Caribbean Epiphany
A few months ago, I found a freeware translation of San Juan to a single-player computer game. Today, I had a chance to try it out.
San Juan is a translation itself, from a more complicated board game called Puerto Rico to a card game. Both are interesting, exploiting a clever if complicated turn order to inject agonizing strategic decisions. Each player, in turn, gets to choose one basic game function: producing goods, for example, or selling goods, or erecting buildings for points and special benefits. When a function is chosen, everyone performs it; so everyone builds on the same turn, everyone redistributes labor at the same time, and so on. And, since most functions get chosen sooner or later, the game revolves around timing the functions to a sequence that most benefits you.
The thought process goes something like this: “I need to get some cash so I can build this sugar mill, so a merchant phase would be good. But I don’t need the cash until a builder phase, and if I do a merchant phase now, Bob will sell his coffee for a lot of money and be free to do whatever he wants. I think Bob needs money, so he might do a merchant phase of his own. I think Carol wants a merchant phase, too, so I can wait for Carol as long as Bob doesn’t produce goods or build first. If Bob decides to produce, it’ll be too late for me to take advantage of a sugar mill, and if he builds, I can’t yet afford a sugar mill. Either way, I’m screwed. Okay, should I sell stuff now, and be certain of getting my sugar mill up and working, or go prospecting first, and hope Bob does a merchant phase?”
People, unless they’re Rain Man, can’t see more than two or three steps ahead in these complicated webs of sequencing, so building a competitive computer player isn’t too bad. Puerto Rico is complicated, but San Juan is simple enough that seeing four or five moves ahead by brute force is feasible, and in San Juan, that’s enough to offset a human player’s natural heuristic advantage. The computer opponent(s) in the freeware translation I got are pretty good, good enough that I lost my first few games by large margins, in the neighborhood of 36-24 and 25-19.
At first, I suspected cheating, but I’d just suffered some bad luck at the outset. Later, I began winning games by similar margins, and winning as often as losing. That made me feel better about myself, but at a price. It’s made me aware of how much San Juan depends on raw luck. I mentioned above that each building provides a specific benefit—anything from reduced building costs to extra production to victory point bonuses. Some combinations of buildings work very well together; others combinations are no greater than the sum of their component parts. Because San Juan has you draw potential building projects randomly from a deck of cards, rather than selecting them from the entire pool, you are dependent on the whims of fortune for the opportunity to build in an efficient, productive order.
Playing against humans, this wasn’t so obvious. Humans are less predictable, especially while learning the game, so crazy score distributions look like the scores we deserved for good or bad strategies. Repeated games between myself (past the learning curve, and no longer improving quickly) and the computer (not learning at all), the wildly swinging scores suddenly looked a lot more random. And suddenly, I’m a lot less interested in San Juan. I’m still up for Puerto Rico; a freeware version I’ve seen clearly penalizes dumb computer tactics, which suggests that good strategies pay off. But San Juan just looks like a matter of luck. Flipping coins is faster and less effort.