Games and Economies
On Games, Part 5: Hell is Other People
In this article, I talk about playing games against human beings, and how human psychology complicates our notion of the “best” move.
“All those eyes intent on me. Devouring me. What? Only two of you? I thought there were more; many more. So this is hell. I’d never have believed it. You remember all we were told about the torture-chambers, the fire and brimstone, the “burning marl.” Old wives’ tales! There’s no need for red-hot pokers. HELL IS OTHER PEOPLE!”
— Garcin, from Jean-Paul Sartre’s play No Exit
2002, Salt Lake City. It was the finals of the 1000m short track speed skating event in the Winter Olympics. Steven Bradbury was representing Australia, and his four competitors — Ohno, Hyun-soon, Li, and Turcotte — were all younger and faster. The field of competitors had been skating all day to qualify. Bradbury reflects on the buildup to the race: “I was the oldest bloke in the field and I knew that, skating four races back to back, I wasn’t going to have any petrol left in the tank. So there was no point getting in there and mixing it up…”
The race kicked off. Bradbury immediately fell to the back. Over a minute into the race, with just two laps to go, Bradbury lagged about 7 meters behind the rest of the pack.
Heading into the last lap, Bradbury was now 15m behind. But then, the unthinkable happened.
Jostling frantically for first place, the top four competitors crashed into each other at high speed and went sprawling across the ice into the barrier. Bradbury cruised over the finish line unobstructed to take first place.
Did Bradbury just get lucky despite being less skilled than his opponents? Or was Bradbury’s deliberate decision to stay back, in anticipation of his opponents crashing, actually a stroke of genius? We’ll revisit this question at the end of the article.
Metagame
So far in this series, we have talked about topics that are endogenous to games: currencies, mechanics, and win conditions. These concepts shape our thinking on how games work in a vacuum, and how to win them. We have noted that a player’s skill is not a currency. Rather, skill is a measure of how well a player can make favorable transactions with their initial “wealth”. We have also hinted at how some games, like social deduction games, blur the distinction between the game and its players. It’s time to tackle this topic head on through a discussion of metagame.
Metagame is “the game about the game” — any approach to the game that draws on exogenous variables. Metagame is most talked about in the context of exploitative play, where you take actions in the game to deliberately confuse or surprise your opponents, or take advantage of a specific weakness or preference they have. But metagame can also refer to personal heuristics in an environment where the best strategy is unknown, not obvious, or not strictly better than alternative strategies. Playing moves that are “optimal” requires only reasoning about the game. Playing moves that “maximize my chance of winning” requires reasoning about the game and the metagame.
Defensive formations in soccer, routes in American football, betting strategies in poker, and fielding positions in cricket could all be considered examples of metagame: approaches to play that are not mandated or implied by the game mechanics in any way, but have emerged organically through years of competition. These patterns go in and out of vogue as the infrastructure of the game changes (e.g. think different types of tennis courts), quality of play improves, and opponents test unorthodox strategies, forcing others to evolve. From an economics standpoint, metagames can be roughly thought of as unstable Nash equilibria, where players coalesce around certain modes of play but can find situations where it is advantageous to deviate.
Metagame frequently crops up in games where players choose different initial portfolios of resources. In Starcraft, players play as one of three races — Protoss, Zerg, and Terran — each with their own strengths and weaknesses. In Overwatch, players choose a hero to play as, each with their own strengths and weaknesses. In Counter-Strike, players choose specific maps to play on, and choose to eliminate other maps from being played on. In Magic: The Gathering, players build their own custom decks. Half of what makes these games enjoyable is the journey to developing an intuition around what combinations of resources yield consistent advantages.
It’s worth quickly distinguishing metagame from strategy, which is the logic of why you think a particular approach will win, and the plan you will execute as a result. For instance, the set of all established defensive formations in soccer, and their respective popularities over time, is the metagame; the reasoning behind a particular formation, the roles players will assume in that formation, and the logic of how they should act in concert to create goal-scoring opportunities is the strategy.
