Thursday, 13 September 2012

Game Design Fundamentals - Meaningful Decisions

As someone who's been making games for over ten years, initially as a hobby and now as a full-time career, I've picked up a lot of wisdom and experience that I probably take for granted. While much of it has been collected from my own experiences developing games, most if it has come from other developers who I've met at jams and events, or read about online.

I want to share some of these in my blog so others can learn from what I've learned. Hopefully there'll be plenty of ideas here that will be useful and will be helpful for you in your own game development!

I'll start out by looking at meaningful decisions. When I design and work on my games I try to make every decision as meaningful as possible. Let's have a look at what that means.

Introducing Meaningful Decisions

Games are built from decisions, be they moment-to-moment, like deciding where and when to jump in Super Mario; or discrete, such as choosing your next tech tree option in an RPG. Decisions open up possibilities for new decisions - creating access to a new power-up, or the next set of skills in a tech tree, for example - and finding the optimal path through the decisions available to you is a big part of what makes games mentally stimulating.

I like to think of a decision as meaningful when it opens up possibilities with one hand and closes some off with another. The opposite of a meaningful decision would be a redundant one - one where there is obviously only one optimal choice.

Examples of Meaningful Decisions

Levelling up a weapon in Alien Soldier is a meaningful decision because it means choosing not to level up a weapon you may need later. So while beating bosses vulnerable to weapon A may be easier after you've levelled it up, you may find yourself eaten alive by bosses vulnerable to weapon B because of the vary same decision.

In Triple Town, every position you can place a grass tile in will create an opportunity to make a bush, but also cordon off positions where it is no longer possible to grow new tiles. When you learn a new move in Pokémon you are forced to forget another - which set of strategies should you sacrifice in order to access this new one?

David Sirlin's Kongai is a fantastic example of a game where every single decision is meaningful and has a noticable outcome on the rest of play. Kongai is one of the most engrossing games I have ever played because every single action matters, and you never find yourself going through the motions.

Meaningful Decisions in Greedy Bankers

I designed Greedy Bankers so that, at any stage during the game, there would be a choice of whether to cash in your best gem now to free up space for more gems, or to keep it and build it up to make it more valuable. The rubble and the robbers change how much you can expect to grow a gem, so your chances of growing a gem successfully change from second to second.

As a more subtle example, look at the positioning of the magic crystals in the iPad version. Magic crystals appear hidden inside special bits of rubble and, when activated, turn surrounding rubble into like-coloured gems. In Bailout Mode positioned the two crystals so that using one of them in its optimal position (i.e. the position that would generate gems of highest value) would also result in the other crystal being turned into a gem. So there's one simple way to create a lot of new gems, but by destroying your second crystal you could be closing off an opportunity to make even more.

Even cashing in gems in multiplayer is an important decision to make. Cashing in a large gem generates rubble on your opponent's side, which they could easily turn into new gems of their own with a magic crystal. So you need to know that cashing in a gem will bring you close enough to your target, or that your opponent will be in a poor position to bounce back (for example, if they don't have any magic crystals on their board).

In each of these examples the better choice is not obvious, and their best strategy is entirely dependent on state of play at that moment (their current score, their opponent's board, the time on the clock, whether there's any robbers nearby).

Why Meaningful Decisions?

Good meaningful decisions have long-term implications on play, and will affect the strategies you will be able to use later on. It makes them high-stakes and high-adrenaline, and increases your personal investment in your performance. Meaningful decisions make you feel smart when you get them right, and leaving you screaming "I should have known better!" when you get them wrong.

In the long term you may find yourself playing differently depending on the choices you made early on - a decision that affects the way you play makes the experience deep and rich, and full of possibilities to explore.

Redundant Decisions

Redundant decisions can add a lot of clutter, and often frustration to a game design, or at the very least are wasted opportunities. Often redundancy forces you to perform the same trivial set of actions over and over. In Skies of Arcadia there were plenty of stat-changing attacks, but none of them affected bosses (the only threatening battles in the game) so they were utterly pointless. This meant that beating bosses was always the exact same routine of charging up big attacks, performing big attacks, then performing healing spells, over and over. I had a similar issue with Final Fantasy VII, which had a list of hundreds of near-identical spells, but I barely found a use for most of them.

