Against score systems (and for success and failure)

Strategy game designers should start thinking about alternatives to “score systems” for their games. In this article I will talk about how and why we use score systems right now, what their weaknesses are, and how we can (as well as why we should) move beyond them. Much of this article is written with respect to designing single player strategy games, but the theory absolutely applies to multiplayer strategy games equally.

Score Systems in Videogames

Score systems have been relied on by all kinds of interactive systems designers since the beginning. Early videogames such as Pac-Man and Galaga had high score boards that players would compete for places on, whereas Super Mario Bros. had a score feature as a sort of extra added feature that really serious players could try to maximize if they got tired of beating the game.

It goes back further than that, of course. Pinball, which laid many of the foundations for videogame design tropes, also used a score system, not to mention some ancient games such as Go. Today, there are strategy games that use score systems, such as games like Civilization, Rogue-likes, or my own Auro: A Monster-Bumping Adventure. They’re appealing to designers because they’re so simple to implement and design; you can pretty easily take just about any simple toy/sandbox activity and slap a score on it, and then it almost instantly feels a little bit more competitive, a little bit more replayable, a little bit more “strategy-game-like”.

I think I can classify the usage of score systems in videogames into three categories. Later, I will propose a fourth.

Category #1 – “Completion Systems”

This type is least relevant to the current article, because its score (if it even has one) is more or less irrelevant. These are systems that may have a score, but also have some other unrelated binary goal that most people pursue (Super Mario Bros., arguably Rogue-likes, any story or puzzle-based system). To the degree that Super Mario Bros. states its goal, I believe it’s to “rescue the Princess”, and I think that that’s how most people play the game. Rogue-likes perhaps a bit less so, since completing them is often very difficult. But most people do describe Rogue-likes as “really hard” and say things like “they’ll kill you over and over”, sort of suggesting that they believe completion to be the goal. It’s also worth noting here that videogames have trained most people to think along the lines of completion; “beating the game”, and so on, so if a system has both a score and completion, people will tend to go for the completion for that reason (as well as another that I’ll get into later). It is particularly useful for story-based systems and lends itself to a system that is designed to be “consumed”; played once and then discarded, more or less. It is equally useful for puzzles for the same reason.

Completion is different from a “score”, in that it’s binary. A score can be any number, but completion has just two states: completed, or not completed. Granted, these days, completion is a lot blurrier, with multiple endings, extra content and so on, but the principle of “beating the game” as a basic design idea still mostly holds up.

Completion is also different from winning. While both completion and winning are binary outcomes, completion marks a point where the user is largely done with the entire system, whereas winning is just a tiny notch in the long life of a usually highly replayable system.

Category #2 – “High Score Systems”

These are systems that output the number at the end of play, without giving you any kind of “win/loss” or “completion” feedback along with it. Classic Tetris comes to mind; one never “wins” (at least not in the single player mode) in Tetris. You play, and at some point the play session end is triggered, and at that point you are given a number that reflects the number of combos and a few other factors.

These have been the most prevalent among strategy/competitive-ish kinds of videogames. Like I said before, it’s largely because it’s easy and extremely flexible. It’s much harder to have a “balance problem” in a system that just spits out some integer at the end, than if the player’s goal is a binary win/loss. Maybe the score will be a few points higher because some strategy or component is too strong, but who cares?

Category #3 – “Score Test Systems”

These are systems that do have a binary win/loss outcome, but wherein the outcome is calculated using score. So, there is some score threshold

These are quite common in multiplayer games. Pretty much every single popular sport—football, baseball, soccer, hockey, you name it—uses this method. They are also equally common in designer boardgames, particularly in Eurogames where “victory points” determine the winner (Puerto Rico, Agricola, Through the Desert, etc). Also, notably, Go.

In single player games, you very rarely see this model, however. Civilization counts here because one of the victory conditions is to reach the end of play (some specific year like 2100 or something) with the highest overall score. With Auro: A Monster-Bumping Adventure, I have tried to really hone in on this idea, and I built a whole single-player Elo system making sure that no matter where the player is at skill wise, they get a match that is balanced to their level.

