order to progress in the game. And for some players, when these more subtle points ‘click’ in their head, they tend to appreciate the attention to detail and the feeling of discovery.

I reported that the A Light in Chorus team had been anxious about going down the discoverable systems route. But I was surprised to discover Yacine had also been troubled:.

I worried about it a lot. For me a big part of game design is how you teach players. I didn’t want to do an instruction video and with text instructions there’s always the worry that players will simply skip over it. I wanted to provide a space for play and experimentation and make it easy for the player to learn from their actions. We spent a lot of time playtesting with new players (once you grasp the concept, it’s pretty easy, so we had to constantly have to find new players). We’d hand over the tablet and watch the player figure things out. It took a long time and a lot of iterating but we only released once we were confident that 98% of players could pick it up and figure it out without explicit instructions.

He Who Walks Away

And finally, droqen had an anecdote that almost made the opening for The Developers Who Won’t Hold Your Hand. I decided against it in the end, as it was too ambiguous for a punchy opening and I swapped it for droqen’s reactions to the linear design of Half-Life 2.

Here’s the anecdote.

There was a guy who played [Starseed Pilgrim] at the IGF booth. He’d been talking to me for… I dunno, 5-10 minutes. What are your inspirations, what’s it about, blah blah blah. I answer him, I guess. I really don’t remember what I said. He finally gets his turn, does the little tutorial, digs up the whole first island and goes into the dark star and dies, goes back to the hub world. Okay.

Earlier I’d decided I wouldn’t help anyone with the game, for my own sanity and to preserve the instructionless mystery some people loved. So I just watch.

He does it again. Same thing. Zig-zag from top to bottom to his death. Okay.

And he does it again.

He puts the controller down and walks away without a word.

I love that memory.

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!