Not that redundant decisions are necessarily an issue. Guitar Hero is a great game even though most of the decisions (press the right keys at the right time) in it are redundant - you either time your presses right and win points, or mis-time them and fail to win points. The game works because it is a test of accuracy and agility before it's a game of strategy.

However, I do remember playing Guitar Hero 3's multiplayer battles on PS2, which did suffer from redundant decisons. In this game mode power-ups were awarded for a given number of points, and allowed you to mess up your opponent. You obviously want to use your power-up as soon as you get it, as it will prevent your opponent from gaining points and winning a power-up of their own. There is no decision to be made, as the optimal strategy is clearly to win points faster by playing more accurately, and win the power first - the exact same strategy that you take in standard play.

If power-ups had a negative effect on the player who used them then choosing to use a power-up would risk damaging your own performance. If they were more powerful during certain passages then there'd be a decision to use them as soon as you get them, when they'd be weaker, or save them up for a section where they'd be more powerful. In order for these elements to add to the strategic palette of the game, there'd need to be valid a reason to use them later, in a different way, or even not use them at all.

Rules of Thumb

In general, I like to think of it this way: any buff in my game should be counteracted with a risk or debuff. Running to get a powerup in a platformer should require you to dodge enemies and traps, risking losing a life, or require a diversion that costs valuable time. Moving to a good striking position in a sports game should leave your goal open, giving your opponent a better chance to score. Choosing to activate a power sooner could prevent you from using it later, when it might be more valuable. Choosing to upgrade your RPG character to a new class could prevent you from learning skills from other classes in the future, or from using skills that you may have valued before. Every time you make a choice there should be a risk that you've made the wrong choice, or it should force you to reconsider the way you're playing.

If the player has no reason not to do use a feature then the game is no more deep or rich than it was before you added it. All you've done is make it one notch more complicated. Another way to think of it is that one new feature needs to create two new strategies. I call this the "two-for-one" rule. This way, the web of strategies can increase exponentially, new features having knock-on effects on other strategies, making for a deep and multi-layered game with relatively little complication.

Meaningful decisions make for high-adrenaline moments and deep gameplay. If you can fit in as many new strategies for as few new game elements as possible then you're onto something good.

I hope this is useful food for thought for anyone interested in game design. If you have any questions please do let me know, either in the comments or on twitter!


  1. I've often heard you cite Kongai as an example of good design. This is very curious to me, because I consider it one of the most conspicuous design failures I've ever encountered.

    From my perspective, the problem is that a player armed with a basic understanding of mixed strategies (the game theoretic terms for strategies with a random element to them) sees only one available move most of the time. So far from every move being a meaningful choice, every move is a meaningless choice. The game is, essentially, rock-paper-scissors (a farce of a game that shouldn't be considered an acceptable template for anything). Surely, as designers, we don't believe that trying to mindread the opponent counts as gameplay?

    Example aside, the actual point you're making here is a very good one. I look forward to seeing what other topics you visit with this series!

    1. Glad you enjoyed it! I'll have to disagree with you about Kongai though. Not quite sure what you mean by a player "armed with a basic understanding of mixed strategies" though - could you elaborate? Most of the time there is more than one move available to you at any one time, as even if one attack does most damage it may cost more energy, or may not cause a status-change that will be useful in the long term. From my memory, the vast majority of situations allow you at least two viable options (beyond switching and intercepting), because so many of the moves work well as set-ups for strategies further down the line. Although perhaps if you explain what you mean about the mixed strategies thing I might get the point you're making a bit better :)

      I think the rock-paper-scissors element of the game is very interesting. It's weighted RPS, so it's not a random chance, and so that element is all about working out what your opponent's optimal strategies are, and then based on your understanding of their behaviour, predicting whether they're likely to do it. Or predicting whether they're going to go for immediate damage or set up a long-term strategy. Understanding an opponent's behaviour is a very useful life skill and crops up in all kinds of great games, from poker, to chess, and even politics, economics and real-world warfare. So yes, I do think it counts as gameplay ;)

    2. Yeah that stuck out like a sore thumb to me too.

      Maybe we just played Kongai and David Sirlin's Kongai is much much better.

