Archive

Strategy

What is randomness? We tend to imbue the term with more authority than it ought to have, to invoke an ideal of completely unknowable numbers and events that come from nowhere, from nothing, beyond nature, supernatural. However, all of the things that we think of when we talk about randomness are actually quite predictable: Dice, cards, roulette wheels, these tools of chance generate seemingly arbitrary results through a series of individually very simple and straightforward physical interactions. These are still causal. These are still predictable, if one had the depth and clarity of vision and the time for calculation to glean those predictions. Computer random number generation is no different – every computer science text about random number generation goes to pains to establish that the algorithm is not truly random because the outcome can be predicted, but the same is true of basically everything we consider random. Thus, in practice, random doesn’t mean not deterministic, it means undeterminable.

Every reaction has an action. Every effect has a cause.

All this may seem pedantic, but where it becomes important is it shifts the way we tend to think about nominally non-random events. Even games that go to great lengths to strip out elements of chance, such as competitive FPS games, frequently have an element of luck to them. Sure, you could perfectly predict the behavior of all of the opposing members of the enemy team, but unless you have intimate knowledge of their psychology you’re not going to be doing that. You’re going to be playing the odds, judging where your opponents are most likely to make their move based on the state of the game and their habits as best as you know them. A lot of the time this will work out just how you expect it to, but sometimes the outcome will be completely different, and even though there’s nothing we would traditionally consider randomness at play the outcome of what is an overall strong play may turn out to be dramatically different, may lead to a win or a loss, based entirely on circumstances beyond the human capacity of insight.

Randomness is about lack of insight as much as it is about chaos of result. If you fail to understand what factors lead to an outcome, it will seem random. A feeling of a game being unfair is just as often because of a poor understanding of what situation leads to what outcome as it is about the outcome itself being biased. Almost every game hides information from its player – even in Chess or Go, these pure games of open-information causality, you do not know the internal state of your opponent, possibly the most vital information of all for victory.

It all comes back to luck, the thing no one wants to rely upon. The more we come to understand what random is, what we mean when we say random, the more we come to understand where the boundaries of what we know and can predict and what we don’t know and can’t predict lie; the more we can control luck. When we understand what these factors are and how they interact, we can begin to make our own luck.

We’ve invented probability to estimate the likely outcomes of overly complex systems using simplified models, and this is hugely useful and hugely insufficient – not least because very few people have any sort of intuitive understanding of what these numbers mean. During the 2016 election, people were passing around the supposed 10% chance of victory Donald Trump had as an indication that he had somehow lost. 10% isn’t 0%, and a 10% chance is actually pretty likely – hardly much better odds than those offered by Russian Roulette, a pastime few people would willingly indulge in. And, when the estimate rose to a 30% chance, people were still weirdly reassured – and felt betrayed by these numbers, regarded them as failed somehow, after he won, even though the numbers said there was a 1/3rd chance of this happening. Perhaps the mistake was that people conflated the percent chance of each side winning with the percent of the populace projected to vote for each side – in which case 70%-30% would have been an unprecedented and implausible landslide. Either way, there seemed to be a generalized lack of literacy as to what these numbers meant.

All of which brings me to XCOM. In XCOM, you’re tasked with defending (or, in XCOM 2, retaking) Earth, after it is attacked by a multi-species conglomeration of aliens with mysterious motives. You give your elite squad of soldiers orders of where to move and what to shoot, and each shot has a percentage chance of success, and it is a hugely effective tool for imparting the core concept that a 95% chance is not a certainty. Success, then, becomes a matter of ordering and improving these chances, hedging bets, and trying to keep outside factors from interfering while you do so.

Unfortunately, the game undermines this simple and vital lesson in a few ways. There are a number of options which have zero risk of failure – indeed, most of your options tend to trend that direction as you get late-game upgrades to shot accuracy and abilities with no chance for failure, but even early on there are very few problems that can’t be solved by the simple expedient of a deftly placed hand grenade. Or three. So, in the end, rather than hedging all your bets, most of them are just backed by a couple of completely reliable fallback moves, and when well played the enemy units seldom get any real chance to counterattack.

More insidiously, though, the lower difficulty levels also sneakily tweak the odds in your favor. The more you miss your shots the more likely you are to hit subsequent shots – the gambler’s fallacy, codified into game mechanic. The odds of hitting are, as well, far higher than they are represented as, and your odds of getting shot are similarly reduced. All of these strange concessions and tweaks work to make the game feel more ‘fair’ – or, read less charitably, to uphold fallacious views of what probability means that have been elevated to narrative necessity.

Who do these views serve? Does the belief that the world is fundamentally knowable and controllable outside a few supernaturally random events make us vulnerable to believing that we are more in control than we actually are? Vulnerable to believing that those who have lost control, who are downtrodden and oppressed, deserve this treatment for having lost grip of the reins? Or does it make us think that the world we live in now is far more intentionally, carefully, and competently controlled than it sometimes seems, that any attempt to seek change within such a system is impossible?

