Game Design (or related) Articles

Discussion in 'Game Design' started by keithburgun, Mar 20, 2014.

  1. Batlad

    Batlad Well-Known Member

    I wrote this in response to the discussion about score goals and binary goals. Mostly from discord.

    Okay so I've long been a binary goal skeptic, yep it could be true but I haven't encountered a compelling argument. Until @evizaer and other's said that a binary goal maximises strategic variety because it allows a range of strategies to be successful. This is a good property for sure. But I think this is still missing the point.

    Lets take a step back and think about the arc of learning and mastering a game system. To start off with you're exploring and just kind of testing stuff out, here you want a lot of strategies to be viable otherwise you won't have enough time to explore the possibility space before cutting yourself off from winning. Play here is very toy like. So you want a nice low goal, a binary goal would suit this purpose just fine. It is clear and directed. (I don't see an issue with a high score goal at this point).

    Mid way along the path to mastery you are comparing the viability of different strategies, and learning extra tricks that give an edge in rare situations. Here you want the pool of potentially viable strategies to be reduced, so you get clear feedback about which strategies are the most viable. A binary goal is still great as long as it has moved, some kind of par could work, but it would be hard for the player to get win percentages, if it changes too much match to match.

    In the final stages of mastery the player really is just figuring out what the very best strategy is. At this point the line between game and puzzle should begin to fade, there probably is only one strategy that expresses total mastery of a system. So the question is what strategy best represents mastery? The one that can attain the highest possible score, or the one that can win in the most consistently across all the variances in procedural generation (including during a match).

    I think this is where it becomes clear that we should favour the latter, essentially the one true strategy is the one that is the attains the highest average score across all procedural content (note there could be several that are equal but that doesn't matter for this conversation, and the best strategy is probably extremely adaptive to different state spaces). This could also be expressed as the highest win percentage when a binary goal eliminates the overwhelming majority of strategies.

    So I think our goal as designers in collaboration with our players should be to map the strategic space of our game system and design goals that push players towards the ultimate goal of the final strategy. Early on players should have a lot of freedom to experiment, but quickly we should push them to throw out their poorest strategies and use only their best, while constantly improving them. Then ultimately at the latest levels of mastery give them a way to measure their success in finding that one perfect strategy.

    Now with this backdrop I think we can begin to compare the design of different goals. But that can wait till next time.
    Last edited: Mar 14, 2017
  2. evizaer

    evizaer Well-Known Member

    I have three separate responses to this.

    1. I think you're missing proc genned difficulty in your analysis. The designer can amp up the difficulty until only one strategies can win. That's the test of the highest level of mastery when you have a binary goal. An SP MMR system, like what Auro has, can ramp you up to that point if designed right.

    2. You need to take proc genned variety into account. Some games will be harder than others and require different strategies to succeed. So there may just be one strategy that can beat the hardest possible level, but it's going to likely be different than the strategy needed to beat another similarly difficult level. Especially if there is much variety in what a level can be. I think this speaks strongly in favor of having a toy-like game mode so you can play with the kinds of variety the game has to offer.

    3. The fastest way for the player to learn a game system is to play with it in a toy-like fashion, but with clear goals in mind. Systematically experimenting the with the game systems is a far better way of gaining a thoroughgoing mastery than to chase after low-difficulty binary goals. There's a lot to explore in an elegant system. The binary goal just doesn't make you explore so much, definitely at lower difficulties. It'd be interesting to plot out goals that cover more of the useful skills needed to play at a high level. I think you definitely want some kind of reliable binary goal to be in the game for competition and comparison of skill purposes, but for teaching and learning we can find better ways of coming up with all kinds of diverse goals that stretch the player's horizons. This is not an argument against binary goals, but against there needing to be ONE and only one in a game.
    Batlad likes this.
  3. Batlad

    Batlad Well-Known Member

    1. Proc gen difficulty is assumed, but needs to be accounted for better in this line of reasoning. I assume that the goal is shifting either through more difficult scenarios or actors, or through harder goals. This is not a question of what type of goal to implement but how best to implement it.

    2. Proc gen variety comes into the fact that a strategy that is used is actually an implementation of a meta-strategy which understands which strategy to implement to exploit a given game state, locally and throughout a particular match. Essentially the ultimate strategy is all the viable strategies applied appropriately based upon the circumstances delivered by the proc gen.

