The basic i think of Dynamic komi in my opinion is the following. The bot must be able to distinguish good from bad moves (this doesn't need much explanation i guess) and a bot '' by implementation" thinks that the opponent is as good as he is. (in most algorithms he is playing against himself)
The main way to distinguish between good and bad moves is that good moves lead to a win and bad moves lead to a loss. (or relative more chanche to a win and less change to a win). The problem can occur that with a to low (for the player) (internal) komi all moves lead to a loss (or a high losing chanche) and then the distinction between good and bad moves is factually lost. (this is what couses brain death moves in a for the bot lost game, the bot is unable to distuinguish between good and bad moves) The chanche that "all moves win" is much less likely some moves are so bad they will lose even against all odds. So the (internal) komi needs to be set so that good moves win and bad moves lose. (and that is in short the whole problem with komi how to make sure that this happens. In high handicap games the player who has black has an easier game but so a bot playing black (under normal komi) will think of almost all moves as good (they all lead to a win) and if he is playing playing White that all moves are losing (there is not even one good move) The internal komi most be set so that the bot thinks that only one (ore only a few) moves are good in the hope that the really good move (I mean the good moves on the board) is that one. During the game this all becomes dynamic. The opponent blunders (or a slightly sub optimal move) The bot will think it has (to) many more moves that will win, (and plays teh wrong one) the bot plays suboptimal and all moves lead to a loss. (and he cannot find any good one anymore) hope this helps On Thu, Feb 11, 2010 at 11:14 AM, Alain Baeckeroot <alain.baecker...@laposte.net> wrote: > Hi > > Le 11/02/2010 à 10:32, Le Hir Matthieu a écrit : >> Hi, >> >> > >> From what I think I understood, dynamic komi is supposed to try to keep the >> game more even. > The dynamic komi is a bias in the evaluation in order to "inform" the bot > that the game is balanced, and prevent it beeing "blinded" by the false > feeling of large win in the beginning of a high handicap game. > > Else with black and 9 stones the bot would "think" > i (bot) am ahead, just take it easy, this game is won, > whereas it is not, the 9 stones are needed, and we human know that white will > slowly but inexorably catch up. > > When the bot plays white, this is less a problem, as montecarlo bot will > correctly play safer and safer as the game advances. Maybe it could help, > by reducing the risks taken? > >> If the computer is black, playing at 9 handi, will the " burden komi" ( >> negative) be 9 x - 7,5 ? - 67,5 ? It sounds highly improbable. >> >> On the other hand, 9 handicaps are supposedly giving an advantage of 90 to >> 120 points, so my natural thought would be that the bot would give itself at >> least a negative komi of that many points ? > > Yes this is the idea. > The problem is to find the right balance, and not be slow when the bot aims at > being safe, and not force the bot to take unneeded risks > (you know better than me the pb slow/thick, thin/light, safe/crazy) > > .../... > >> I am going to play a series of high handicap games ( as white) against >> pachIV on kgs, that explains why I'm curious to know about how this >> komi-stuff works precisely. > no idea how/if this is implemented in different bots. > > Alain > _______________________________________________ > 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/
