50% win rate

Discussion in 'Game Design' started by Bucky, Apr 20, 2017.

  1. Bucky

    Bucky Well-Known Member

    Keith got this result from information theory; if you only get information about your performance from winning or losing, the game can optimize your learning by tuning the difficulty so you win your better half of the games and lose your worse half of games.

    Even within those assumptions, the win rate should be just slightly above 50%. Because you're learning, your skill should be slightly higher than the game thinks it is, so you would win more games than you lose.
  2. SwiftSpear

    SwiftSpear Active Member

    We've had a recent obsession over the feed back provided by end state. Good games give you feedback a lot more frequently than when matches end. Civ works perfectly fine if you stumble your way through 45 hour long matches and never actually finish a game.

    I'm currently unconvinced that end state feedback is more special and important than any of the other feedback a game gives you.

    It definitely IS feedback, and feedback is valuable. End state feedback can be central to your game design. I don't see a good reason why it has to be central to your game design any more than any other type of feedback. I can imagine a game which never provides end state feedback and is still perfectly playable.
    KindFortress and adrixshadow like this.
  3. Nachtfischer

    Nachtfischer Well-Known Member

    Define "works perfectly fine"? I mean, in that case you either played without even looking at what the goals are, which would be fine if all you'd want is a toy-like fantasy simulation thing, but defnitely not fine for a strategy game, because to strategize you need a goal as an anchor point to strategize around. Or maybe you were playing around a specific goal, but then nobody ever reached it and so everything became kind of meaningless after all. What did you actually do? Did you play well? Which game did you actually play to begin with? Did your opponents play the same game? Strategy games need a goal and they need some form of outcome to evaluate your actions in light of that goal.
    keithburgun likes this.
  4. Erenan

    Erenan Well-Known Member

    Even if you interrupt the game and never finish it, moves are still good or bad with respect to potentially reaching the goal. What matters is how clear the impact of your moves is in that regard when you process them analytically. Maybe actually winning or losing gives you some measure of clarity on whether your moves were good or bad in that sense, but presumably you can sort of tell whether moves are good or bad in that sense even if you don't actually finish the match, right? I mean, based on mid-game heuristic signals and whatnot.
  5. Nachtfischer

    Nachtfischer Well-Known Member

    @Erenan: Sure, but that only works by you projecting what will happen, who will reach the goal first etc. There still needs to be a theoretical end state and you need to have sufficiently reliable information regarding your predictions. It's like deciding to concede a match early (or deciding that the AI in Civ concedes a match early because you're so far ahead). So bascially there is an ending in those cases.
  6. Erenan

    Erenan Well-Known Member

    Okay, wow, I didn't see SwiftSpear's "I can imagine a game which never provides end state feedback and is still perfectly playable" remark. Curious how SwiftSpear expects this to work without any end state feedback at all. I thought he just meant what I said in my post but apparently not.

    EDIT: Like, the only thing I can think of is if the game has an ostensible goal which is logically not reachable so it can only go on forever until you quit, so you can try to achieve the goal not realizing that it's impossible. Of course, in this case all the feedback you receive is actually mistaken because nothing really gets you closer to the goal. But you might erroneously feel like you're making progress.
  7. fodazd

    fodazd Member

    I can also imagine a game which never provides end state feedback and is still perfectly playable.
  8. Erenan

    Erenan Well-Known Member

    Can you describe it please?
  9. SwiftSpear

    SwiftSpear Active Member

    Your brain doesn't melt down and fail to learn in any way due to lack of feedback. You get better at the game, form heuristics, adapt to interesting situations nearly indistinctly from your response to games which end far more frequently.

    I don't think there's any argument that learning the game is hurt in any way until maybe we're talking about the really finicky optimization level problems where you're already playing pretty close to strategically optimal. Even there it's pure conjecture that there's any learning problem.