The classic and most flamboyant example of metagame comes from the beloved wine scene in the movie The Princess Bridge, where Vizzini must guess which glass of wine Westley put the poison in. His hilarious monologue touches on Westley’s intelligence, the Australian origins of the poison, and previous fight sequences in the movie — variables that are all exogenous to the game.
Which glass would you have chosen? The optimal move, as we now know, is to be immune to iocane powder.
Metagame matters more in some phases of a game than others. Most games follow an implicit three-act structure: opening, middlegame, and endgame. Each phase of the game emphasizes a different set of skills:
- Opening: “prepare to fight” — acquire resources efficiently and set yourself up to derive the maximum value from those resources
- Middlegame: “fight for advantage” — use your resources to attack and defend with the goal of ending up with some trump cards
- Endgame: “convert the advantage into a win” — use your trump cards to hit your win condition
In the language of the previous article on games as portfolios of assets, the opening is where you acquire your portfolio assets and justify their initial cost; the middlegame is where you increase the market value of your portfolio by investing heavily in unrealized gain; and the endgame is where you cash out, either by turning your unrealized gain into realized gain (converting influence to power), or by forcing your opponents to liquidate at a loss.
In games where the win condition is a race against time, these phases of a game may not look very different in terms of actual gameplay, since the goal at any point in time is simply to stay ahead of your opponent. The distinction is then more about how you pace yourself and vary your play-style at the start, middle, and end of a game. Metagame still plays a role as competitors try to identify and exploit each other’s weaknesses in real time.
In games where the win condition is a race to the bottom, metagame tends to matter more at the beginning of the game, where information may be scant, competitors may be plentiful, and the “optimal” move is all but impossible to judge. In Suzanne Collins’ The Hunger Games, metagame considerations abound. Should you run into the Cornucopia to acquire resources — and which ones? Should you make alliances with contestants from other districts? Should you head to the edge of the arena or stay closer to the center? Should you stay put in one place or move around? How pivotal are the sponsors to your survival? While some readers (entertainingly) debate whether these questions have optimal answers, there are too many variables to assess accurately. Your ability to make it to the middlegame is heavily calibrated to what players have done in the past, what players are doing now, and how you respond.
In games where the win condition is a race to the top, metagame tends to matter more at the end of the game. In the spirit of a race, players will often pace themselves until they are ready to make a dash for first place. When you should start making that dash, and whether you can sustain it, are the stuff of metagame. Sound familiar?
Regardless of win condition, metagame matters most when the game is “unsolvable” in some important way, either theoretically or practically.
Donkeyspace
Let us dwell on the difference between the “optimal” move, and the move that “maximizes my chance of winning”.
The optimal move is the “objectively” best move — the move that is least exploitable by any opponent. In rock-paper-scissors, for example, the “optimal” move is to perfectly randomize between the three options.
The move that maximizes my chance of winning is not the least exploitable move, but rather, the move that best exploits my opponent. For example, if I am confronted with an opponent that only plays scissors, the move that maximizes my chance of winning is to always play rock. By playing this strategy, I, too, must play exploitably, with the understanding that my suboptimal play defeats your suboptimal play.
This idea of deliberately playing a suboptimal move to exploit an opponent’s suboptimal move is captured in Frank Lantz’s concept of “donkeyspace”, which he coined in an article in 2011. Donkeyspace is the set of all suboptimal moves in a game. Frank argues that in real life, player strategies follow chaotic orbits in donkeyspace, where player A (for whatever reason) plays suboptimally, then player B plays suboptimally to try and exploit player A’s move, then player A plays suboptimally to try and exploit player B’s move, and so on. Complicating the idea of donkeyspace is the fact that players are estimating the optimality of each move incorrectly as well, hastening the descent into donkeyspace. The claim here is that sometimes you have to play stupid to beat stupid.
There are two dimensions on which you can try to exploit your opponent — their game, or their metagame. Exploiting their game means you are taking advantage of their skill level: you expect your opponent to fail to recognize the optimal move or execute it consistently, meaning they are more likely to fall for tricks. We tend to see this in perfect information games, where experts are well-versed in what the optimal moves are, or have a much better intuition for why a suboptimal move is suboptimal. In 2014, when chess world champion Magnus Carlsen played Bill Gates in a friendly televised match, Carlsen deliberately played a bad move — 7. …Ne5? — predicting Bill Gates would fall for a simple trap, which he did.
