I am very interested in Improvisation in games. One of my favorite things in games is knowing its systems inside out and then being able to play with them in unique ways. To be presented with novel problems I have never solved before and using the tools the game gives me to overcome them, preferably under time pressure.

It’s no surprise that a lot of my favorite games strongly rely on improvisation; being able to quickly digest new situations and devise a novel solution to it. A list of my favorite improvisation-forward games might include Starcraft, Spelunky and Panel de Pon. These are games I love deeply. Games I have dropped hour upon hour into and never felt guilty about. Games that I am proud to be good at and still have room to grow.

But today I want to talk about Improvisation specifically in the light of  Ben Ruiz and Matthew Wegner’s upcomming brawler Aztez. Aztez is still a ways away from release but Ben and Matthew stayed with us for a few months in Mexico so I have had the joy of playing early versions. Before Aztez I had never really played brawlers before (I don’t count River City Ransom, fun but shallow) and dropping into Aztez has been like discovering a new unspoiled continent for me. It’s very good at improvisation and I want to discuss why.

Broad Tools

First Aztez offers you a lot of inputs, a lot of tools. I’m not going to enumerate every move in Aztez but in terms of variables to work with it has: damage, reach, knock-back, knock-to-ground, knock-to-air, stun, grab, parry, lead-up-time, cool-down-time, combos, foward movement, move to air, move to ground, etc… etc… Ben takes these variables and crafts moves out of them which in turn make up your complete tool set. These tools are varied and balanced, they each play a different role in solving the problems the game will present you with.

Combinatoric Problems

This is the beating heart of an improvisation game. What use are tools without problems to solve? Games that use randomness or unpredictable complexity are great at giving players a vast set of unique-but-simmilar puzzles. Aztez has different types of enemies but lets take a case where you are fighting four guys of the same type. Now how many problems do these enemies present you? Very many. The case of on being to your right and three to your left is different from all on the left, the enemies may or may not be attacking, they have different spacing, they have different amounts of health, they may be stunned, they may be on the ground, they may be in the air. The number and state of the four guys defines your current puzzle. Now you decide which tools to use.

I call these problems combinatoric because they are a combination of many simple states. Each single state is easy to understand and the correct response to it is known. But it is their ability to be combined that is their strength. Not only does this generate many new problems for the player but, and this is important, they are all simmilar states. On the face of it this might seem like a disadvantage. You might think you want as much breadth as possible but if you were generating very different states then players wouldn’t get to use the things they have already learned. You want to present them constantly with puzzles that are simmilar to problems they have solved before, so that they have some idea of how to solve them, but problems that are still different, so that they are forced to improvise a slightly new solution.

Many Possible Answers

Think of improvising in music, there is no correct jazz solo, although some solos are better than others. This is part of the joy of improvisation games. By allowing a large number of possible solutions you maximise the player’s chance of finding one. Obviously this has to be balanced with challenge. In Aztez there are always many actions that solve the puzzle but there are many more actions that lead to death. This also leaves room for style. Different players will tend towards different types of solutions. In Aztez you might focus on controling the enemies or on being hard to hit or just brute-force dealing brutal amounts of damage. Players will naturaly develop different skills depending on what tactics work for them early on.

Time Pressure

You could have all of the above without time pressure, but time pressure adds a beautiful flow to the game. Without time pressure there is a temptation to spend forever maximising your solution, to sit and stare and calculate. With time pressure you are forced to focus on the bigger picture and to rely on trial and error to figure out the details. This is more fun, why? Who knows, that’s  the way the human brain is built. Time pressure frees your frontal cortex from the minutea.

Incredipede, Fantastic Contraption, and the game I’m working on now don’t really use these principles, many great games don’t. But I want to start making games that embrace improvisation. Games that allow players to be artful. I’m even learning to play the flute so I can have a better understanding of improvisation. I hope in the future to make games that let you be a virtuoso every bit as much as Aztez does.

