Theory: Test Decisions in the Right Context

Some small percentage of interesting decisions are always going to be fun. Some interesting decisions never are. Most, though, are fun only when given a context in which they can shine. It’s vital to understand what situation a decision needs, and to test your game accordingly.

Many well-known, big-budget games have decisions that work as gameplay specifically because of how they’re presented. Choosing where on the screen to click with a mouse, for example, is completely trivial in a turn-based environment that allows the player to take as much time as she wants. Yet, that activity supports the entire FPS genre, because it’s not at all easy to click just the right spot under extreme time pressure.

Fighting games are of the same ilk. Lots of fighting game decisions are easy once the player understands the game (“I should parry this projectile, that’s strictly better for me than blocking or jumping”). Others are more-or-less math problems: there are two options, with the first having a higher expected value than the second. On paper, the player alway makes the right choice. When one has to find the right move in a sixth of a second, doing the math or executing the correct action is a lot harder—and the difficulty helps turn what would otherwise be a trivial exercise into a tense journey through donkeyspace.

Both of those examples rely on time limits, but there are other contexts that can affect whether a decision is interesting. While in school, for example, I worked on a game loosely based on regattas. Players formed teams, and then teams rowed against each other to win a race. We were to discover the hard way that that has the potential to be interesting . . . so long as the game isn’t an entirely cooperative game about rowing in tandem.

If you’ve ever been out on a boat yourself, this probably doesn’t surprise you. Rowing can lead to beautiful scenery, it can be a kind of moving meditation, one can use it as a social event. It’s unusual, though, for the actual act of rowing to be intellectually challenging. We struggled for weeks to build a system that would make coordinating a team of rowers fun in and of itself. Two weeks went by with the team rowing boats using different systems, just by ourselves without any opposition, searching for a way to turn that into a fun puzzle.

Mercifully, that process ended (as is so often the case with design problems) when we gritted our teeth and took our not-really-working game out to playtest. The results were dramatic—it was great fun! Players enjoyed it, we enjoyed watching them play, everyone had a good time. What had happened, to bring about such a remarkable change?

All we did was change the context in which the same decisions were made. Throughout our internal design work we played a cooperative slice of our cooperate-to-compete game. In doing so we put all the weight on the puzzle, and the puzzle couldn’t bear the load. It wasn’t deep enough to be a game unto itself.

A competitive environment, on the other hand, made the relatively simple puzzle we had built deeply rewarding. It was hard enough that players sometimes made mistakes, and those mistakes now existed in a context where they mattered. At the same time, it was easy enough to allow for successful cooperation more often than not, and players enjoyed solving each turn’s puzzle in order to progress in the race. Competition made the exact same mechanics fun.

Our game didn’t have the wrong decisions; we’d just been testing those decisions in the wrong context. If you can predict that your game will work best when certain things obtain, be sure to test in that context. It’s worth saving those two weeks of heartache!


Theory: Doing RPS Right

Rock-paper-scissors (“RPS”) dynamics are sometimes held up as fundamental to game design, a core principle that makes balance possible. Taking rock-paper-scissors too far, however, can lead to reductive games that are only interesting during character selection. It’s vital to understand that what’s interesting and valuable isn’t the RPS, but the interesting decisions a good RPS setup can contribute to.

The Story So Far

To my knowledge, the popularity of RPS comparisons in game design started with an article on Sirlin’s old website. It’s worth reading the article in full (and I’d love to see it preserved on the redesigned site), but his core argument went something like this: rock-paper-scissors, as we all played it as children, has little to no strategy because there’s no basis for preferring one move over another. The opponent is probably choosing more or less at random, so there’s no way to predict what he or she will do, and you’ll just have to play randomly, too. However, it’s possible to inject strategy into rock-paper-scissors by giving each move different payoffs; if rock is worth 2 points and everything else is worth 1, you know that both players want to play rock, and you can use that knowledge to craft a plan.

Sirlin followed up with another article, this one on how RPS is implemented in fighting games. Again, it’s excellent and well worth your time (and also worth preserving). While I can’t trace the spread of its ideas, I can say that this article was extremely influential among fighting game players, and that in the years since its publication I’ve seen its arguments and conclusions repeated many times. I think it’s fair to say that after this article, RPS was off and running.

Today, RPS has become ingrained in game design thinking. However, saying that RPS with differing payoffs can be interesting doesn’t mean that it’s always going to be executed correctly. In particular, it’s not great when the only meaningful RPS decision is made early.

