On Sun, 2007-09-30 at 12:41 -0400, Don Dailey wrote: > >> 4. correctness of random move selection strategy. > > > > Pick a random empty position. If illegal or eye-filling, remove from > > consideration the list and repeat. > > Same basic idea. I start by taking all filled points and removing them > from consider but I reinitialize after a capture.
The wording sounds like it may be wrong, and the implementation may be tricky. You start will all points and remove the filled points? If a selected move is illegal due to suicide, will you consider it again on later moves prior to a capture? Rejecting eye-filling moves from future consideration seems ok, but only if it's rejected for only a single move color... It could be the key capture play later. > >> I just thought of something. I think I initialize the statistics array > >> with 1 draw per move as a cheesy way to avoid divide by zero error. > >> Could this be affecting the performance? Perhaps at low levels like > >> this it has a noticeable effect? Would it make the program especially > >> vulnerable to an identical program that doesn't do this? > > > > Draw? Is that a valid outcome of your random simulations? I would > > assume your counters would be integers (and a draw value of 1/2 wouldn't > > work). > > A win is 1 and a loss is -1 so I set all the game counters to 1 - which > is like a draw. Of course in the simulations a draw can never happen. This seems strange to me. I guess you sum up all simulation results and count how many simulations there were? It sounds like if you did that, the winning rate would be (result_sum + num_sims)/(2 * num_sims)... or 0.5 + (result_sum/num_sims). It seems a little odd to me, but it works. _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