We only see a tiny, narrow slice of the world in front of us, and anything outside of that view may as well be RNG. It might be time to test our luck.

If you enjoyed this essay, please consider supporting me on Patreon. Support at any level lets you read new posts one week early and adds your name to the list of supporters on the sidebar.

One of the most common ways to evaluate a game design decision is in terms of “risk and reward”. Usually we assume that that whenever the player takes a risk it should be to attain a commensurate reward, and so we try to encourage the player towards risky play by offering such rewards. Risk-taking is something worth encouraging, so the logic goes, because it increases the tension and therefore the excitement of the gameplay.

This assumption raises some questions. Does risk actually make the game more exciting? Since there’s always a threat of failure in any challenge-based game, if the ‘risks’ provide rewards that increase the long-term chances of success, aren’t actually risky overall – they’re just the most dangerous inflection point of a strategy. If they’re risky because they have a random chance to fail, they likely fall into one of two categories: Either an unnecessary chance of creating a failure where none exists, or a necessary gamble to take in order to gather the resources needed for success. Either way, the risk is usually either always worth taking or never worth taking, and the game becomes just a test of luck and of the insight to know whether the coveted resources are necessary to victory. Conversely, if the risk is a test of skill, then it becomes something similar to the luck test but with unknown odds of success – but, again, the player either needs the resources or they don’t and they either have the skill to mitigate the risk or they don’t, and in either case the strategy is simple and straightforward.

The trade-off of risk and reward is, by itself, an incredibly tedious way of balancing a game. Once you know how a statistical game is optimally played, it stops being very interesting: Blackjack is not interesting because it’s a good game, it’s interesting because there’s money on the line, so unless you want to ratchet up the stakes of your game to include real life consequences (beyond wasted time), the risk/reward model exemplified by the casino is not one to emulate.

There are a lot of tools that are useful to describe some aspect of game design, but they are hazardous to use prescriptively as a blueprint for what a design should look like. Genres, as well, are great for describing fiction, but sticking too closely to their conventions is anathema to the imagination. The issue with “risk and reward” is that the risks and the rewards aren’t actually what’s interesting about the challenge of a game. There are two things that are interesting about game challenges: planning and mastery. The most satisfying experience in a game is coming up with a plan and then executing it – or failing to execute it and having to improvise a new plan and execute that. While viewing a player choice as a risk and a reward can give an insight into how these strategies will take shape, it almost never shows the whole picture.

You might be wondering what specific tree the branch up my ass came off of at this point – that is to say, you may be wondering what actual game design decisions I have in mind when I say that this faulty metric has led designers astray. The first example I have is probably a contentious one, because I know lots of people really like it, but I think that parrying in the Dark Souls games is garbage. You have a game that rewards careful analysis, positional play, and timing, and then you also include a mini-game that lets the player ignore all of those things if they can hit the button at the right time. “Do or do not, there is no try” may be helpful advice for space wizards, but it is a pretty lousy way to design a game. By the metric of “risk and reward” parrying looks like great game design – you take a risk of eating an attack to the face for the reward of distributing an attack to someone else’s face! – but in terms of giving the player something interesting to do it fails. It’s Guitar Hero with the sound turned off. It’s a Quicktime Event with no button prompts.

Shields in Dead Cells share most of the problems with parrying in Dark Souls (which makes sense since that’s what they were clearly inspired by), but a much bigger issue are the cursed chests. In Dead Cells, a roguelite game where each run is unique, you frequently find cursed chests. These chests contain a fairly useful reward – a bit of money and item-unlocking currency, a high-level weapon, and the equivalent of a level up – but in return they curse you, which means that if you take any damage before the curse is lifted you instantly die (the curse is lifted after you kill 10 enemies). These become an incredibly awkward piece of design, though, since both the risks and the consequences of those risks increase rapidly to the point where there’s essentially no way for the rewards to keep pace. Early on, if you find a cursed chest there’s very little reason not to take it: If you die you don’t lose very much, and it might give you just the item you need to pull your run into shape. Past that point, though, you start to risk completely losing 30 minutes or more of gameplay, and having to completely redo the relatively rote early levels, in order to get an item which you’ve already probably got something more useful than and gain some currency you don’t need. So, in this case, not only is the trade-off not very interesting, but the choice is usually obvious based on your situation.