RPS Done Right: Interesting Decisions that Support More Interesting Decisions

Go back to Sirlin’s second article for a moment. He describes a series of situations in which both players make a choice, and one of them just loses. Attack beats throw, period, with the attacker doing damage and the thrower accomplishing nothing. It’s a hard counter, just what we would expect from RPS: rock doesn’t sort of beat scissors, it SMASHES scissors and gets ALL THE POINTS.

However, that decision isn’t the be-all and end-all of the match. Rather, a match will involve many such choices. Indeed, much of the interest of the game comes from the choices building on each other: “last time he went for the throw and got me, I think he’ll try that again even though it doesn’t do as much damage as an attack.” Evaluating moves becomes a multi-layered process, as one judges not just their value in the abstract (rock is worth 2 points) but also how the opponent seems to be responding to those values (this opponent is being tricky by playing lots of paper).

We see the same building of decisions in other good RPS games. Starcraft, for example, has units that beat other units—but players build many units over the course of a game, and can expect to skirmish several times before the match is decided. As a result, players will make a number of important and engaging building decisions. Over the course of a game one might shift gears in response to what the opponent is building, feint, condition the opponent to expect one thing before building something else, and otherwise make interesting choices throughout the game’s duration.

Magic has only one RPS decision, but it still uses that choice as a springboard for more decisions later. It does this by incorporating RPS into an important, but not definitive, early choice.

It used to be said that there were three broad strategies in Magic, and that they interacted in RPS fashion:

Aggro (“aggression”) beat Control, because it won before control could lock down the game.

Control beat Combo, because control stopped combo from assembling its Rube Goldberg victory condition.

Combo beat Aggro, because aggro didn’t stop combo from assembling the Rube Goldberg machine.

While the model spoke in definitive terms, however, making the right choice at deck selection (e.g., playing aggro against someone who was playing control) never actually decided the game. Rather, it provided an advantage which had to be carried through in play. The player on the winning side of the RPS decision still had to manage random card draws and the opponent’s resistance, which could still be potent even coming from a position of disadvantage. Making a good RPS choice thus offered advantage but guaranteed nothing, and so the game’s in-play decisions remained meaningful.

RPS Done Wrong: One and Done

For a contrast to Starcraft, Magic, and other good RPS implementations, consider a fighting game where some characters beat other characters RPS-style. That might be balanced. If Ryu beats Zangief, Zangief beats Dhalsim, and Dhalsim beats Ryu, then everyone has about an equal chance of losing and the game is fair.

However, that game has only one important decision: which fighter to choose. It’s all downhill after that, as one plays out the inevitable result of the character select RPS. None of the decisions in the actual game are very interesting, because Ryu is just going to clobber Zangief no matter what their players do.

Weighted RPS mechanics, then, are not an inherent good. They can in fact be very dangerous, locking in results and turning the rest of the game into going through the motions. If RPS is going to be used, it’s critical that the game not hinge on a single, early RPS choice, but rather that the RPS decisions create further interesting decisions as the game goes on.

This problem is not strictly theoretical. Several miniatures wargames have wrestled, or are wrestling, with the problem of an early RPS choice that dominates the game.

Minis games’ RPS issues revolve around how in-game armies are built. Broadly speaking, minis games have a system wherein heavy armor (tanks, flying tanks, whatever the game’s setting has) is largely invulnerable to infantry; infantry is good at dealing with anti-armor specialists; and anti-armor specialists destroy heavy armor. As a result, games can be decided as soon as the players show each other their armies; if one player has way more tanks than the other player has anti-armor specialists, the tank player will pretty much have the run of the field. It’ll be particularly egregious because that game will probably still take two hours to finish; two hours is a long time to have no interesting decisions.

We’ve seen a number of solutions to this problem over the past decade: allowing pieces to serve in more than one role so that players can always use all three of rock, paper, and scissors, for example, or trying to opt out of the RPS situation entirely by flattening the differences between armor and infantry. I’m sure we’ll be seeing more in the future. Even if the problem is entirely solved, however, its persistence over many years will stand as a reminder that leaning on the RPS tripod can leave a designer flat on the floor.

Keep Your Eyes on the Prize

Rock-paper-scissors isn’t the ultimate in game design; it isn’t even a good summation. What it can be, if done right, is a source of fascinating choices for players. The key to using RPS well is to always keep focus on the decisions, asking whether the RPS is making them more interesting—or less relevant.

Something Completely Different: Context-Sensitive Fighting Game Controls

File this one under “projects for a 25th hour in the day” . . . .

