3 ms·
It looks like he's choosing the worst piece by looking one step ahead. In general, choosing strategies this way can back you into corners. I'd be interested t
by jrp 17y ago
It looks like he's choosing the worst piece by looking one step ahead. In general, choosing strategies this way can back you into corners. I'd be interested to know whether in this case, looking ahead 1 move is enough to get a global win as well (for the AI).
- eru 17y agoThe paper "How to Lose at Tetris" (http://www.geom.uiuc.edu/java/tetris/tetris.ps http://www.geom.uiuc.edu/java/tetris/tetris.ps) says that you don't even need to know what the player is doing. And that even almost all [1] random tetris games are bound to be lost, even with perfect play. [1] "Almost all" is a technical term in probability theory and means "with probability 1" (which is not the same as "all games".)
- brazzy 17y agoI wonder whether that is equivalent the same technical term from set theory which mean "all except for a finite number of exceptions", and thus with a finite underlying set, it's correct to say that "almost all" elements have a property when in fact none of them have it...
- jrp 17y agoIt's similar, but with a finite set it wouldn't show up unless you had some outcomes have probability 0 (in which case, why not just delete them from the outcome space?). Where you need the idea is in continuous spaces - like a uniform [0,1] random variable. With probability 1 it's not a rational number, etc.