Games are inherently about getting better, progressing towards mastery and there are two ways for games to give you this feeling of mastery:

1. Skill Mastery: give you systems that you can learn and improve at (i.e. improving your multitasking is Starcraft)
2. Number Mastery: increase some number over time (i.e. increase your level in Final Fantasy)

Games are almost always a combination of the two. The core of the game is something you can improve at and by improving you are able to increase some number that is a gauge of your skill. I think the less of a gamer someone is the more easily they confuse these two things. The most casual games have almost no skill mastery but strong number mastery systems (Farmville being an amazing example). The least casual games usually have number mastery very tightly coupled to skill mastery with Starcraft and Chess’ ladder systems topping the chart. Game designers, who think about games all the time, should be really good at distinguishing between these two things and that helps them like Spelunky, which has wonky number mastery.

Spelunky is a hardcore skill mastery game. No matter how many times you play you always start the game naked as the first time you played. In the video above and in other places Derek talks about how Spelunky has a non-traditional difficulty curve: it starts very steep in the mines, gets even harder in the jungle and then eases out with the ice and the temple. But the rate at which people aproach skill mastery is the opposite, they start bad and then get better and better and better. This means the game’s number mastery system (getting to later levels) is telling you that you are improving very slowly even when you are improving very quickly. Even a future master of the game will spend a lot more time dying in the mines then they will ever spend in the late-game temple levels.

The result of this is that the game is very frustrating to players who look to number mastery for their “I’m improving” fix (since improving at something is why games are fun) and less frustrating to players who “see past” the numbers and know that they are greatly improving their skills despite not being able to beat the first level.

Does this mean that some people are better at self-gauging when they are improving at things than other people? Or that they trust their own sense of improvement over external indicators? Or that they value external indicators less? Probably a combination of all three. I would also wager that the more games you play the more true all of these things become. As time passes you play less and less for the numbers and more and more for the raw experience of playing and learning, of experiencing novel environments and learning to master them.

So game designers, because we play a lot of games, probably like Spelunky more than people who have played less games. You could also argue that being in indie games self-selects people with a strong attraction to internal indicators over external ones. Making an indie game requires months or years of work with little clear, positive, external feedback. Meaning we have to be good at self-guaging when we are improving.

It may be that all of these things make it easier for players like me to enjoy Spelunky, but I doubt that it makes Spelunky better for us. I think we would probably enjoy it just as much if the numbers reflected our growing skill more generously, and maybe all those people who find it “too hard” would realise that the game is just lying to them.

There are a lot of game ideas out there and it’s not obvious which ones are going to be fun. I have a file on my desktop called gameIdeas.txt which is a list of over 40 ideas ranging from “One button game: afghan kite fighting” to “knitting!…?”. Picking which game to start is easy, which one makes your mouth salivate the most? Picking when to give up on it is much much harder for me. Even when a game isn’t fun after months of work and exploring I still feel like just one change, one decision, could make it great. On the other hand, the only two of my prototypes that have worked out and been fun enough to finish were fun right from the start. Fantastic Contraption and Incredipede were both fun (for me) almost immediately and Sarah’s games, Rebuild and Word Up Dog, were fun right away. Is that a general rule? I wrote a psudo-random collection of eight games designers I know to try to find out.

I wanted to know if their successful game was fun right away, I figure if all successful games are fun right away then I can skip the months of grinding on ideas that seem promising but just aren’t working out. So without further ado, here is the data:

Derek Yu, Spelunky

Was it fun right away? Yes

“Spelunky was honestly the smoothest development I’ve ever had”

Justin Ma, FTL

Was it fun right away? No

“We felt the idea (which was very abstract) could be fun but the actual prototypes were not enjoyable for months.  We could see hints of interesting gameplay but it wasn’t really fun for a long time.”

Jonathan Blow, Braid/The Witness

Were they fun right away? Yes

“Braid was fun in the first week. The Witness probably took 2-3 weeks, if only because it is 3D”

“I do think this depends on one’s level of design experience, though. Part of being a good designer is being able to just home in on what is really good, without having to spend a lot of time slogging through mud.”

Jan Willem Nijman/Rami Ismail, Radical Fishing

Was it fun right away? Yes

“At Vlambeer, most games come together within a few days. If they’re fun, we work on them for a few weeks. If they’re fun to work on for a few weeks, we turn them into a project (or not).”