So how do we try to make the choices in the game interesting, if not by measuring their risks and rewards? The key to whether a choice is compelling usually lies not in what we risk or we sacrifice, but in what we need to take into account to make that decision. If any given choice could be good or bad based on the situation, that generates an interesting thread of thought to follow – assuming those externalities themselves are interesting to navigate. If a choice will always be great in a particular scenario and you know that the scenario will be in play when you take the choice – IE if fire weapons are extremely useful against the ice monsters and the next level is populated entirely with ice monsters – then it’s not really an interesting choice whether or not to take it, since you know it’s optimal. These sorts of obvious best choice situations can be good for pushing the player to try a new mechanic, but aren’t interesting in and of themselves. Conversely, if you know the next level could have ice monsters or robot monsters or a dark labyrinth, and while the fire sword isn’t great against the robots it’s fantastic against the ice monsters and also can help light the way through the labyrinth, but the laser is more generally viable against the robots and ice monsters but has limited ammunition, but you’re really most comfortable using the poison scythe and generally prefer it – this starts to become a really interesting choice, one generated from the specific combination of the situation and how you in particular feel comfortable playing the game.

A great example of this kind of decision-making is the choice of whether or not to take a given card in Slay the Spire. When presented with a set of potential cards to take, you weigh them in terms of their general usefulness, their usefulness in the deck you have now, their usefulness in combination with other cards you might get in the future, the likeliness of getting those cards, what boss you’re expecting to fight, and more. Every decision has a risk and a reward, sure, but the designer didn’t determine what the risks or the rewards were in an excel document, these risks emerged from the nature of building a deck, and the reward is of seeing a machine you have built work flawlessly.

There’s a lot you can learn from thinking about a given player choice as a risk and a reward, but there’s even more that can be obscured if you trap yourself into seeing it only through that lens. Every player decision has to have context, has to have its place in an overall strategy that emerges from the player’s engagement with the game’s situations and tools. If it does not, it’s a coin flip or a Quicktime Event.

I’ve continued to play a ton of Slay the Spire over the last few weeks, though I’ve dropped off a bit. Even when I’m not playing the game I like to think about the game. I’ve been thinking about one card in particular quite a bit recently: Prepared.

In order to talk about this card and why it’s interesting I’ll have to talk a bit about how the game works. Put briefly, each turn you draw five cards, and at the end of the turn you discard your entire hand. During each turn you have a certain amount of energy with which to play cards: You start with three energy and most cards cost one to play. In between battles, you get a choice to take one or none of three cards, and judicious selection of these is key to achieving success as the battles get more and more difficult.

Now that you have some context, Prepared is a card that costs 0 energy to play, draws an extra card, and discards a card. It is commonly regarded to be a bad card. I don’t think it is.

There’s a couple of naive ways of looking at Prepared. The first is just looking at the 0 and the draw a card and saying “ooh, free card draw!” Of course, it actually isn’t free draw, because the card you end up drawing is just the next card in your deck, which you would have drawn anyway if you’d never bothered to take Prepared. The second naive perspective is the reaction to the first, saying essentially “well this does nothing but discard a card, so it’s bad”. The first part of this is true: The second part isn’t.

There’s two assumptions that go into this read. The first assumption is that discarding a card is bad, which it frequently is not. Once you’ve spent all your energy for the turn, extra cards are useless, so until you start generating enough energy that you have some left over after using all the cards in your hand Prepared costs you nothing at all. The second assumption is that discarding a card is not good, which sounds a lot like the first one but I think is subtly distinct: One is the understanding that much of the time there is no penalty for discarding, the other is the understanding that there is frequently utility in discarding. On the more obvious end of this, there are a number of cards and relics which take advantage of discarding in various ways – in describing Prepared as a bad card people often stipulate “except in a discard deck” because of these. However, there are also curses and negative status effect cards that you want to discard before the end of your turn to keep from taking damage or other negative effects, giving it a utility outside of these specialized synergies. If you have 3 energy and most of your cards cost 1, as at the beginning of the game, prepared is almost never bad and is occasionally quite good.

Then there’s even more obscure and surprising uses. I’ve used prepared as nothing but an empty card that costs 0 alongside abilities that activate with every card played. I’ve used it to dump cards which were perfectly fine but also not good enough to justify using them when I could empty my hand and use a relic to draw more cards. I’ve used it to pull an extra card so that next turn I would have 0 cards left in my deck and be able to use an ability that’s only usable under those circumstances.

This is what’s wonderful about Slay the Spire. Everything has a use and everything works together, sometimes in delightful and unexpected ways.

I do think it’s interesting, though, seeing how these concepts of card advantage and deck thinning sometimes fail to transfer from games like Magic and Hearthstone, where most people learn them, to a game that is, on the surface, very similar. Thinning your deck is if anything even more important, since you end up cycling through it multiple times in a combat – however, the utility of cards that only serve to grind through the deck faster is questionable, since by adding them they themselves thicken the deck, which less of a concern in a game with a set deck size. Drawing more cards in Slay the Spire doesn’t control your long-term prospects the way it does in these games either, since you’re going to draw a new hand next turn anyway, but solely offers you a way to burn energy to achieve advantages on the current turn.

In all honesty, Prepared still isn’t a very good card. It’s bad in a lot of decks, and in a lot of others it makes no real difference one way or the other. Still, the specifics of when and how it’s good, and the assumptions that people bring into the game about why it is useful, or fails to be so, are interesting.