Skip to content

QWOP and Simulation Design, Pt. 1

July 9, 2012

You are QWOP, our small nation’s sole representative at the Olympic Games. Use the QWOP keys to move your legs. Ideally you will run 100m…but our training program was under-funded. Remember, it’s not about whether you win or lose. (

When creating a simulation of a sport, there are really two primary goals. The first is to reproduce the rules of the game as they exist in the rule books, that is, to accurately model the game that is being played. This is a relatively easy task, as sports rules tend to be fairly unambiguous and well defined. These features mean that the rules that define a sport lend themselves well to the strict limitations of a digital representation. The second task is harder, because it requires modeling everything else, things that we simply take for granted in the real world. These run the range from physics simulations (rebound behavior), to physiology simulations (player exhaustion), to psychological simulations (teamwork). Typically, it’s in this second area that games distinguish themselves.

That isn’t to say that the first task is entirely without it’s own pitfalls, especially in the good old days of video gaming. In a piece for ESPN, Patrick Hruby reported that properly modeling all 11 players on a side in football was a particularly difficult task for the first Madden NFL game thanks to the limited memory available to games at the time.

Some of the pain was technical: making a game on a computer, the Apple II, that didn’t have enough memory, pixels or disk storage. No sound chip, either, and only one joystick port. The machine could produce four colors, sure, but only if a programmer knew all the dirty tricks. Anything beyond seven-on-seven football caused the on-screen action to slow to a crawl. (ESPN)

But it isn’t just technological limitations that provide hurdles to proper sports representation. There’s also the fact that video game players often don’t really want to play a fully accurate game experience. Most people who sit down for a game of Madden don’t actually want to spend 3 hours playing a single game. Test Drive: Le Mans players notwithstanding, most people don’t have the time or dedication to replicate the rules of such matches completely. This commonly means that play experiences are truncated for time vs. their real world counterparts.

This can introduces some strange results. The effects of playing a 9 hole golf course, or a 3 inning baseball game are pretty clear. There’s less variety, and fewer opportunities for comebacks, or for endurance players to be well represented. However, relatively little violence is done to the underlying game design. But something strange happens when you start playing with time-clocks in games with continuous running times. It seems innocuous enough to reduce the play clock in a soccer game to 1/3, so that an entire game can be played in about 30 minutes, as opposed to 90. However, as noted in Jesper Juul’s Half-Real, because the spatial dimensions are unchanged, it invents a strange world where everything is actually moving at 1/3 the speed. If you were to reproduce the game on a real world stage, it now takes several minutes for a player to cross the field. A kicked ball glides through the air at a leisurely pace.

The art of simulation design then, is to choose what you care about modeling, and what you don’t. There’s an old joke where a dairy farmer asks a group of scientists to help him to maximize the milk output of his cows. The veterinarian recommends using a more nutritious feed. The psychologist recommends reducing the cows’ stress by giving them more pasture time and less time in the barn. The physicist says “Assume the the cow is a sphere…” The joke, such as it is, is about how physicists by necessity work with abstractions of the real world, because only certain details are relevant to each investigation.

This kind of modeling process should be familiar to most designers of simulation games because for most simulations it is either impossible or undesirable to create a perfect representation of the activity being simulated. The job of the game designer, like that of the physicist, is to determine which aspects of the activity are relevant to the goals of the simulation. In sports simulation modeling, there are roughly 3 major approaches one can take:

Reproduce the Spectator’s experience

There is a board game called “Pit“, which purports to simulate the activities of a wall street trader. The goal of the game is to be the first to get all of a certain commodity in order to “corner the market” by trading with other players in real time. Since the game takes place in real time, trading is accomplished by each player yelling “Wheat! Corn! Barley” at one another in accordance with what they’re trading away and what they’re willing to take. Pit isn’t a very good simulation of a market, in that it doesn’t actually model money, and the actions taken by the players don’t really have much to do with what a commodities market trader really does. But anybody who’s ever seen video of the stock market floor can tell you that it feels like what it those traders appear to be doing: yelling at each other and waving pieces of paper in the air.

Pit really exemplifies this style of simulation design, which concerns itself not so much with what the participants in the original activity are doing, but with what spectators of the activity perceive the experience to be like. For obvious reasons, this is most common with simulating activities that take a large amount of domain specific knowledge, such as flight simulators. As noted in the above example, differing choices of the spectator may lead to a simulation having relatively little to do with the core activity it purports to simulate. A game that represents military operations based on Rambo and other such action movies may not be accurate, but it can still accurately capture the audience’s perceptions.