Auro: A Monster-Bumping Adventure

More on Score Test Systems

For the past few years, I sort of accepted the idea that Score Test Systems are basically the optimal design pattern for strategy games. It was only in the past year or so that it started to dawn on me that there are issues there and that we might need to move away from score systems in general in strategy game design.

But first, let me explain why I thought (and still think) that of the three listed systems, score test systems are the best. I will do so by demonstrating the problems with Category 1 and Category 2.

Category 1, completion systems, are quite obviously incompatible with strategy game design. Strategy games are inherently about the pursuit of mastery, and should be deep enough to be played many times. While ultimately, even strategy games can technically be “completed” through “full mastery” (or even close-to-full mastery), that’s not their design intent.

Category 2, high score systems, are a bit more interesting. There are a lot of practical problems with high score systems, such as the problem of ever-expanding match length in Tetris or Rogue. Many of these issues, including that one, can be overcome. But the fundamental problem with high score systems which cannot be overcome is that they are not even contests. Using my system of interactive forms, a videogame that uses a high score systems is a toy, not a game. But let’s talk about what that actually means and why it matters.

Winning and Losing

A strategy game is an interactive system that has some kind of goal. That’s got to be the most uncontroversial way to talk about what a strategy game is that I can possibly come up with; probably literally everyone would agree with that statement.

Let’s get a little bit more granular. In my system of forms, I consider a game (strategy game) to be a type of contest. That means that you are pursuing a goal and being tested on whether or not you meet that goal. You either pass, or fail to pass that test, by meeting, or failing to meet that goal.

You might ask, why is this “testing” component important? Why can’t we just give the player a goal, but then not necessarily have the testing aspect? For instance, you could (and most high score systems do) have a un-testable goal like “get a high score” or “survive as long as you can”. These “goals” ask you to maximize some value generally, without any sort of “check” on the other end.

On a practical level, the response to this would sound like: well, at what point is a score considered “high”? How long do I have to survive before it qualifies as “as long as I could”? If I survived for one second, was that as long as I could? If I survived for 10 hours, was that as long as I could? At no point does the player get any feedback with regards to the goal.

It is undeniable that human beings can follow a general command to “maximize” something, but this is really more of an exploratory “finding of the edges of the system”—a toy interaction—than it is any kind of test. I would call this something like a gameplay guideline or a suggestion, rather than a goal.

While it may meet some definitions of the word “goal”, in terms of how we use the word interactive systems generally, it really doesn’t make sense to say that a general command to maximize something qualifies as a goal, because of the lack of a “test” or “check” aspect. Even completion-based systems have a clear, enforceable, detectable state where you either have or have-not completed the thing. These are more of a test than a toy (and indeed, puzzles add a “goal” to toys in my interactive forms).

In short: high score systems do not have a goal, and they can not be strategy games, contests, or even puzzles. They are formally, and in practice, toys, even if they may have some of the other apparent trappings of strategy games.

Why is the “Test” Important?

Okay, so I’m calling them “toys”, but so what? A better question might be, why do we care? Why do we care that we are “testing” the player rather than just giving them a general gameplay guideline?

Well, firstly I should say that there is a place in the world for sandboxes, simulators, and various kinds of toys. I am not saying “don’t make toys”. What I am saying is that toys are distinct from strategy games, as a form, and if you’re trying to make strategy games but using a high score system (which happens quite often), you’re undermining your strategy game massively. You’re making a toy, and if you knew it, you could make it a much better toy, free from the kind of restrictive, careful “balance” that strategy game-design thinking tends to push you towards.

But why should we preserve the idea of a kind of system that is based around a kind of test? What is the value of the “failure” and “success” states? I think people on the one hand take these things for granted, especially when it comes to multiplayer games (largely for social, and not game design reasons). On the other hand, I think they find these qualities mostly or totally unnecessary for a single-player game. I mean, who cares if you won in a single-player game, right?

I don’t draw a big distinction between single player and multi-player when it comes to strategy games. (I’ve written before, actually, about how I think single player is actually ideal for strategy games.) In either case, you have a player making inputs in pursuit of a specific goal. The fact that the goal is a specific, enforceable thing means that the player can build strategy that (hopefully) leads to it. Strategy is impossible without a binary goal.

