Recently I had a discussion with one of my online friends about Go. I told him that despite my tendency to “abandon” the game for months at a time, I always seem to come back to it. I had difficulty articulating why that’s the case, but it led me to thinking about what makes a good game. In general, I think a good game should create an environment in which the player possesses the ability to affect that environment as a player, not as a character. This can include metagame elements such as completion times and scores, as in classic arcade games, or it can be based on an individual’s command of the playing environment, as in the perfect information abstract games. This concept can be called “consequentiality,” the idea that the player’s actions have a real effect on the outcome of the game. I think Go epitomizes the concept. Beyond the turn-based structure of the game, it makes few assumptions about the players or the flow of the game. For a counter-example, let’s look at online RPGs…

Anyone who’s talked to me about games will realize that I hate MMORPGs. With the exception, perhaps, of the socipathy simulator EVE Online, these are the exemplars of games based around being time sinks. That’s their business model; a player pays $14.95 per month, so it is against the company’s interests to make it possible to “finish” the game. The games are structured around loot, and there is no functional difference between a “good” player and a “bad” player. This may seem nonsensical if you’ve never played one of these games, but allow me to illustrate with two imaginary items.

To someone who plays a MMORPG, these item tables will look immediately familiar. They will tell you that the first weapon produces 5–14 damage, for an average of 9.5. 9.975 over many hits if the game uses a 1.5x critical modifier. The second weapon produces 9–25 damage, for an average of 17, or 18.275 over many hits. The second weapon is obviously better: if you give a “good” player the first item and “bad” player the second, the bad player will win every fight. The “balancing factor” in most MMORPGs is the loot model. Obviously the second item is better, but it’s “rarer.” The first sword might be available from a standard weapon shop in a safe town; the second drops 1% of the time from the Mad Mushroom of the Fungus Hills, which spawns only once an hour.

Now, randomness isn’t necessarily bad. Roguelikes like Crawl and Nethack rely on similar damage, combat, and loot models, and I consider them very successful and well-executed games. The difference is in constraint. In Crawl, for example, the amount of loot is limited because levels can be cleared and repopulate only very slowly. The food supply is a limiting factor, a de facto time pressure forcing you to explore and requiring decisions about balancing resource use with “farming” for loot. MMORPGs have no such constraints. The world is static, support structures tend to be extensive, and the outcome of your game—such as it is, given that MMORPGs have no failure states—is determined by the time you have to spend on it. This is why they’re so repetitive; the game design is not based on concepts of success and failure, but upon chance. With no time constraints, the chance becomes dreary inevitability.

So. My favorite games remain games of strategy and tactics (Go), of reflex and aim (Team Fortress 2 and other action games), and of limited resource management (Crawl, Spelunky). All are consequential: a player’s action determines his success or failure, and his skill at the game can mostly mitigate random effects. There are critical hits in TF2, while Crawl and Spelunky are heavily randomized, but all rely more heavily upon the player’s understanding of the playing environment than upon inconsequential elements such as luck-based damage or repetition.