    3. I think you're right, splitting goals into learning goals and ranking goals will be the way forward.

    So while I think that traditional high score systems are bad, and that binary goals which adapt to the player are better, I'm not sure that they are the best possible solution for all strategy games. I think this is still early days and knowing when and where to deploy different categories of goals to help players master a system, and that this should be an active area of research.
  4. Hopenager

    Hopenager Active Member

    I hadn't really thought about it in those terms, but I partially agree, though ideally the player wouldn't ever have to do calculation even outside of the match.

    I guess you can divide games into a "heuristic-inventing" part and a "heuristic-testing" part, and so far my whole thought process hadn't really consciously taken into account heuristic-inventing. I've been mostly focused on maximizing the fun of heuristic-testing. I see your point that by including a timer, especially a short one like I suggest, doesn't allow for the player to really engage much in heuristic-inventing during the actual game since to play well they need to focus all their time on thinking about what to do, whereas without a timer the player can just stop playing whenever they like and say "oh, what if I tried doing this thing that I've never considered before."

    I agree that it's kinda weird to split up the heuristic-learning and heuristic-inventing, but I don't think that if a game did that it would necessarily require a "large amount of offline study time." It seems to me that the heuristic-testing part is far more significant and time-consuming than the heuristic-inventing part. It's pretty easy to come up with strategies that seem like they might have a chance of working, but testing them to see whether they really do work is the hard part. Also, I think the most effective way to study is to play the game (or if that isn't true by default, games should be designed such that this is true. This links up with an principle I've thought a lot about before, that "the most effective way to learn about a game should be to play the game") and so generally in practice the amount of offline "studying" the player should to to be ideal should be very low.

    Additionally, it seems to me like the most effective time to do heuristic-inventing is generally a time very soon after to actually playing a match of the game. I think strategies often emerge when there is a situation where it is particularly easy to see that the certain strategy would have been useful, and that gets the player thinking about how that strategy might be useful in other situations. For example, if a bunch of enemies happen to get close to one another in XCOM, throwing a grenade will seem like the obvious best move when it comes to mind, and when they use it they see that it is indeed very effective. Then they might think "wow, that was really effective, I wonder if that would still be a good strategy if there were fewer enemies in that area" or "hmm, I wonder if I can make those types of situations more common by trying to a bunch of enemies into the same place." Generally, it seems to me that having concrete examples of gameplay situations in recent memory lets you invent strategies vastly better than you would be able to outside of the context of a match like in "offline study time."

    So, for a short game (<10 mins), I think that if the player takes a short time after the end of a match to think about how the game went, that will be enough heuristic-invention time. For longer games, it might be a good idea to have times where the action stops and the player has an opportunity to just sit there and think for however long they feel like, like making the next level not start until the player presses a button or something. This would give the player the opportunity to reflect and do some heuristic-invention throughout the game to make sure the situations are fresh in their mind, since unlike a short game it might be too hard to keep all of the situations you encountered in a match in mind for reflection after the match. Of course if you do that, you need to make sure the information the player has about the upcoming gamestate is VERY minimal in order to avoid having the ideal strategy be to just sit there and calculate in that time.

    I'd never really considered any of this before, so you've definitely given me a lot to think about. Thanks :)
    richy and Batlad like this.
  5. Bucky

    Bucky Well-Known Member

    In my experience, the best time for heuristic-invention is talking with another expert player. A close second is while writing strategy articles. In both cases, being forced to turn a heuristic into words tends to clarify both what the heuristic is and how to expand on it.
  6. Redless

    Redless Well-Known Member

    New "remedial" article up on my blog. I think this is a pretty basic and uncontroversial point, but one that many designers seem not to get, so I think it's perfect for, if someone would be so kind enough to put it there.
  7. lepreconor4

    lepreconor4 New Member

    It's been taking me ages to write anything so I decided I just want to get something out. Here is the deterministic solvability proof,

    I'll be working on describing an algorithm to solve this and the non-deterministic solvability problems at some point in the future. I also read Keith's article on the solvability of games and can't help but wonder if I might have been part of the cause of it. I disagree when he says computation is adjacent to game design. Similar to how mathematics and computation underlies computer-science I feel mathematics and computation also underlies all of game design. We are dealing with mathematical constructs and understanding of how they work is required if we are to make any of the grand claims we desire about turn timers, complexity and depth, randomness or win/loss ratios.
  8. cuc

    cuc Member

    Last edited: Mar 25, 2017
  9. Kdansky

    Kdansky Well-Known Member

    Obviously not an article, but still relevant: Day9 talks about Starcraft Broodwar, and the key sentence is this one: "Broodwar is not primarily a strategy-game."