Years ago I saw a really excellent post—I can’t remember whether it was on Shoryuken’s forums or elsewhere—regarding the challenges of learning fighting games. The poster argued that one of the major issues for new fighting game players is that the buttons aren’t labeled in a useful way; they have thematic names that don’t express what they should actually be used for. At the time it was difficult to resolve that in a satisfactory way, but it strikes me that this problem could definitely be addressed with current technology.

The concern goes like this: most fighting game characters have attacks that serve defined tactical purposes. By and large they will have a long-range strike that controls the opponent’s movement; a quick and unpredictable attack that probes for weaknesses in the opponent’s defenses; a slow but powerful smash that’s meant as the last move in combos. One of the major hurdles for players who are new to the genre is recognizing that the moves available have these specific uses.

Part of what raises that hurdle so high is that the various attack buttons aren’t labeled “control,” “probe for weakness,” and “smash.” They’re “medium kick,” “light kick,” and “heavy punch,” which sound thematic but don’t do anything to explain how they ought to be used. New players thus find themselves without the information necessary to make decisions about which attack to throw out, and are obliged either to button-mash or to figure it out on their own. Either way, it’s a substantial barrier to clear before they can start to get full value out of the game.

Moreover, the knowledge a new player gains from one character isn’t transferable to another. Millia controls with kick and probes for weaknesses with crouching kick; Ky controls with slash and probes for weaknesses with kick. Picking a new character means starting from about 50%; the new player (hopefully) knows that different moves have different applications, but has to figure out which move is for which use all over again.

When controllers just had whatever markings they had, and the labels needed to be appropriate for all characters, this was a very difficult problem to solve. However, today we can do context-sensitive controls. Imagine something like this:

The controller is a tablet or smartphone. Instead of being labeled light, medium, and heavy kick, buttons have descriptive labels appropriate to the character. Not only is there a button marked “control opponent’s movement,” it’s the correct button for that character. New players immediately see that each attack is for a specific purpose.

What’s more, the buttons change. If the opponent jumps, attacks that aren’t useful in that situation have their buttons greyed out—the attack will still work, but it’s clear that that’s a bad button to press right now. Attacks that are useful will stay their normal color and will get new labels, like “anti-air.” In single player mode, the game can be paused so that players can look down at the controller and find out what might be useful to do.

Fighting games are, I feel, one of the hardest genres to get into—but also one of the most rewarding. Context-sensitive control labeling would make it a lot easier to access both what a specific character can do and how new players should think about their moves generally. And it’s so doable now . . . if there’s time . . . .

Theory: Marvel Contest of Champions and 2D Fighting With Few Controls

I like fighting games and I like comics, so I couldn’t resist giving Marvel Contest of Champions a try. If nothing else, I wanted to know what the control scheme was like; after years of playing fighting games on an arcade joystick, my thinking on how to control a game like that had gotten stale. To my surprise, I discovered that MCoC’s tap-and-swipe system works better than it seems like it would. There’s only so many things you need to build a legitimate fighting game, and tapping and swiping enable all of them.

MCoC is a 2D fighting game. That means each player controls a martial artist, and those martial artists fight back and forth on a flat plane. In this case the martial artists are Spider-Man and Captain America instead of practitioners of karate and muay thai, but the colorful characters don’t change the underlying gameplay.

2D fighting games have two core concepts that make them work: the attack-block-throw relationship and controlling space. MCoC features both.

Fundamentals of 2D fighting games

Almost every 2D fighting game I’m familiar with–I would go so far as to say every 2D fighting game released in the last 25 years except one–has rock-paper-scissors at its core.

Blocking (rock) nullifies the damage from attacking (scissors)
Attacking (scissors) does damage to an opponent who is trying to throw (paper)
Throws (paper) inflict damage on a blocking (rock) opponent

Much of the strategy in 2D fighting games comes from manipulating opponents into making the wrong choices, so that their damage is nullified by timely blocks and they are not blocking when the time comes for one’s own attacks. That manipulation is possible because the different choices have different payoffs; knowing what the opponent wants to do makes it possible to get into his head, predict his moves, and bait out the moves you want him to make.

2D fighting games also involve a battle to control space. When Ryu throws a fireball in Street Fighter, he takes control of the lower part of the screen; since the game occurs on a flat plane, the opponent cannot advance while the fireball is approaching. Thus, Ryu’s fireball prevents the opponent from taking the offensive. By controlling space, Ryu controls the game.