A point of clarification: are games like football or Go not strategy games, since they use a high score? Technically, these multiplayer score based games are score-tests, not high-score. They do have a clear, binary goal you’re going for: more than the opponent.

Strategy

We say “strategy” a lot, but it might be good to really take a hard look at what it actually is, so that we can figure out what it also is not.

In a weightlifting contest or a 100 yard dash, I think it’s fair to say that there is not a lot of “strategy” going on. I’ve had arguments about this from people who say “actually there’s a LOT of strategy”, and they go on to point out some tiny little fringe aspect that involves strategy. Okay, yeah, for any task there will always be different ways of achieving the task. But when we talk about “strategy” as in “strategy games”, I think it’s fair to say that the kinds of different strategies employed in these contests don’t really register.

In fact, if you are designing a contest, you make attempts to reduce the amount of “strategizing” possible. In an arm-wrestling match, there are lots of requirements about how you can sit, where your other arm must be, and various other regulations that ensure that strategy plays as diminished a role as possible. If we wanted more strategy in an arm wrestling match, we could have the players standing, allowed to use the other hand any way they want, maybe add some symbols on the ground that mean different things, and so on.

Generally, the strategy in an arm-wrestling match is “use all your arm power to push the other guy’s arm down”. There are a number of ways this is distinct from strategy-game-strategy, but the most important one for the purposes of this article is the quality of sacrifice.

Strategy in strategy games almost always involves doing things in an indirect way. It’s not just “do X”, it’s “do Y, which will allow me to do Z, which will allow me to do X”. Usually in these kinds of cases, steps like Y and Z are a sacrifice, or an investment. You’re spending some resources now for a benefit later. Most possible investments are risky; you don’t know the solution to the game or what’s coming, so your invest may not pay off when or how you expect it to.

Knowing the structure of the match (how long, roughly it will be) and having a vision of the end goal (me reaching the target) allow the player to build a plan that involves those investments or sacrifices and leads to victory (hopefully). It allows the player to use strategic long arcs.

So let’s go back to our “maximize your points” idea, where you don’t have a binary goal. Now, how do you build strategy? There are two options:

  1. You choose some number of points to “go for”. In this case, you’re basically house-ruling a binary goal! Players tend to do this pretty much all the time if they’re playing a score-based strategy game. Haphazardly, on the fly, they prescribe new binary goals, and if they reach that one, they make another one. (Kind of like how people act when they’re playing with a toy.)
  2. You just play as conservatively as possible, making only the most safe, “guaranteed-to-be-good” investments.

If you don’t have a target number of points you’re going toward, it makes no sense whatsoever to take any risky investments. Investing only makes sense with regard to some known span of time and some target amount of resources. Little investment means few strategically-meaningful long arcs, and that means limited depth. Therefore, a binary goal is more or less a requirement for strategy games.

That’s been my understanding for a long time, and what I notice about having a binary goal is that it makes design a lot harder, but in a good way. It’s harder because you’re now fitting parts into more of a structure; as I say in Clockwork Game Design, the goal is the anchor around which your other mechanisms revolve and get their meaning.

Category #4 – Scoreless Tests

One of the failures of Auro was the fact that it lacked long arcs. I never made the conscious decision to do this, but Auro became a purely tactical game, and while it is a deep game purely on the tactical level, I regret that I wasn’t able to utilize long arcs in a better way to make it a deeper game.

Practically speaking, because the arcs are so short, this means that Auro is sort of arbitrarily long in match length. As people have noted, you can kind of express a full Auro match in one screen worth of monsters (maybe two). Everything that’s going to happen happens and resolves in about two screens, except for score, and health (which really can be looked at as “the opposing score”, in a way).

This experience, plus many conversations I’ve had on the topic, plus the experience of playing many score-based videogames has led me to believe that even if you use a binary goal with your score-based game, scores are still going to be causing you design problems. One of the reasons the “only short arcs” thing happened with Auro, I think, was the fact that I was working on a score system to begin with.

I’m not saying that because Auro was a score based game, it therefore lacked long arcs. You can have a score-based test system which does have long arcs, like Go for example. What I am saying is that in general, a score-based design will tend to lead you toward smaller arcs and a more homogeneous system, because scoring is almost inherently about repetitive collecting of a resource.

