Folding Space

Recently, games-wise, I’ve been taking a bit of a break from my usual artsy and the pretentious indie fare to play a bit of Poker. Poker Night at the Inventory that is– what, did you think I had the money to play for real? I’m afraid not.

I keep on being startled by just what a fantastic game Poker is– at least the Texas Hold ‘Em variant (I am not particularly familiar with other variants). Surprisingly, or at least surprisingly to me, the video game experience which I am most reminded me of is FTL.

In FTL, you play the captain of what is usually a skeleton crew manning a Federation cruiser, jumping between stars just a few steps ahead of a pursuing Rebel fleet. With each jump, you use a bit of fuel and a bit of time, both of which are limited resources. And, with each jump, you make a series of decisions which can either help you, such as providing another crew member or a weapon for your ship, or harm you, such as leaving you a skeleton in a burning husk of a ship drifting through an eternal black void.

It’s a great game for as much time as it takes to explore the many different exotic scenarios, find out what the expected returns on each play are, and which particular resources are necessary to complete the game– at which point it becomes a tedious slot machine. That took me about 40 very enjoyable hours, though, so I would highly recommend checking it out.

Playing Poker, the flow of the game is much like that of FTL. Each hand is dealt, and our limited resource pool dwindles, and we are presented with a scenario. From here, we must determine what approach is most likely to lead to long term success based on a statistical analysis.

It’s funny how many games, when you get right down to it, boil down to statistical analysis– this lends some credence to my earlier arguments on the main difference between work and play being context, but I digress.

However: Unlike FTL, Poker never becomes a simple slot machine. Despite the fact that the parameters and inputs of the game are vastly more limited than in most strategic games, and the undeniable role that luck plays in the outcome, Poker never becomes a simple game of pat choices the way that FTL does once you understand it. For one thing, the number of possible outcomes for any given hand are far more nuanced and intricate than determining whether I have the armaments necessary to take out an enemy ship or which unit I should use to augment my forces against the army I have scouted out.

There is no calculus that can tell me whether, if I double his bet, my opponent will take the bait or be intimidated into taking his money away– a mind game entirely separate from the statistical analysis of determining whether I probably have the stronger hand in the first place. It’s almost like two separate games, one played with numbers and one played with wills and wits, woven together to form an intricate whole.

Nevertheless, I suspect that it won’t be long before the AI in this game starts to wear thin, before I see through the thinly spackled dividing lines between the baits they’ll take and bets they’ll fold to. Which is not to say that people don’t have their own quirks that a player can learn and exploit– quite the contrary, they probably have more– but they also have learning and inquiring minds that seek out the player’s own weaknesses, bad habits, tells, and turns them against that player.

Computerized opponents never learn about us, but oh we learn about them.

We always find the limits of single player games eventually. Any game made to be enjoyed by a single player is limited in a way that allows the player to account for all given inputs. Any randomized numbers can be accounted for and averaged out to create an optimal play. Single player games, particularly those which don’t rely on physical coordination or reflexes, can invariably be solved to provide, if not consistent victory, the highest consistent chance of victory.

And thus it shall always be, until we create AIs who learn about us, and begin to outguess us. Of course, then we might start feeding them misleading information about our behavior and game them that way, but then perhaps they’ll become smart enough to account for that behavior as well and– how many layers deep will we have to go, do you think, before we cease to be able to outsmart our machines? How much deeper past that before they begin to find us tedious and predictable?

It’s another kind of Turing test with more immediate applications. And, really, once we near that point, are these even single-player games any more? The distinction begins to blur. As technology progresses, as games become more connected and AIs become more advanced, there’s going to be a trend of convergence between single and multi-player games. This won’t be absolute, of course– What would be the point of a multi-player Braid?– but games which we tend to think of as single-player will become multi-player, as players drop in and out to take up the mantles of NPCs in these worlds, and games which we tend to think of as multi-player will become single-player, as our allies and enemies rage-quit and are seamlessly replaced by bots who have learned to emulate their behavior.

Or perhaps not. I am no futurist (in that I do not demand exorbitant speaking fees to share my predictions). And maybe the lines I’m drawing here are drawn in the sand at low-tide, and maybe as our understanding of the world and of what games can be in it progresses they will all become laid bare as predictable, as tedious, as uninteresting slot machines that can only play on our compulsions.

I hope not. But if it be so, I hope that these insights will help us to build newer and better games, games so sophisticated that even we cannot break them.

Someday. Someday.