Dan Cook, Triple Town/Leap Day

Was it fun right away? No

“Of the prototypes I’m working on now, I’d say that the best ones convert over a few weeks and then we really know we have something fun after 2-4 months.”

Marc ten Bosch, Miegakure

Was it fun right away? Yes

“Miegakure was the third prototype in a series of ‘games in higher dimensions’ prototypes. Ignoring the first two prototypes, Miegakure is incredibly similar to its original vision”

Michael Boxleiter/Greg Wohlwend, Solipskier/Gasketball

Were they fun right away? Solipskier: Yes, Gasketball No

“I do think it’s important to get something down that’s interesting as soon as possible, something that you can play over and over, but there are a lot of games that I would never be able to make if it had to be fun in a few days or a week.  It’s really hard to gauge early on, and I don’t really have any good rules even now.”

Mike also mentioned that most games he’s worked on have been fun right away with a few exceptions.

Cactus/Dennis Wedin, Hotline Miami

Was it fun right away? Yes

“It was fun straight away from when the basic gameplay had taken shape, which took less than a week.”

So 5.5 out of 8 replies were “it’s fun right away”. Clearly not every successful game is fun in the first week but a lot of them are. And if you can make a great game without months of smashing your fists against the wall then why not! I like that Vlambeer has gone so far as to build this into their process, judging games by how quickly they become fun rather than how fun they might be down the road. Marc’s aproach of exploring the same design space with totaly new games is also very appealing. If I want to make “knitting!…?” it makes sense to try a bunch of disparate ideas in the design space instead of settling down with one and trying to force it to work.

FTL and Dan Cook are the counter-examples, and they are strong counter-examples. FTL won the Excellence in Design Award at GDC this year and I personally respect Dan Cook as a game designer more than just about anyone (Leapday is totally amazing). Standing opposite to Vlambeer, Dan has built the months-long search into his process and the resulting games speak for themselves. Clearly you can make great games by taking a rough idea and through teasing, exploring, and experimenting make it into something wonderful.

But that’s the thing, all that teasing… I hate stressing about whether I should keep working on a “promising” project or whether I’m groping in a blind alley. The fun-right-away rule takes away that stress entirely and maintains a good chance of finding a great game. It’s the way I’m going to prototype my next game. It’s all dim sum from here on out!

Spelunky Logo
Originally uploaded by gerbil.llama.

I’ve been playing Spelunky all day and just wanted to get a few thoughts out of my head.

Most importantly: Spelunky is Nascar racing.

At some point I was trapped in a hotel room and ended up watching the Rednecks Turning Left show. I discovered that cars break down alot in Nascar. There are all sorts of things that can go wrong in Nascar from mechanical difficulties to cars running into eachother. What is interesting is that the percentage chance of something going wrong is proportional to how hard you’re driving your car. So if you take off like a hare you’re car is going to fall apart and you’re going to lose. But if you’re slow like a turtle everyone will pass you and you’ll lose. So Nascar is the game of balancing the two goals of going fast and not breaking down. It’s fundamentaly a calculus problem. Which seems to be amusing to the human brain.

Spelunky is almost exactly the same thing. But in Speluky speed is replaced by collecting gold.

Fascinatingly the goal in spelunky is not to beat a set number of levels. It is to get 100,000 gold to the guy standing at the end of level 4. But you can make multiple trips to fill him up. So if you take him 50,000 the first time you play and 50,000 the 2nd time then you’re done. But you can also play through ten times and take him 10,000 gold each time.

Spelunky is also fucking hard. Making breakdowns inevitable. So the longer you spend in the level the more likely you are to die. But the only way to get more gold is to spend longer in the level.

So you have to do the same kind of calculus Nascar teams do. What is the optimal amount of gold to grab each time through the level? If you try to get all 100,000 in one trip then you may die hundreds of times before you make it. But if you just grab 1,000 you have to play it at least 100 times!

This is a fantastic meta-game that gives each run purpose and meaning as every play you make is seen in the light of this risk/reward calculus.

What would make Spelunky really awesome is if Derek explained to you that this guy who wants 100,000 gold exists instead of making you wander off to the Spelunky Wiki to figure out how the fuck to save your progress.