Thinking of your game in terms of points, and these short loops of getting points, I think lends itself to a game that repeats many short arcs. It also lends itself to thinking of a game as sort of arbitrarily expandable. Indeed, any score-based win/loss game is arbitrarily expandable—just increase the number of points required to win! The inherent nature of “gathering points” is, perhaps, less suited for a strategy game than for a contest. A strategy game is a structured thing with a beginning, middle and end. A contest is a measurement.

But what are some alternatives? Well, I’ve already started to deploy one with the next Dinofarm Games game, Alakaram. Alakaram is sort of a spiritual sequel to Auro, but with a much larger map. Killing monsters creates zones of “influence” that decay over time. There are six totems throughout the level, and in order to win, you have to capture all six at once, which requires moving around the level in a smart way and planning where to place and upkeep your influence, when.

There are other examples. Chess is actually one of the most popular. Chess is not about getting a certain number of points—it’s just about capturing the opponent’s king.

Another example is League of Legends, maybe, with “destroying the opponent’s nexus”. With multiplayer games, it seems a lot less weird to think of non-score-based binary goals. (Leave me a few other examples you can think of in the comments!)

My other game, Push the Lane (which currently has a Kickstarter running, check it out), is currently still a score based game, but I’m currently thinking about doing something to reduce the score thresholds to something like 3 or 4, or change the score based nature of it completely.

If you take nothing else from this article, I hope that you at least think twice about how easy it is to put a score system into a game; how little balancing is required to make that “work” should suggest to you that you’re not doing what could be done with your system.

To clarify, I do think it’s possible to have a score-based binary goal game that is an optimal strategy game design. But I think you have to be really careful with it, and I am also pretty sure that if there is a target score it should be something very low – more like 3, or 6 or something that feels more discrete and less continuous. It is the continuous nature of the score that leads strategy game designers to think more like contest or toy designers; to think less structurally, and to fail to achieve the depth that their system otherwise would be capable of.

Thanks for reading!

Special thanks to Aaron Oman, a Patreon supporter. You can also support this article and articles like it by supporting my work on Patreon.com.

  • Isaac Shalev

    Untestable goal systems allow players to define milestones and shit them dynamically as they achieve them. The designer declaring some binary win/loss condition somewhere along the line (survive 20 min, say) doesn’t change very much. If you’re struggling to survive more than 5 min, the fact that there’s some success state at 20 isn’t meaningful, other than to communicate to you that it’s possible. But there are many ways to communicate that without delcaring 20′ to be the win state.

    I hear what you’re saying regarding the toylike nature of play when you’re simply competing for a higher score. But this same behavior pattern exists in games with binary testing. If you’re playing that kind of game at a low difficulty level, the existence of a win/loss state is irrelevant. You’re not going to lose. The question is only how much you’re going to win by. You’re literally toying with the game. So it’s not the binary-ness that we’re after, it’s the testing. We need the condition we’re striving to achieve to be contingent, uncertain, but within the sphere of the possible. Which leads to the inevitable conclusion that if we fix such a condition at some static point, we have created something which is a cipher for those for whom the goal is entirely unreachable, a game for those for whom the goal is well-calibrated, and a toy for those who have well surpassed it. In other words, ‘game’ is a relational word, not a defintion that can be fixed to the system itself. It’s validity and descriptive power is grounded in the quality of the interaction of the player and the system.

  • harlan entler

    So with your definitions, Go is not a game, and doesn’t involve strategy?

  • CraigStern

    This is interesting, but I want to hear more about why you think a point-scoring system necessarily induces players to “play as conservatively as possible, making only the most safe, ‘guaranteed-to-be-good’ investments.” Isn’t that really a matter of how you design the scoring system and how you design the levels? As in: can’t you put optional high-risk, high-point-reward side objectives into your levels to force players who want to achieve the highest possible score to challenge themselves?

  • Jake Forbes

    Thanks for elaborating on the topics. On several occasions you’ve made me rethink what constitutes a good scoring system so I’m glad to have read your latest thoughts here.