    A goal doesn't have to be an end point. We're completely comfortable with life goals like "make as much money as possible" or "travel as much as we can". Civ does have goals, it just doesn't let you win or lose in a neat little 30 minute loop. Not to say it's better than neat little 30 minute loop games, it's just not innately worse do to lack of feedback. It gives you just as much meaningful feedback.
  10. SwiftSpear

    SwiftSpear Active Member

    Stock market, the game, for example. It never ends, you just keep making or loosing money and comparing yourself to the performance of others in relative time blocks as a measure to rank your skill.

    [edit] I'm not arguing for no goals. Even goalless games don't really have no goals, they just let the player select their own goals (although this isn't really appropriate for a strategy game in any case I can easily think of). Strategy games do need goals, especially competitive strategy games. Goals just don't necessarily need to have a binary success or failure state.
    Hopenager likes this.
  11. fodazd

    fodazd Member

    When you get feedback for each "round" of play (like Poker, Scrabble, Civ, most fighting games, some strategy boardgames, etc.), then the learning process is not actually linked to the end state. The end state is just an arbitrary cutoff-point. Remove it, and you have a game that does not give end state feedback and is still perfectly playable.
  12. SwiftSpear

    SwiftSpear Active Member

    I mean, many round based games break if you try to play them forever, they naturally wind up to a stopping point.

    I want to be VERY clear, I'm not arguing there aren't other reasons why a tight 30 minute game play loop isn't preferable for a strategy game design. I'm arguing feedback is one of the least relevant reasons to have a 30 minute tight game play round. I can spit out a half dozen reasons why a short discrete game is preferable, all of which I think are better reasons than feedback.

    What's bothering me is this thread is like "well, game end feedback is so important lets draw a bunch of other inductive conclusions from that assumption". Feedback is one of the less important reasons to prefer developing games of this format.
    Last edited: Apr 21, 2017
    Redless likes this.
  13. adrixshadow

    adrixshadow New Member

    Garbage. It has nothing to do with information theory and everything to do with player psychology and motivation.
    The games are not played by robots.

    Besides I am of the opinion that the 30% winrate is the best since too much winning leads to a scrub mentality. Only when they suffer and know they suck do they start to Git Gud and learn properly.
    Nobody starts good at playing an instrument. Their winrate when they start is 0-5% and they increase by learning but that doesn't make them quit.
    Since a non-competitive game tends towards 100% as you become better by your own merits, since the challenge of the system is not infinite it makes your achievement even more sweet.
    As for 50% match-making that is absolute heresy. If every match becomes a coin toss you aren't getting any feedback on your skill at all, you just muddy through with your current skill.
    Instead have a tiered system where the bottom has around 30%winrate and the best around 60%. Have an Quake 3 like intro cinematic whenever you reach the next tier so that you know you are going to get your ass handed to you. But that also makes it sweet when by your own effort you dominate everyone from your tier for a time.

    Also my opinion is just as legitimate as everyone here since its all unsubstantiated shit anyway.
  14. SwiftSpear

    SwiftSpear Active Member

    You're being particularly uncharitable to his position. He's arguing from the position of a highly developed ELO system already in place. You're pragmatic assumptions of "the way quake works" being ideal are primitive at best. Ranked systems are clearly better on a per player basis than unranked systems. Granted there's a bunch of problems with them still.
  15. Batlad

    Batlad Well-Known Member

    That's not actually how rational discussion works. People present arguments and their evidence. If the evidence isn't good enough then no one will accept the argument and you should probably abandon it as well until new evidence comes along.

    Information theory is relevant and so is player psychology, ignoring either one would be a terrible mistake, because player psychology interacts directly with and processes information. Instead what we need to do is find the right balance between the two and see where the implications lead us. Its also fine to let some one to follow the implications for one aspect for a while before we see how it jives with the other.

    Now as for my opinion.

    I think having an unknown end condition is problematic (if there is an actual end condition), if there are multiple things the player's might mistakenly optimise towards. As this is just bad communication on behalf of the designer.

