What is "analysis" (as opposed to "calculation")?

Discussion in 'Game Design' started by MichaelSinsbeck, Nov 3, 2016.

  1. Hi everyone,

    new user here. I have been following Keith's work for a while, but now finally decided to register here and join the discussions.

    I would like to discuss and better understand the concept of "analysis".

    Keith wrote about it in his last article Minimize calculation (in games worth playing). In short, this article says that, in a game, the players should ideally perform as much analysis as possible and as little calculation as possible. Unfortunately, the article does not give a positive definition of what analysis is. It only says that it is the kind of thinking that is not calculation (which is a negative definition: a definition by saying what it is not):

    I think it would be useful to have a positive definition of the term analysis.

    I have actually proposed such a definition as a comment to the article, but haven't gotten a response. So what are your thoughts?
    keithburgun likes this.
  2. Nachtfischer

    Nachtfischer Well-Known Member

    MichaelSinsbeck and keithburgun like this.
  3. keithburgun

    keithburgun Administrator, Lead Designer Staff Member

    Sorry I missed your comment. Welcome to the forums! But yeah, those are pretty good definitions.
    MichaelSinsbeck likes this.
  4. Thanks for the quick response. In the following, I will work with these definitions:
    Calculations: Making a decision based on your understanding of the rules.
    Analysis: Making a decision based on observations/experience from previous games.

    By these definitions, even simple thought about the rules count as calculations. For example, in the German board game The Settlers of Catan, the goal is to gain victory points. After reading the rules, I can understand that I gain these points by building towns, and to build a town, I need roads. So one of the first things I will do is build a road. And I can even do this the first time I play the game. So, the thought process I just mentioned would count as calculations and not as analysis.

    My questions: Do you agree that the mentioned thought process counts as calculations? And if yes, would you rather have a game without any calculations? That would mean that in the first couple of play-throughts of this game you would act randomly until you see some patterns. Only then could you start becoming better at the game.
    Bob Shelline and Nachtfischer like this.
  5. Jon Perry

    Jon Perry Well-Known Member

    I support the proposed distinction here, but I'm not sure I like the term analysis for this. Analysis feels extremely broad to me as a term, almost as if it would include calculation. Like if I'm making a detailed analysis of how to improve my game, part of that analysis would seem to involve seeing where and how I could be calculating things better and more efficiently.

    You say that analysis is "the process of forming heuristics", so why not just adopt the term "heuristics" and abandon the term analysis? Heuristics vs. calculation seems like a better framing to me. It's immediately clear from the two words how the two decision-making processes differ.

    More colloquially we could explain the difference as making choices based upon "rules of thumb" vs. "mathing it out."

    As a side point, it's also worth noting that while heuristics can be learned from repeated plays, they can also be gained from discussion with other players, reading strategy guides, etc. So they're not only experience-based. Or at least they don't have to be based on your personal experience, but could be based on the experience of a whole community.
    Last edited: Nov 7, 2016
    Leartes and MichaelSinsbeck like this.
  6. evizaer

    evizaer Well-Known Member

    Reading this thread (and the others on analysis* v. calculation) makes me think that strategy games can only aim to do one of two things:
    • "pleasantly" overwhelm the player with complexity per unit of time, or
    • create tricky contests of calculation.
    This seems quite unpleasant.
  7. richy

    richy Well-Known Member

    Calculation vs analysis to me is trying to draw a distinction between non-interesting and interesting, where interesting is the holy grail of strategy games. I think there's an element of validity to that.
    The consensus here seems to be that the process of *finding* the heuristics, the learning process, is the pleasant and interesting part. If you're saying though that in the end heuristics are just calculation shortcuts and ultimately strategy games are still just calculation contests, albeit obfuscated calculation, then I would kind of agree.

    But I would also speculate that even though gameplay favouring analysis over calculation isn't going to make a game interesting on its own ("interesting" IMO is a blend of many, mostly subjective, factors), it's probably something most kinds of gamer would agree was a step in a +(interestingness) direction. Hence a good guideline.
  8. evizaer

    evizaer Well-Known Member

    I think there are several elements to enjoyment of strategy games:
    1. The joy of mastering a process and exhibiting your agency. This is just a baseline positive that comes with any game where skill apparently matters much.
    2. Exploring the strategy space is fun for some people. This includes the fun derived from being forced to think differently to solve problems.
    3. Seeing what happens is ultimately fun in more cmoplex, broader-scoped games like the Paradox grand strategy games. These games give you an interesting thing to spectate while also allowing you to become involved.
    In strategy games I'm mostly interested in 2 and 3.

    Exploring the strategy-space is more interesting at the level of heuristics than it is at the concrete level of calculation--probably because calculation is a lot harder for humans to do and feels too much like rote memorization, whereas training intuition happens without much conscious effort if you enjoy playing and/or spectating the game much in the first place. You get intuition "for free" if you merely enjoy the game enough to immerse yourself in playing it, so it compounds on existing enjoyment. Calculation is much more effortful regardless of how much time you spend with the game.

    Using a timer to force the player into using intuition when they could be calculating is an unsatisfying design outcome for me. Seems like a superficial fix for a problem that could be better ameliorated in some other way.