    You asked for examples of scoreless strategy games. One of my favorite mobile games, Dream Quest, sort of fits this bill with pretty unique win conditions.

    Dream Quest is a rogue-like made up of simple card battles with the meta-game being mostly about deck building. Victory has three distinct tiers reinforced by the achievement system: 1) Clear the dungeon and get to the final boss. 2) Knock 25% off of the boss’ HP, and 3) Kill the final boss. The first condition is a challenge at first, but moderate to easy once you learn the ropes. The second condition is the realistic win condition that remains satisfyingly challenging even when you have played dozens of games. The third is nigh-impossible to defeat with any consistency and is effectively a stretch goal to check off once. The game’s achievement system reinforces this very well by having conditions 1 and 2 reinforced with distinct achievements per class, while the 3rd condition is a singular achievement. It’s a little weird where “Victory” is just making a strong showing against the boss, not defeating it, but whatever, it works.

  • Jake Forbes

    Another aspect of scoring and win/loss that is especially relevant to games with victory points is that players often have a great deal of control on determining when the game ends. In a game like Puerto Rico, Race for the Galaxy, or most any Euro without a fixed turn schedule, there are strategies that reward forcing an endgame quickly and others that benefit by delaying that end. It can be problematic in a some poorly implemented cases where stalling can drag past usefulness, or where it’s effectively an act of desperation for a lucky comeback, but it can also make for compelling strategy when your chance at victory depends on properly anticipating the endgame.

  • Go is a game and does involve strategy because it has a score based binary goal (getting more points than the opponent).

  • Oh yeah. I meant to play DQ, never got around to it. Thanks for the suggestion.

  • harlan entler

    What if I’m playing tetris with my buddy, and we are trying to get a higher score than the other?

  • Then it sort of works as a score test, but with some weird behaviors. There you have a “target score”, but what’s weird is you kinda keep going after you’ve already won the game. The game should know what the target score is and, I guess, end when you get 1 point more? I don’t know. Some real game design would be necessary to come up with good coherent rules for that.

  • harlan entler

    you can just make it a game of horse, where it goes back and forth until one person isn’t able to exceed the other’s score.

    but I feel that the definition is very strange, if such a small adjustment changes the nature from game to toy so easily.

  • Jesse Welton

    “If you don’t have a target number of points you’re going toward, it makes no sense whatsoever to take any risky investments.” I reject this assertion. There are a number of ways to encourage risky investments by rewarding offering the player risk-vs-reward tradeoffs. Offer challenging optional objectives. Increase a score multiplier for progressing faster than is safe. Even simply changing to a pure high score system can encourage riskier behavior: Take Go, and change it from a score test to a game played to maximize your score against a fixed AI. All the same strategic considerations are present, except that now, when you are ahead, you still have an incentive to make risky plays to maximize your score instead of playing safer stabilizing moves that merely retain your lead.

  • sedj

    Unfortunate typo aside, I think this is a really insightful comment. I generally like Keith’s definitions of “toy”, “game”, and “contest”, but Isaac has an interesting point about those definitions in relation to the level of skill of the player.

  • Isaac Shalev

    Hah! I had to re-read my comment a few times to find the spelling mistake.

  • MichaelSinsbeck

    I agree with you. I think Keith’s argument is incorrect here. A point-scoring system does not necessarily mean that the match-length will change with player skill.

    As a simple counter-example: Take Auro and remove the target score and replace it by a point-scoring system. In all situations I can think of, I would play exactly the same way in both variants. Removing the binary win-condition would not make the strategic thinking disappear.

  • Yeah, that’s true about Auro, but that is a problem with Auro. Auro has nothing but short arcs to begin with. It’s purely tactical.

  • Max Hospadaruk

    first, he wasn’t saying that uncapped high scores *always* make players risk averse, just that they can. Just one inch above the text you quoted he’s list 2 different reactions players *can* have to uncapped scores.

    Second, your examples are basically just re-stating his 1 and 2:
    – “Offer challenging optional objectives.” which we either arbitrarily choose to complete (#1) or safely skip (#2)
    – “Increase a score multiplier for progressing faster than is safe.” so playing safely but for longer will still get you to the same score? see #2…
    – “…you still have an incentive to make risky plays to maximize your score instead of playing safer stabilizing moves that merely retain your lead.” do you? assuming in that system that a loss to the ai results in a score of zero, unsafe plays are only wise if (#1) you’ve set an arbitrary score goal for yourself- otherwise the best way (on average) to maximize your score is to secure as big of a “guaranteed” win as you can without ever risking a loss (#2).