    It's nice to see someone as mainstream as Day9 having figured that out too.
    liltiger likes this.
  10. Batlad

    Batlad Well-Known Member

    Hopenager and Redless like this.
  11. richy

    richy Well-Known Member

    Nachtfischer and keithburgun like this.
  12. keithburgun

    keithburgun Administrator, Lead Designer Staff Member

    Read the first few paragraphs of the @Batlad article and have some feedback.

    This is weird phrasing for two reasons. "Most" of the challenge? More than 50% of the challenge comes from the pressure of not losing --- as opposed to what? What's the remaining percentage?

    (Also it's weird that you use an image from Symphony of the Night, an interactive system which almost certainly shouldn't have "losing" of any kind.)

    I have no idea what kinds of games you're talking about here where players either win or stop playing. Some examples might have gone a long way. I suspect you're talking about puzzles?

    How does increasing the challenge "avoid the problem" of them "either winning or stopping playing"?

    This feels really disjointed from the stuff that comes before it. Probably needs a new paragraph? Did you write an outline for this article?

    You're cramming way too many ideas into each sentence. We're flying by concepts faster than they are getting fleshed out or explained. Somewhat related to the information-cram is the fact that this sentence has a "however" and a "but" in it - you generally want to avoid having more than one of those per sentence.

    The AI determines the loss condition? Also, even if there is an AI, the designer still has to impose the loss condition.

    How are timers a loss condition? I assume you mean "once the timer runs out, you lose, assuming you haven't won", but that should be explained.
  13. Batlad

    Batlad Well-Known Member

    The first real paragraph is very crammed because instead of spending a third of the article setting up the background which is not used for the rest of the article, I wanted (and was advised) to get through it as quickly as possible.

    Some goals are innately challenging to reach, they might require you to learn calculus or or some weird interaction. Now given enough time to pursue them, maybe you wouldn't consider them challenging any more. But, putting that limit in contributes to their being challenging.

    I just wanted something with a memorable loss screen. I kept the article intentionally broad so I could focus on the issues I think are interesting especially towards the end of the article.

    I'm using the colloquial definition of games here. Even most games that have loss conditions don't really use them as loss conditions, they just let you load a save or go back to a checkpoint. I'll change it to include an example.

    In a symmetrical game versus an AI, the AI winning is the player's loss condition.

    Edit forgot to address this:
    Adding a challenge eliminates the problem of triviality and time rather than skill determining outcomes. So the problem created as an implication of 'winning or stopping' not winning or stopping itself.

    Edit: also
    The but here is in brackets as an aside that does not warrant its own sentence, since it is a dependant clause. Here it is specifically used to highlight a counter example to the topic of the sentence, that AIs that support multiple skill levels of player are hard build.

    Thanks for the feedback, I'll split this paragraph as suggested and explain things more clearly.
    Last edited: Apr 17, 2017
  14. keithburgun

    keithburgun Administrator, Lead Designer Staff Member

    Yeah, but the designer has to define what that condition is.
  15. Batlad

    Batlad Well-Known Member

    Sure but my point is with a symmetrical game you can have have the goal the same for both sides. So you can avoid specifically designing a loss condition if you want. However, without an AI opponent you must specifically design a loss condition, if you want there to be one. Which you should.

    Edit: Okay I've updated those paragraphs based on your feedback. I'm hesitant to reduce their density though since as I've said they only form the backdrop for the article and are better off the topics of their own articles. Though they may not even warrant that.
    Last edited: Apr 17, 2017
  16. Redless

    Redless Well-Known Member

    For the record I love Batlad's article, my favorite article since Hopenager's generalizability and beyond. Great job @Batlad!
    Batlad likes this.
  17. Hopenager

    Hopenager Active Member

    richy, Redless, Batlad and 1 other person like this.
  18. richy

    richy Well-Known Member

    Agree totally re. the alternative to win and loss as feedback. One suggestion maybe is not use the word "score" in describing that kind of result, because it evokes "high score" type scoring. That kind of feedback doesn't need to be in the form of a numerical score as such anyway - it could be a grade or something else. (I remember getting graded "rookie" by a Gorf machine or something in an arcade years ago - we were Brits and none of us had ever heard the word before - we assumed it must be a good thing :D)
  19. Batlad

    Batlad Well-Known Member

  20. keithburgun

    keithburgun Administrator, Lead Designer Staff Member

    Final paragraph was probably my least favorite part (got a little GHASTLY for me!!) but yeah it's good! And I'm also really pleased to see this in The Atlantic - that's like a paper real people read!

    Comments are predictably defensive. It seems that "Reductionist" is really over-used in this kind of situation. Seems like anything other than "Mass Effect Rules!" is basically reductionist.

Share This Page