Not all 2D fighting game characters have fireballs, but they all have ways to control space. The player’s goal is to use each character’s unique tools to assert control over space, take control of the game thereby, and turn that advantage into a victory.

This video, made by David Sirlin, is a great visual explanation of controlling space. Take a look; the relevant discussion begins at 0:58.

The fundamentals in Marvel Contest of Champions

Everything one would expect from a 2D fighting game exists in MCoC. The rock-paper-scissors relationship is firmly in place; MCoC uses “heavy attacks” in place of throws, but the effect–damage inflicted on a blocking opponent–is the same. So too is the struggle to control space in evidence, with Iron Man’s repulsor beams standing in for Ryu’s fireballs.

What’s striking is how few “buttons” MCoC needs to accomplish those things. Movement is thoroughly simplified; players can only shift toward and away from the opponent by swiping left or right, with no jumping, sidestepping, or other movement options. Yet, “toward” and “away” are enough to create space for oneself and reduce the opponent’s space. Hitting the opponent is also very basic–tap, swipe, or tap and hold–but that’s enough to enable attacking and throwing, which are all that’s needed.

In some respects MCoC reminds me of Divekick, the “art game” of the fighting game world. Divekick is the one modern 2D fighting game without rock-paper-scissors; it’s all about controlling space, with a total focus on jumping into the air and positioning oneself to dive down on an opponent who’s trying to do the exact same thing. Although they play very differently, both games are about stripping away the cruft that has affixed itself to the 2D fighting genre in order to explore the essentials of how such games work.

(Well, MCoC is also about incentivizing spending using a freemium model.)

I’m always fascinated by the question of the most minimal thing that would count as a game. Divekick and MCoC are interesting because they push that boundary within a specific genre: they’re both trying to find the smallest number of elements one can include in a 2D fighting game while retaining the strategy and fun. The fact that they both use minimal controls to do so is surely interesting . . . .

Theory: How to Make Losing Fun

Part of the reason why we have so many sayings to the effect that “winning isn’t everything” is that winning is closely tied to having fun. Yet, it’s possible to make a game fun for players who are currently losing–even for those who have no hope of victory. Providing measurable goals losing players can meet separate and apart from overall victory enables them to walk away from the game with a sense of satisfaction.

The recent poster child for game design that’s fun even when the player is failing is Dark Souls. For those who haven’t played it, Dark Souls is an action game. A very, very hard action game. “Prepare to Die,” its ad copy declares, and die the player will. Many times.

Yet, Dark Souls can be a very satisfying experience even as it clobbers its player. Progress in Dark Souls is easy to measure; monsters are always waiting in the same places, and so one can always tell when one has gotten a little further. Last time the ghoul waiting in the hallway beat me; this time I beat it. Those tiny but clear bits of advancement let players put the controller down with a sense of accomplishment, even if the end of the game is still very far and many deaths away.

Dark Souls’ puzzle-like form–enemies are always in the same place, paths always lead in the same directions–allows for concrete sub-goals. However, there are other ways to introduce objectives that are satisfying even though they are short of winning. Role-playing games, for example, use story for this purpose. Winning might be tens or even hundreds of hours away, but the next chunk of plot and character development is much closer. If the player is enjoying the game and its story, reaching that intermediate point is a powerful incentive and satisfying when it happens.

Games that call heavily on player skill also have this dynamic going for them, although they approach it from a different perspective: they encourage the player to create her own goals. Fighting games, for example, involve a tremendous number of skills. Being good at fighting games is incredibly difficult, so much so that most players will never achieve it (myself included; I top out at a journeyman level). In other words, the vast majority of fighting game players will never “win.”

Fighting games are, nevertheless, deeply compelling, because even if a player will never reach the highest plateau she is constantly achieving things. The first time a player successfully does the forward-down-down + forward motion for a dragon punch feels great. Being able to do it consistently is even better. Comboing into the dragon punch from a standing close hard punch is better still. Pulling off Evil Ryu’s one-frame link to wipe out half of the opponent’s health in a single flurry of attacks is amazing! There’s a never-ending series of little goals fighting-game players can set for themselves, and they maintain players’ interest in climbing the next rung of the ladder even if the player knows its top will always be beyond reach.

These examples offer three different mechanisms by which players can have satisfying, engaging goals short of winning: a puzzle structure that allows one to see progress in concrete terms, a story that is doled out in limited amounts to leave the player wanting more, and a skill-driven model in which players take pride in each small accomplishment. What overall lessons can we derive from the examples?