    I think your objection actually has more to do with an unwritten #3 that Kieth didn’t mention in the list : risk taking can be fun when there’s nothing at stake, even if statistically it’s not a sound “strategy.” This is true, but it doesn’t change the fact that it is the players sense of fun driving the risk behavior, not the nature of the score system.

  • Max Hospadaruk

    feel like he covered that with
    “1. You choose some number of points to “go for”. In this case, you’re basically house-ruling a binary goal! Players tend to do this pretty much all the time if they’re playing a score-based strategy game. Haphazardly, on the fly, they prescribe new binary goals, and if they reach that one, they make another one. (Kind of like how people act when they’re playing with a toy.)”

    this is what a player is doing when they “want to achieve the highest possible score to challenge themselves”

  • Max Hospadaruk

    yup, this is a great comment – a system might move through many of these definitions depending on a player’s skill.

    I think it actually connects pretty well back to what he’s saying about score systems maybe not being the best way to test strategy games in every situation – for all its faults, chess never truly becomes a toy even to great players because there will always be another game where the binary win condition eludes you.

  • Jesse Welton

    Keith’s #1 and #2 are two possible reactions players can have to an uncapped score, but a player does not need to choose a target number of points in order for playing to maximize score to encourage risky behavior.

    – Completing optional objectives contributes to maximizing score without a target score. Since a player need not adopt a target score, this is not #1. Since a player who wishes to maximize their score should attempt the challenge if they can complete it, even if they may sometimes fail, this is not #2.

    – In a game with a score multiplier, it is not necessarily possible to reach the same score by playing longer. The game could be finite, or could become more difficult over time without increasing the multiplier. In either case, the player will eventually be unable to continue racking up points. Maximizing your score requires taking on the extra risk of increasing the multiplier early in the game.

    – In the Go example, your assumption that a loss to the AI results in 0 points is unwarranted. 50 points can be worth 50 points whether the AI has more or fewer. You also assume that the player is attempting to maximize their average score. I think players generally emphasize their highest scores over their average scores in a high score system. However, even under your assumptions, yes, sometimes you can absolutely increase your expected win margin by playing riskier moves than you should if your objective is only to have more points than the opponent. You don’t have to set a goal, you just have to ask yourself, “Can I make more points by playing A, or by playing B?” A is a safe play that simplifies the situation and leads to a straight-forward end game. B starts a fight which will cost you points if you lose the fight, but will cut off and kill a large group earning you many points if you win it. Even if B could lead to a loss, if the potential gain is big enough, you might still choose it if you want to maximize your expected average score.

    I agree that players will often adopt risky play for the fun of the challenge. And I’m not going to attempt to maximize my score if it is not fun for me to do so. But this is not about that. You should be able to see that in the examples above, it is strategically sound to take risks in order to maximize score.

  • CraigStern

    I believe the “you” in that quote is the player, not the developer. (Hence, the reference to “house rules”–if the developer puts a rule into a game, it’s not a house rule, it’s simply a rule.) And if a player chooses not to go for the maximum number of points in a score-based system, then that’s the individual’s choice not to maximize their score–it’s hard to see how the mere presence of a scoring system induces them to play conservatively.

    Since I’m thinking about it again, a scoring system that rewards difficult achievements without penalizing for failures might be another design approach to avoid encouraging conservative play (e.g. a system which awards points for each objective achieved, but does not deduct from the score for, say, damage taken or units lost).

  • Max Hospadaruk

    “Since I’m thinking about it again, a scoring system that rewards difficult achievements without penalizing for failures might be another design approach to avoid encouraging conservative play ”

    this I really agree with. I was generally thinking about high score systems with a “lose” condition (like tetris or pac-man). when you take that away you definitely enourage more experimentation – and the game is even more definitely in the “toy” category of being more about experimentation than strategy (not an inherently bad thing ofc)