    3. OK, so to elaborate a bit:

      RPS is a solved game. The optimum strategy is to pick each option with probability 1/3. Weighted RPS is similarly solved in the general case, with again a single mixed strategy being played over everything.

      But never mind the theory - let's focus on Kongai. So, in a typical situation in the game my character is beating the opponent's current character and there are about four things I might want to do:

      * Intercept, because they probably want to swap out at this point.
      * Invest power in my biggest hit, since it will finish them if they don't swap out.
      * Opt for a small hit since it's more energy efficient (in terms of damage per energy).
      * Take this opportunity to switch character ourselves since the opponent will almost certainly not intercept.

      Loads of tasty choice, right? Well, not really. The top three options just form a weighted RPS subgame. So we just solve that (more likely by experience and/or intuition than exactly, but it makes little difference gameplay-wise) and play near-optimally. The bottom option will sometimes generate deeper gameplay depending on the characters we and the opponent have available, but the payoffs are so tiny compared to the actual combat that clever play is very unlikely to influence the outcome of the match.

      I think I see from your comment where our viewpoints differ, though. The idea that you can "understand someone's behaviour" in a way that lets you predict their move in a weighted RPS game to any degree at all is implausible to me. I suppose if the opponent is a six year old or has utterly failed in their analysis of the game then it might happen. Chess has no psychology at all and poker has very little when played well (it's mostly a myth designed to appeal to fish). Politics? Yes, absolutely, but that's because your opponent is actually one of the pieces on the board, so to speak.

      I think the interesting boundary case for psychology in games is Mafia (often called "Werewolf"). That seems to be a game which genuinely manages to bring psychological considerations into its gameplay. Even then, they diminish rapidly if all the players are good... although not quite to zero, I think.

    4. I disagree that RPS is a solved game because the human brain does not have a random number generator. Humans instinctively form patterns, and think in terms of "I haven't played scissors in a while so it would be more random to play scissors" - that means that even "random" actions can be anticipated if you get the way your opponent is playing. In Kongai, the weighting changes every round, so you can't just sit with a calculator and expect to win 50% of the time. There's a reason that winners of the game stay consistently at the top rankings.

      Chess has plenty of psychology at high level. The human brain can only see a finite number of moves ahead, so anticipating your opponent's long-term plans is an important part of play.

      The design of the attacks is never just as simple as a big attack for high energy or a small attack for low energy. There's a lot of long-term strategy in the game, and effects such as punishing switch-outs with damage, attacking opponents on the bench, forcing an opponent to switch out so that their next character cannot avoid your next attack... there's not many attacks per player but, because of the range of different effects they can have (on you and your opponent) there's a wealth of long-term strategies there for you to uncover. The best moments (and these happen often) are when you've manoeuvred your opponent into a situation where they have no favourable choice.

      May I ask, how many times you played the game before you dismissed it as "just RPS"? I assure you if you'd stuck with it longer you'd have started to discover some long-term strategies and realised it's not quite as simple and "solved" as you might think. Perhaps the failing of the game is that it's easy to dismiss as random chance before you get into it.

    5. I don't remember exactly, but I played it twice. First time was shortly after it appeared when I played for only an hour or two before concluding that there was nothing interesting going on with the few characters I had access to. Much later when I had a far better mix of characters I tried again and played in occasional bits of spare time for about a week before getting bored of it. So no, not much overall.

      Incidentally, your description of the psychology in chess is actually a description of something that isn't psychology at all. Yes, anticipating long term strategy is important (and interesting), but it's a rational process of analysis.

    6. Personally, I don't think the process of understanding player psychology and a rational process of analysis are that different. In fact, I'd say Kongai is very much about rational analysis. You make a rational choice based on your (a priori) best strategy, your opponent's best strategy, how your opponent has been playing so far, and the level of risk involved in your selection being the wrong one.

