You example is perfect for explaining the strengths of Monte Carlo. And I think your analysis is wrong. A Monte Carlo program will choose the group that maximizes it win expectancies. If it see's all groups as 85% then it would clearly go for the biggest group since that has a bearing on the probability of winning.
The beauty of it is that if you have a complicated situation such as yours, but imagine each group has different odds of survival and there are many of them of different sizes. A human will probably take the naive approach of immediately defending the largest group because that is often the right choice. Or he will intuitively choose the move that is ON AVERAGE going to win the most stones. But the monte carlo engine takes a more measured and intelligent approach. It does a pragmatic calculation to determine which exact move gives it the best chances of winning the game. What less are you asking for? And it's not STUPID about this. Usually the biggest group is the one you want and it's not like the monte carlo programs purposely will go after the little group - they are not trying to minimize their score as some people seem to believe. They are doing what every player should be doing - maximize your winning chances. It's irrelevant whether they are good at this or not or if they make mistakes that cause them to lose games - the point is that they are focused on the only goal of winning to the best of their ability. It's just plain silly to focus on a less relevant goal and I don't understand the compulsion to make them play for less worthy goals. If it made them play stronger - then sure - go for it. But it doesn't work that way. I rather like this style of play. When the game is over and you have lost, I think it makes you look like a chump to fight as if you still believe you are winning. And even more I think trying to win big makes you look like a fool - at least it makes you look arrogant. I remember as immature children we would get upset if our opponents resigned (at chess.) We wanted the thrill of victory. And it wasn't good enough to just checkmate them, we had to wipe out every piece and pawn and then proceeded to checkmate them. As an adult you want to move on to the next game - this one is played out and no longer interesting. Monte Carlo programs are like this - the game is no longer interested and they don't pretend it is. With the greedy approach you recommend, it's possible that you could convert some losses to wins - but you have to program the machine first to win less! There is little point in that. - Don Dave Dyer wrote: > Here's a more likely scenario: Approaching endgame, there > are 10 "resolved" fights that remain to be played out. The > program estimates is won 5 of them and lost 5 of them, > each with 85% confidence. The sizes of the groups is > such that any single switch from won to lost will swing > the game. The probability of one such flip is pretty high. > > If one of the groups is much larger, you should go for the big > group, because if it switches, the game is flipped irrevocably, > whereas if one of the small groups flips, it might be corrected > by switching status of another group. > > That's why you should play for the big win, or to avoid the > big loss. It swamps all the other uncertainties. > > Of course, current programs don't see the situation with > anything approaching this level of analysis, but given > a choice or a big win or a small win with slightly higher > probability, I suspect that detailed analysis would be > similar to the above. > > _______________________________________________ > computer-go mailing list > [email protected] > http://www.computer-go.org/mailman/listinfo/computer-go/ > > _______________________________________________ computer-go mailing list [email protected] http://www.computer-go.org/mailman/listinfo/computer-go/