    Ultimately, though, almost all games (I can only think of one possible exception, and it's for different reasons that are irrelevant to how we discuss strategy games) have to come down to quick pruning with intuition followed by a bout of calculation. You just can't avoid it. As the community gets better at the game, the pruning techniques because a part of the culture and the game becomes more about calculation.
    Jon Perry likes this.
  9. That is a nice suggestion. I like the term Heuristics. It does speak more for itself than analysis does.

    As you mention this, I just realize there is one more way of learning heuristics: a well-chosen theme. If, for example, a resource is called "gold" than I will assume that having more of it is always better and automatically start collecting it (which means that even on the first play-through, I am better than playing randomly). In this case, I use knowledge from outside the game to be better at the game.

    I am not sure, though, if becoming better at a game that way is really desired by designer. In my understanding, games should try to be interesting systems on their own, without simulating real-world phenomena (or at least I believe that games do have a value, even if they do not teach us anything about specific real-world phenomena).
    Jon Perry likes this.
  10. I think the solution here is to design a game, in which intuition is a stronger and more successful approach than calculation. That way, players would automatically use intuition, simply because it is the better strategy.

    I am not sure, though, if such a system exists.
  11. Jon Perry

    Jon Perry Well-Known Member

    Other than a timer, the main way strategy games try to curtail calculation is by making sure the lookahead tree has a steep complexity curve that takes off exponentially after a few moves and prevents further lookahead past that point due to mental fatigue. The steeper the curve the more likely it blocks even the calculation savants among us. Randomness is one of the main tools as designers we have to manipulate the slope of that curve, since randomness can be used to massively multiply the number of branches in the lookahead tree.

    Edit: You could call the steep complexity takeoff im talking about an "information horizon."
    Last edited: Nov 8, 2016
  12. evizaer

    evizaer Well-Known Member

    It's not really an information horizon, though. for the game itself. It's the information horizon for whatever individual player is overwhelmed by complexity at that point. They still will try to calculate the best move in a subtree that they think is valuable. You would intuit your way down to a subtree and then calculate the values of the individual moves in there--that calculation can be just as challenging and tedious as looking ahead many moves in Chess, even if your branching factor is much higher.

    So maybe the guideline here is to make the branching factor of your game high enough that intuition becomes a bit more important than calculation in narrowing the decision down to the seemingly-optimal move?

    This would say, for instance, that Arimaa is strictly a better game than Chess (when played with the same timer), I think, for the sole reason of the piece movement rules.
  13. Jon Perry

    Jon Perry Well-Known Member

    This is true. I think an information horizon is almost always a subjective thing, dependent on the limits of the mind of that particular player. The only way I know how to create a truly objective information horizon that applies exactly equally to all players is to introduce opacity into the games rules, which seems undesirable.

    EDIT: Although most human minds are maybe similar enough, that at a certain complexity most everyone is stopped from calculating.

    I would agree with all of that. High branching factors probably tend to force the use of intuition more.

    I don't think it's only the branching factor that matters though. There are probably other variables that are important. But the point is that certain tree shapes are probably more likely to lead to intuition being more useful.

    EDIT: Actually the other variable I think matters is the one I was talking about above, how "steep" the takeoff is. I think you want just a few branches at first, enough that anyone can search them, and then all of a sudden, a few moves out, just an explosion of branches, to the point that almost no one can calculate it.
    Last edited: Nov 8, 2016
  14. Jon Perry

    Jon Perry Well-Known Member

    Here is roughly how I imagine the tree of a game that relies more on heuristics than calculation. It's a somewhat extreme example to make the point.
    Basically, everyone can calculate two moves out, and see 4 possible outcomes clearly. This level of calculation should be in reach of any halfway decent player.
    But then each of those outcomes suddenly explodes into so many possibilities that from then on calculating them all properly would be nearly impossible. So the player must therefore rely on learned heuristics about which of the 4 outcomes seems most favorable.
  15. richy

    richy Well-Known Member

    ^ So in that diagram the "information horizon" would be 2 moves away right? And new information is revealed each turn which collapses the highly-branched trees down to a small number of concrete possibilities.

    Like if the game involves driving a car on a straight road towards the horizon. A wide range of obstacles and other stuff might appear on the horizon, but each turn we find out what did appear, and that turns the many potential branches into a couple of actual branches. Then in the 2 turns it takes us to catch up we can plan the best way of reacting.
  16. Akkete

    Akkete Member

    I think that is a very good explanation of the information horizon. Also, I think randomness is very useful in achieving this. Not only because a random event produces many branches in the game tree (that could be achieved with just player decisions), but because random events can’t be pruned. If the branches in the game tree are player decisions, the player can only do look-ahead on the ’promising moves’ and ignore the ’obviously bad moves’. Pruning does require heuristics, and being good at pruning is an interesting skill, but it does make the game feel pretty tedious. Luckily, pruning doesn’t work very well on random events, because they always have a fixed probability of happening, whereas players will try to make winning moves, which makes it easier to predict their moves the better they get at the game.
  17. keithburgun

    keithburgun Administrator, Lead Designer Staff Member

    I also want to just say I love that picture, @Jon Perry - good job.

Share This Page