2013-03-19: Player as Programmer III: Rules, Levels, Running, and Multiplayer
The source for this post is online at 2013-03-19-pap3.scrbl.
The last two weeks I discussed the utility of the concept of a player as a programmer when understanding fun in games. This week, I continue to discuss these concepts.
-
The high-level points from before were:
Replayability: The best games are fun to write and to execute.
Story Heavy Games: Story is a liability when creating a fun game, because it takes attention away from the gameplay and is inherently un-replayable, no matter how good it is.
These proclamations are mostly about what is not fun and don’t give a lot of advice about what is fun.
1 Fun to Write: Foundational Rules
I think the foundation of fun when writing functions is: reasonable rules with understandable consequences that reward mastery. One of the most frustrating experiences when playing games is when there don’t seem to be general rules controlling what’s going on, where every situation is capriciously different than those situations before it, and where earlier knowledge and experience does more to hurt you than to help you.
Many puzzle games have this feel, where each situation is so totally different that there’s not really a sense that it’s the "same" game. Many run-and-gun and shoot-em-up games have this feel as well, when they seem to reward only pure memorization and not the building of a general skill. The most insulting form of this is the modern Quick-Time-Event, which is completely unpredictable and follows no rules or internal logic. Another more modern form of the lack of coherent rules is when a game has many "mini-games" or special levels that obey totally different rules. An example close to my heart is Bayonetta, which is almost a perfect game but suffers from a weird shooting mini-game, a weird racing level, and a weird flying level. The turret sequence is an incredibly common lame mini-game.
Strategy and role-playing games seem to obey this framework most rigidly because they so clearly define their systems in which all game interaction takes place.
Most platforming and action games tend to obey this fairly well also, but are far more likely to introduce unique situations on a per-level, per-encounter basis. These "gimmicks" tend to give variety to the foundation that might otherwise feel bland. Eventually those gimmicks stop and the game reaches a vocabulary plateau on which there’s a complex framework to build levels on.
Summary: A foundation of rules that rewards mastery and has clear consequences is necessary to create an environment where learning can take place, and thus is necessary for a programmer to work and thrive.
2 Fun to Write: Levels
Levels are interesting from a programming perspective. You could understand each level has its own unique game, because a different function could be designed to perfectly beat each level. (Indeed, this is a fairly accurate description of what a tool-assisted speedrun is: a unique gameplay script for each level.) Yet, when that is the most productive way for a human to play the game, then it suggests that the game does not have coherent rules that don’t discriminate between levels. This suggests to me that levels are best seen as:
A way to gradually train the player in the rules to write the function they need to be successful at the game.
A way to constrain the overall game design to obey rules to create independent experiences.
A way to lengthen a game that is fun to execute (so you have more time to execute it)
In my opinion, the game that best demonstrates the strength of the level concept is Super Mario Bros.. This single game in 1985 has an incredibly minimal ruleset and very few gimmicks, yet creates a wide variety of levels. Each new version of Mario takes the same rule framework, adds a few new gimmicks (like Poison Mushrooms, new power ups, new enemies, etc) and then expands the set of levels to be larger. Each game is not different, in the sense that it requires a new training/design/programming phase to learn to play it. Its continued success suggests that it is a Fun to Execute, since the learning is relatively minimized each time.
Summary: Levels are a tool to structure learning and structure play, as opposed to a way to divide new and different learning and play.
3 Fun to Execute: Why?
At first, it seems strange that any functions would be fun to execute, because it so boring normally to be a computer: no one enjoys doing basic arithmetic or manually stepping through the execution of their programs. Indeed, it is insulting to our humanity to do things that a computer could do. This suggests, however, that the functions that are fun to execute are the ones that a computer could not do. This, I believe, is the key to understanding which functions are fun.
Games where we could easily imagine creating the perfect player programmatically are unlikely to be fun, because either it is boring to actively be that player and "run the algorithm" or because it is upsetting to know that the player exists, we could write it and run it, but to do so is insulting, so we don’t, and we don’t perform as well as that player would.
An example of this for me is Minesweeper, because I know exactly what I need to do and it is just a matter of doing it over and over again.
A more subtle example is something like
Spelunky where a major
component of the game is "being careful". The ideal player goes
through a monotonous routine over every pit—
Simply because a computer player exists doesn’t mean the game isn’t fun for humans. For example, the Infinite Mario AI doesn’t play at all how a human would, so it is not an attack on Mario’s fun that this AI exists. Although it does suggest how to "unlock" the fun that humans can have: they need to be overwhelmed to the point where the only way to succeed is to develop subtle heuristics to guide when to apply the rules you’ve learned playing the game a lot.
When I play games I can feel myself doing this very consciously: I’m doing a level of Mario and I notice a small pattern that I "know" how to solve, I do it, and keep going. I’m playing a shmup and I see a bullet pattern coming toward me and I get a feeling to dodge in one direction or another. I’m playing an action game and get an inkling that a certain attack is incoming, so I fire up how to avoid that situation. It is extremely satisfying to build up a large repositories of such patterns and their response and then build little plans about how to apply them.
The key, however, is to be able to first overwhelm the player so that there’s nothing better than heuristics at their disposal. There are a few common techniques to overwhelm:
Overwhelm the space of possibilities by increasing chance, with randomness or with the complexity of the player response mechanism, so the player can’t effectively simulate.
This can be dangerous, because players rely on a certain level of internal simulation to be able to predict the consequences of their own actions. If you take this form of overwhelming too far, then you risk hurting your rule foundation.
An example of a game that does this extremely well and clearly is a roguelike such as NetHack where the frontier of predictability is very large and close, so the player has to act in the presence of missing information a lot.