You can also exploit your opponent’s metagame —their personality, preferences, or heuristics for approaching the game, independent of skill. This is captured in the adage “don’t play the game, play the player”. Social deduction games like poker are notorious for this kind of dynamic, where players go in circles trying to read each other and disguise their own emotions so they can trick their opponents into burning money on losing hands. The final poker scene in the movie Rounders is a great example of exploiting metagame in action.
Another great poker scene in Molly’s Game teases out the psychological warfare of donkeyspace very explicitly. The protagonist Molly provides narration on Harlan, a pro poker player who loses a hand to a terrible poker player nicknamed “Bad Brad” because Brad accidentally makes a series of genius bets that convinces Harlan to fold a winning hand. Harlan is outclassed by a string of “optimal” bluffs even though Bad Brad is so bad at poker that he has no idea he bluffed optimally. Harlan blows a fuse and goes “full tilt”, losing all his winnings over the course of the evening.
Broadening the definition of skill
Human players introduce a chasm between the “optimal” move, which should be inferable purely from factors endogenous to a game; and the move that “maximizes my chance of winning”, which factors in the metagame. Playing the game optimally is what we have passively understood to constitute skill. But successfully navigating donkeyspace in an ocean of competing metagames and coming out on top is a skill of its own, one that sits above rigid notions of “correct” play and should be incorporated into our definition of skill so that we paint a more complete picture of what it really takes to win against human beings. Especially in social deduction games, sometimes metagame is all there is.
In a number of interviews after his race, Steven Bradbury said that he felt the four other racers had been under a lot of pressure to come first, each for their own reasons, which might have influenced them to take unnecessary risks. He later went on to write an article titled The Medal I Struggled to Accept, where he explained how, in the moment, he felt apologetic about winning in the way that he did. (For context, Bradbury had won some of his earlier heats in similar fashion, where frontrunners crashed into each other at the last second; he had also benefited from certain competitors being disqualified even though they had better times.) Ultimately, Bradbury did convince himself that he deserved the medal. One quote in his article stands out for me:
If you happen to achieve your goals at the end as well — which I was incredibly lucky to do as probably the luckiest individual gold medallist in Olympic history — that’s great. But the way I won gold didn’t change the fact that I had trained five hours a day, six days a week, for 12 years, to put myself in a position to get lucky.
Bradbury is a fantastic skater to make it into the finals of the Olympics: he is genuinely skilled at the sport itself. And Bradbury’s remarks do seem to suggest that he could have been closer to the front of the pack had he wanted to — he had the capacity to go faster and was deliberately holding back. But Bradbury’s observation of the metagame in his sport, and the strategy he crafted around it, only serves to augment his skill, not detract from it. There will always be some luck involved in pulling out the win, and playing suboptimally always carries the risk of ending in a loss, as suboptimal plays are prone to do. But at the highest levels of competition where everyone has mastered the game, metagame can and will be the differentiator — the way you put yourself in a position to get lucky.
In my view, when everyone plays fairly in a competitive game, “deserving” a win should be agnostic to the path to victory, because winning is all that matters.
Before next time
In the context of the market economy framework, metagame is reminiscent of irrational patterns of meta-behavior that markets frequently exhibit, undermining the hopes of academics that markets are actually efficient. Bubbles fueled by irrational exuberance, flash crashes, the Monday effect, and various other forms of market failures big and small occur because large populations of market actors can be swayed to trade the same way based on information that is irrelevant to the “true” value of the assets. Emerging victorious from donkeyspace is a lot like the Greater Fool theory of investing — it doesn’t matter if you buy an asset for a ridiculous price as long you can find a greater fool who you can sell it to for an even more ridiculous price. Ultimately, games that are steeped in a rich metagame may be less about playing well and more about not making the last mistake.
In my next article, we will discuss styles of play and contextualize this in terms of three phenomena in market economies: depreciation, leverage, and inequality.