First, a good sub-goal for keeping losing players engaged is measurable. The player can tell when she has met it. Achieving something isn’t as much fun when the result is in doubt, so there’s no uncertainty about the accomplishment.

Second, these goals are independent of winning. They may involve actions which are conducive to overall victory–hitting a combo contributes to winning a fighting game, and getting part of an RPG’s story is a step toward the game’s conclusion–but they don’t rely on reaching that lofty plateau, or even being ahead at any particular time. It’s possible to get the satisfaction of these achievements even if the game has turned severely against the player.

Third and finally, they’re desirable. None of these are mocking “most improved” awards. They allow one to progress along an interesting axis, even if it’s not the most competitive one.

It’s easy to make winning feel good. Getting losing to feel good is harder–but not impossible. The key is to provide measurable, desirable goals that can be achieved independent of beating the opponent.

Theory: Make the Right Choice the Default, Part 1

I love fighting games–Street Fighter, Guilty Gear, King of Fighters (especially ’98 and, for all its warts, ’03), Virtua Fighter, Capcom vs. SNK 2, Marvel vs. Capcom 2. The change list for Ultra Street Fighter 4 came out recently, and it reminded me of something I saw a long time ago–a design rule that I think makes a lot of sense but that many games, especially fighting games, get wrong. If a given option is almost always the right choice, it should be the default.

Street Fighter 4 is a good example of what happens when the default is the less-desirable option. For those not familiar with its genre, SF4 is a two-player game in which each player controls a single martial artist. The players use their chosen martial artist’s kicks, punches, and unique abilities (e.g., breathing fire or throwing rocks) to defeat opponents. SF4 is fun, popular . . . and has a somewhat silly way of handling players knocking each other down. It makes it hard to get up fast and easy to get up slowly.

In SF4, as in most fighting games, it is almost always best to get up as fast as possible after being knocked down. This is for two reasons. First, it gets the knocked down player back on offense more quickly–and being on offense is how you win. Second, and perhaps more importantly at high levels of play, the time a player spends knocked down is time the opponent can spend repositioning and setting up his or her next attack. Minimizing that opportunity is very important.

There are rare occasions when staying down is good. If the opponent comes at you with an attack that will meet you as you rise, it might be advantageous to stay on the ground. The attack will pass harmlessly over you, and then you can get up and counterattack. However, these situations are unusual; in most cases it’s still best to stand quickly and use your full arsenal of martial arts maneuvers to deal with the attack. (Fighting game aficionados will understand me when I say that you would rather quick-stand and DP.)

(Unless it’s a cross-up, in which case DPing might be wrong, but you still don’t want to be down, you want to get up and block backwards, since being down doesn’t stop them from continuing the block string and just turning it into a meaty.)

(OK, sorry, back on topic.)

SF4’s mistake is that it makes getting up slowly, which is almost always wrong, the default. If you get knocked down and do nothing, you will get up slowly and be at a disadvantage. Getting up fast, which you want to do at least 95% of the time, requires an extra joystick motion done with precise timing.

The fundamental problem with this is that it doesn’t make the game more interesting. Since you should do it virtually every time, it’s just adding rote behavior. Get knocked down, tap down as you hit the ground to quick-stand. It doesn’t even sound interesting when you say it!

Having slow-standing as the default also leads to what Mark Rosewater calls “feel-bad” moments. It’s entirely possible for a player to know that quick-standing is right, try to do it, and fail. Missing the input just makes the player feel embarrassed and frustrated. Since fighting games are often played online, where internet lag can cause the game to think an input was mis-timed even when the player did it correctly, these “feel-bad” moments can occur with substantial frequency.

Last but not least, slow-standing as the default makes the game harder to learn. Fighting games are not easy to play. They involve enormous execution barriers–it’s hard for a new player to get the fire-breathing and rock-throwing to happen consistently. Clearing those hurdles is only the beginning, because then the player is ready to start the real journey of learning fighting game strategy. That could be a book unto itself, but suffice it to say that to play fighting games well one must make split-second decisions in an environment of uncertainty. Saying to a new player “by the way, on top of everything else you need to tap down 95+% of the time when you get knocked down” is pretty rough.

I love SF4, but I can’t deny that it suffers from all of these issues. Quick-standing is a rote element of gameplay. I feel bad when something goes wrong and I miss it, especially when it seems like lag was the cause rather than an error on my part. It was a checkbox I had to spend time filling before I could “really” play the game.

OK, so the way SF4 does things isn’t ideal. Why is quick-standing as the default better? I’ll talk about that Friday.