    I think though that once a player has an end goal in mind (or at least a heading direction) the internal workings of the game can provide a lot of feedback on how to get better. So post game feedback isn't the most important concern (as swiftspear claims). But, I think if you're going to provide post game feedback you should make sure it is as effective as possible. So while 50% win-rate feedback is an improvement in terms of post game feedback, if the cost of implementing it decreases the within game feedback by a larger amount. i.e. the per game feedback is reduced, then we probably shouldn't do that.

    So I guess what we should be optimising is per game feedback. Though maybe even better is the amount of feedback we provide per unit time.
    Last edited: Apr 21, 2017
    Redless, Bob Shelline and deluks917 like this.
  16. Batlad

    Batlad Well-Known Member

    I forgot to add that feedback throughput is not the actual value we are optimising, but the amount of feedback a player can actually utilise (thanks swiftspear for reminding me). So don't overwhelm a player with feedback, give them the right amount of feedback that they are ready for. Having opt in feedback seems like a good way to do this.

    Also misinformation is super bad and maybe should be avoided over having good feedback.

    Because of these things maximising feedback efficiency per unit time, might be a good approach.
  17. adrixshadow

    adrixshadow New Member

    And I say my 30% winrate is better then a 50% winrate, how are you going to argue against that?
    Without evidence it is a moot point because its just number we pulled out of our ass without taking into account all the factors.
    And I am saying ELOs that tend towards 50% winrate constantly is the most evil thing imaginable. It sucks the fun out of everything.
  18. Batlad

    Batlad Well-Known Member

    Everyone else here is able to engage with this discussion. If you're not able to understand what constitutes evidence then maybe sit this one out.

    Edit: Maybe while you're there you could listen to or read some things.
    Last edited: Apr 21, 2017
  19. SwiftSpear

    SwiftSpear Active Member

    You're just making a statement, you're not exploring the reasoning behind your claim at all. ELO systems seem really good to me, I've enjoyed participating in them in the past. Better than Quake 3 anyways. I can't prove you actually like what you claim you dislike. But your feelings are also totally irrelevant to my opinion on this matter. What evidence are you expecting? All the most popular competitive games use an ELO style ranking system to match make.

    ELO systems solve the problem of players quitting the game because it's too frustrating to constantly be the loser. They solve the problem of good players reaching a point where they can no longer be challenged by the game in casual play. Psychology has indicated in the past that people prefer to believe they're better than they actually are. Seems to me people will be more compelled to play a game they win more than they lose.

    It's not good enough to just say "I don't like X, do Y instead". You have provided no reason why 30% is good other than it feels right to you.
  20. adrixshadow

    adrixshadow New Member

    50% winrate on average is a flatline, there is no ups there is no downs, every other player engagement theory says this is wrong.
    Every match will be a coin toss and their expectations is no matter what they do it will always be a coin toss,they play bad they play good it is all a coin toss. That breaks whatever feedback heuristic they use.
    Players feel good when they actually improve, to improve they need to be motivated to learn so a roadblock is necessarily and when they achieve it they should be rewarded with consistent victories for a time.
    THIS IS NOTHING NEW. ELO systems already work this way, only that they are far too granular, they need their range expanded.
    Ranks should be embedded in the minds of players. Whenever they reach a next rank it should be the next challenge to overcome.

    Furthermore Ranking by itself is not good. To really drive the player's expertise they need coaching. Things like social dojo's or clans where you are given tools like recording and analyzing matches and high level players can comment and find the weakness of lower players and what to do to improve is more important.
    All professional sports has coaching, with games we could easily create virtual structures like dojo's and keep old professional players relevant, they can make training available for a wider audience.

    In games developers have always used systems unthinkingly, while it works somewhat that does not mean it is the best we can come up with and should stop thinking.

    If you believe that 50% is perfect then you literally stop thinking. I do not believe 50% is perfect, maybe its 30%, maybe its 90%, maybe the question is stupid from the start and is the wrong question and other factors at are play.
    Last edited: Apr 22, 2017

Share This Page