    7. So suppose I decide to generate my random numbers by (say) rolling the dice sitting on my desk here? I don't tell you, I just quietly do so. Your psychological insight into my choices is then... what, exactly?

      The short version is: you are the Sicilian from 'The Princess Bride'! ;-)

    8. If you do that you'll generate a sub-optimal strategy, surely? Either you'd not be taking the weighting into account, or you'd be missing out on long-term set-ups. Especially if I play to minimise my risk. Perhaps that's something we need to physically test.

      Unfortunately I've not seen Princess Bride so I don't get the reference ^^;

    9. OK, you're not allowed to write any more blog posts or make any games until you've watched it! In fact email me your postal address and I'll have a copy sent to you.

  2. bateleur: donkeyspace!

    Sure RPS is solved in the sense that there's a strategy against with no other strategy can do better than even against, but a) as Alistair says it depends on having a random number generator and b) if your opponent isn't playing that strategy then it's possible to do a lot better.

    (this tends to come out a lot better in real-time games though, because under time-pressure we do worse both at solving the equations and at generating good random numbers.)

    1. Thanks for posting that link - interesting stuff. Having a look at his GDC talk now ;)

    2. brog: That article makes exactly the same mistake with Poker that Alistair makes with chess. It confuses the process of reasoning about the opponent's bids with psychology. There is indeed a huge amount of the former in poker, but it's all probabilities and EVs.

      The question "What if the opponent doesn't play well?" is never interesting to me in any game. And no, that is not how poker works. Read a bit about Chris Ferguson if you don't believe me.

      Even in realtime play RPS is never a very good model. Look at Street Fighter IV (and related games) where the top players view "mix ups" (the term for an RPS-like situation) simply in terms of average outcome rather than as a source of gameplay. Indeed, a mixup situation is often the payoff for an a successful line of attack rather than being where the conflict occurs!

    3. Again, I don't really see these disciplines as that different, and you need a solid level of reasoning (damage output vs energy drain, and setting up future strategies, alongside level of risk) to perform well in Kongai. Your opponent's state, and their previous behaviours play a part in your own reasoning.

      At the end of the day, I thoroughly enjoyed playing Kongai and was massively hooked for a long time, so it must have been doing something right - if it was purely a game of chance I probably wouldn't have invested myself in it as much as I did ;)

  3. As my point of view, the problem is that a player armed with a basic understanding of mixed strategies (the game theoretic terms for strategies with a random element to them) sees only one available move most of the time. So far from every move being a meaningful choice, every move is a meaningless choice.Thanks.

  4. Approaching your exploration along these lines, you will let the examination procedure control the bearing of your future paper as opposed to attempting to discover pertinent things that would match what you require. This methodology is much quicker and creates obviously better comes term paper writing services about for both learners and those needing time for writing assignments. As you ace this calculation, you can apply it to more entangled and genuine tasks.

  5. You've the lavish with choosing a transient home loan credit according to your individual situations. As a representation, the people utilizing adverse credit positioning need to oblige a home loan advance to get couple of months utilizing a restrictive focused on including bettering financial assessments rapidly.

  6. You may have gotten twelve calls to solidify your understudy advances, however don't let an inconsiderate or amateurish individual on the telephone abandon you with a terrible feeling about understudy credit combination. Merging your understudy advances can spare you thousands and talking with the opportune individual and comprehending what to search for in an organization could make it the best call you've gotten throughout the day! payday loans

  7. In the current monetary atmosphere where credits are not all that promptly accessible as they used to be it is helpful to recognize what your alternatives are before petitioning a credit. A secured advance is a credit that is secured on your property, and is accessible to individuals that have a home loan on their property who likewise have enough value left in their property. Cash Advance

  8. Title advances have the same highlights as a secured change, close by a singular perspective. While secured credits don't chart the kind of certification that will suffice it, title moves especially oblige vehicles or whatever other vehicle to go about as security. Auto Title Loans Chicago