I find that Super Mario Bros. takes the second approach of increasing the complexity of the player response. Although the Infinite Mario AI does exactly this, it is not feasible for a player to actually simulate the exact outcome of all their choices, so they take more modest and coarse simulations and build their play around them.
Overwhelm the space of information available to the player, so they can’t effectively track it all discretely and must instead use heuristics to capture a useful fraction of it.
This can be dangerous, because it is easy to create a situation where a player is frustrated that they can’t harness it all, because they know that if they did, they could "crunch the numbers" and take the perfect course.
I personally find it very frustrating when a RPG has a sub-system that would be easy to do correctly if the data were just available in a computer parseable form. In fact, for Persona 3 I wrote a program to compute the ideal way to combine the monsters for your goals based on a plain-text database I found in a FAQ. I find the game to be designed poorly because since this approach is possible, they don’t just embed a solver for it in the game so you can spend your time doing the fun part of the game. (And indeed, in Persona 4 Golden they removed the need to repeatedly try combinations until you were randomly assigned the skills you wanted, now you can just select them.) Other examples are when you discover a way to increase a resource infinitely and can’t "show the game" so you can just get as many as you want. For example, in Dragon Quest 9, once you unlock the teleport spell, the cauldron, and have a certain (small) amount of money, you can generate an infinite amount of money by exploiting an arbitrage opportunity with an item being worth more than its ingredients. It would be nice to just automate this and get an infinite amount of money than to have to actually do the mindless process and get the money.
In any case, some games do this quite successfully. I think that shmups do this with their complex bullet patterns: all the information you need to dodge precisely is apparent and obvious, if you tracked everything, you’d be perfect. But there’s just too much to actually do that, so you use heuristics and avoid things using different techniques than actually tracking each bullet.
When I look at games that are almost fun, I almost always find a way some overwhelming has been taken too far to break down the ability to form rules and heuristics to play.
Summary: A fun game overwhelms the player into discovering and using heuristics over discrete algorithms by offering overwhelming possibilities, information, or consequences.
4 Multiplayer Games
In general, I would say that all multiplayer games are designed around
overwhelming the possibilities of what your opponents could be
doing—
Unfortunately, this kind of overwhelming possibility is inherently based on predicting human action. The problem with that is that humans are capricious and do not need to follow any rules. Thus, when they enter your game as an inherent piece of the system, they destroy the ability to have a firm foundation of rules. This is why playing a game online can be so frustrating and unsatisfying: you can’t actually train and learn, because your opponents can break rules at any moment. Another way to look at it, that is commonly used when training new humans into a game’s culture, is that if you want to play against other people, then you have to play the person and not the game. In the Street Fighter/fighting game community, this is commonly understood as the "real" game.
For my taste, these are simply not games because they do not lend themselves to the traditional ways of playing and learning games. Instead, I like to refer to them as "arenas". Starcraft, Street Fighter, Call of Duty, etc are new arenas in which you can compete with other humans. Now, due to how games are made and marketed, it is very common for every multiplayer game to be combined with a single player game that replaces the humans with AIs that can be learned. When we’re discussing that, then, of course, these arenas are "games" to me. In many cases, however, they are not good games. Many fighting games, for instances, are not designed to have good or interesting AI opponents, since the human arena is the primary aspiration of the developers.
I don’t necessarily have a problem with "arenas" like this, but I don’t have taste for them. I find the experience of playing online to be very annoying because other people don’t have the same linguistic standards I do, I have to deal with lag and other issues, and in general I need to coordinate with other people, which is annoying.
I am skeptical, however, of the development of new "arenas" that are on par with the arenas humans have created over time. It is rare for a new arena to be better than something that has stood the test of time like Poker, Chess, Football, Go, etc. In particular, I observe that many modern multiplayer games that people are excited about have very limited strategic spaces: the small number of openings for Starcraft as opposed to the comparable space of Chess openings. I don’t think humans have yet created a computer arena that can realistically compete with physical arenas, but I agree with the zeitgeist that Starcraft or DotA is probably the closest we’ve come.
Summary: Multiplayer games are fundamentally broken as learning environments because humans are capricious and do not need to obey rules. Therefore, to play and appreciate multiplayer games, you need to appreciate psychology, which is quite distinct than why normal games are appreciable.
5 Fun to Input
A unique form of fun can be extracted from the way you interact with certain games. For example, plastic instruments and jumping on giant buttons while "dancing" can be fun for its own reasons independent of the fun of the "game". (In fact, most of these games are the most devoid of fun in normal circumstances: memorizing the discrete order of button presses with no reason for those buttons over others.) I’ve found that most of these forms of fun quickly lose their excitement and the core input style of the D-pad and buttons is far more efficient and applicable to a wide variety of games.
I personally find that some popular inputs are negatives for me. For example, I can’t stand touch input because it is so imprecise and lacks the ability to have individuated digital actions. Furthermore, I hate to use mice (in general) and so I don’t like games that use mice. Finally, I don’t like games that use the keyboard as input, because the buttons are all so similar, you can’t have a feeling associated with each one, like with a controller. As another point against the keyboard, it has far too many buttons and doesn’t encourage economy in designers.
On the other hand, I find the arcade stick to be a uniquely satisfying input that can make almost any game more pleasing through it’s nice clicks and gigantic movements. One of the main reasons I like to play Street Fighter (and similar) games is just because it is so nice to execute the commands on the arcade stick.
I don’t think input is a significant enough source of fun to be worth designing around. I think it is far more likely to guarantee your game is limited in its appeal by choosing a strange input system.
6 Conclusion
As Alexander Pope said, to compute is base and to program is human. It is wonderful to feel that you are learning a system and building a repository of ability to solve new instances of that system. When the system is so complex, we can’t imagine making computers that solve it quickly and easily, yet we can do it ourselves, we feel even more human and wonderful. This is the experience that good games create.