steve uurtamo wrote: > don, > > >> But I also discovered that there seems to be no benefit whatsoever in >> removing them from the play-outs. I have no real explanation for >> this. But it does tell me that the play-outs are very different in >> nature from the tree - you cannot just use the same algorithms for >> selection and prioritizing moves. >> > > did you use the same heuristic in the playouts when pruning them? > i.e. that no other stones are anywhere nearby? > > Yes, of course. Nearby means "touching" in this case because this is based on 3x3 patterns where the center point is the point in question.
I am of course aware that there are exceptions to this pattern, the move I am throwing out in some hopefully rare situations could be the best move. But my intuition tells me it should be more dangerous in the tree, and should make the play-outs far more relevant on average - but it doesn't seem to work that way! > in any particular playout, they may be the killing move for a group > that ends up getting built near to the edge of the board, or a successful > monkey jump. so removing them from somewhere deep in the tree > as a rule would be bad, but if nothing is anywhere nearby, removing > them as a first move choice is pretty reasonable. > I don't actually "remove" them permanently. They will still get tried when the simulation count gets high enough - I don't believe in absolute selectivity unless you can prove it's admissible. > they make up a fairly small fraction of possible moves, and this is > magnified the deeper you go into a search (so the net effect would > be diminished, even with the heuristic). > But the effect on the tree is pretty large. It's very useful to prune even a few moves and get a big total benefit if the pruning is sound. In practice you get MORE pruning as the game progresses and the center fills with stones until the edge starts getting touched. > anything you can say about what not to do on the very first move > of a search, if it applies to every board situation (ha ha ha) is > great, though, if it's fast enough to check for. > Pruning should be progressive. If anything, you should look at more moves near the root of the tree, not less. Of course this is subject to hard hard you have to work at it, and how sure you are of the rule you use. It's no good to see that an edge move is actually important but then not be able to see it when the position actually occurs on the game board. > in fact, if *nothing* is *anywhere* nearby, a move on even the > second line is bad. third or fourth is much better. > My patterns were collected statistically and moves to the edge were never played by my strongest program playing at long time controls, and thus the 3x3 edge pattern was generated by the automated pattern harvester. I think when I did 5x5 patterns I got some patterns just like you describe, moves to the 2nd line were rare if nothing was within the 5x5 pattern space. But if I remember they were not considered as horrible as edge moves. I don't remember if they made my veto threshold or not. - Don > s. > _______________________________________________ > computer-go mailing list > computer-go@computer-go.org > http://www.computer-go.org/mailman/listinfo/computer-go/ > > _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
