Week 8: The lazy strategy must lose

If one way of playing beats every other, the rest of the design is decoration. For a while the best way to play the specimen was to run away.

27 April 2027

Is it me or is it the game? Write the laziest player you can: if it beats the intended way to play, the game is at fault. Running away once did, in an older version of the game: 102 s against 87 s.

Open the slides

Close Quarters is built on one idea: getting stronger means walking back into the fight. For under an hour one evening, while it was being built, that was false, and nothing on screen showed it.

The claim#

A strategy that wins regardless of what the game throws at it turns every other option into decoration. Players find it eventually. A simulation finds it in a second, if you think to write the lazy player as well as the good one.

The hour running away was correct#

Experience used to be scattered generously across the arena. A simulated player that only ever fled survived a median of 102 s. One that went looking for experience survived 87 s (56f46ba). The game’s premise had been quietly reversed by free experience worth as much as the kind you fight for.

Two changes put it back. Experience dropped by a kill became worth 2.5 times the experience lying free, in that same commit. And nothing is collected at a distance any more: you have to walk onto it (010a79d).

Where it stands#

Three strategies, the 100 ms player, 51 runs each:

The intended way to play is now also the best way to play. That is what “balanced” means for strategies in this course.

What you leave with#

Three degenerate strategies of your own for the specimen, their numbers against the intended strategy, and, if one wins, the one constant you would change.