Often, this kinds of simulations are derided by people with relatively more knowledge of the subject matter as “arcade-y”, or less rigorous simulations. However, which this is often directed as a criticism, it’s one based purely on personal preference rather than that a rigorous simulation has any kind of intrinsic value. While designers who choose this route are often criticized as being lazy for leaving out features, that isn’t necessarily the case. From a game design perspective, perfect simulation is the easiest path, since all the systems are defined for you. Choosing only to include specific elements requires more effort, and more familiarity with what elements of the activity being simulated match best with the designer’s goals.

Reproduce the Participant’s experience

Just as reproductions of a spectator’s experience are often treated dismissively by people knowledgeable about the subject matter, those people tend to prefer simulations designed to represent and capture the experiences of the participant, often because those experts make up class of participants whose experience is to be reproduced.

While this is often associated with “rivet-counting” accuracy, that isn’t necessarily the case, because as in any modeling task, some abstraction can still be applied regardless of the type of experience that’s being simulated. In the game God Of War, the player’s character is occasionally tasked with opening a metal porticullis by lifting it off the ground and throwing it upwards, essentially analogous a weightlifter’s snatch. To model this, the player is tasked with pressing a single button quickly and repeatedly. The act of pressing a button is nothing like the act of lifting weights, but one of the designer explained that they chose this method to replicate the feeling of what the character is doing. Lifting the gate is hard for the character, and pressing a button quickly is hard work for the player, so the latter activity was chosen to model the former. Again, the simulation is an abstraction of the task being represented. The job of the game designer is to determine what those abstractions will be.

Note that this also isn’t necessarily about reproducing the experience as though the player of the simulation was the one executing the tasks being simulated. This distinction is particularly important when considering simulations where the player’s skill level is significantly different from the skill being simulated. If my mind were, by some strange twist of fate, inserted into an NFL quarterback, my experience after the snap would be a sensory overload mixed with sheer terror for mere moments before I was flattened by the opposing team. An actual quarterback at the professional level experiences that completely differently, not only in their ability to withstand the weight of 250 pound men hitting them, but in the amount of data they’re able to process and react to in that time.

David Sirlin refers to this as “Presence of Mind” in the context of fighting games, but it applies equally well in any number of high stress situations. Experienced players, having intimate knowledge of the different ways that each encounter can play out, report that time appears to slow down so that things that would be entirely lost on a less skilled player become clear, and reaction and even planning become possible within the same amount of time. So, the designer may decide to use techniques to artificially lengthen the amount of time available in the simulation in order to allow a layman to experience events the way a professional would. However, this amounts only to changing the decision of who the participant to be represented is, and doesn’t change the underlying design philosophy.

Reproduce the mechanics

In some, relatively rare, cases the designer may decide that his simulation model shouldn’t focus on the experience of anybody involved, but should instead focus on replicating the mechanics of the activity, with varying degrees of abstraction that may render those mechanics foreign even to experts in the field. The recent rise of motion control has lead to a vogue for this type of simulation. Games like Dance Central and Wii Sports seek to force the player to reproduce the actions being simulated with varying degrees of verisimilitude.

The distinction between doing this and simulating the player’s perspective may seem relatively thin, but it doesn’t need to be. When I’m playing Dance Central, I’m essentially playing a full body version of Simon Says. This isn’t the feeling that an expert dancer feels when performing, who will probably be focused more on moving to the rhythm, remembering the choreography, and hitting his marks on the stage. I’m doing the same thing the expert is doing, but it feels completely differently. Wii Sports is similar, although more abstracted. Wii Sports wants me to feel like I’m hitting a ball, but it doesn’t expect me to be doing the spatial calculations that an expert tennis player would, or planning my shots or my body placement. All the simulation is focused on the reproduction of the one mechanical activity.

While this philosophy lends itself towards motion control schemes, it isn’t necessarily limited to such. A game called Trespasser famously attempted a somewhat bizarre control scheme wherein arm and wrist rotation were each mapped to separate buttons, giving the player unprecedented control in manipulating the character’s arm, hampered only by the fact that it’s complexity made it almost unusable for most players. While not a particularly successful experiment, it does demonstrate that even highly detailed control systems need not rely simply on 1:1 motion detection. Abstraction, even abstraction to the point of unusable obfuscation, is possible. QWOP follows much in this tradition.

Part 2: Where I actually talk about QWOP.


From → General

One Comment

Trackbacks & Pingbacks

  1. Superlicious | Superlevel

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: