P.S. I'm sure their motivation behind this is to keep it simple and easy to compute. I suspect that most of the time it would probably come out the same using their cap, the square root, or the log of the scores anyway and I don't expect them to change based on this suggestion.
But it did occur to me that in the endless debate over dynamic komi and how you should score wins and losses, using the square root of the final points taken would be a vast improvement over just summing the points taken. It would have the desirable characteristic of giving some credit to bigger wins and yet not taking too much risk for those big wins. I have to believe that has been tried - *has anyone tried this in their MCTS program? * I would start the counting at the komi point, not at zero. For speed the square root could be table driven since there are only a handful of possible scores. Don On Thu, Jul 12, 2012 at 9:41 AM, Don Dailey <[email protected]> wrote: > Like you I agree that they should not not use this at all. However, > given that this is what they want to do I would argue that using using the > square root function (or something like it) is really what they want. > If they believe 150 is no better than 50, but that it's suddenly linear > below 50, what sense does that make? It goes from being linear to being > nothing. So using the log of the score, or the square root, anything > like that, is a huge win and gives them the basic behavior they really > want without having an arbitrary cap. > > And by the way, the square root is not just as arbitrary as capping - > unless you define arbitrary as anything you might do (which in a sense it > is.) So maybe arbitrary is not the right word here. What we are > looking for is something that is a more logical means to an ends - and the > cap is not nearly as good as the log or the square root. > > Don > > > > On Thu, Jul 12, 2012 at 9:25 AM, Álvaro Begué <[email protected]>wrote: > >> The square root is just as arbitrary as capping at 50. The only >> function I really like is capping at 0.5. >> >> Álvaro. >> >> >> On Thu, Jul 12, 2012 at 9:16 AM, Don Dailey <[email protected]> wrote: >> > Truncating to [-50 .. 50] seems rather arbitrary to me too. There >> should >> > either be no truncation at all, or if the concept is to not "over >> reward" >> > big wins it should be replaced by a function such as the square root of >> the >> > score. In this way you get progressively less credit for bigger and >> > bigger wins. >> > >> > Don >> > >> > >> > On Thu, Jul 12, 2012 at 5:48 AM, Rémi Coulom <[email protected]> >> wrote: >> >> >> >> Why truncate to [-50..50] ? >> >> >> >> On 10 juil. 2012, at 22:20, Łukasz Lew wrote: >> >> >> >> > Fellow Go enthusiasts, >> >> > I would like you invite you to: >> >> > >> >> > Kaś Cup - a peculiar computer Go tournament. >> >> > >> >> > There will be prize pool of total 100$, yay! >> >> > It will take place on 5th of August on KGS. >> >> > >> >> > The peculiarity will come from the scoring method. >> >> > While this will be a Round Robin, the score for each game won't be >> +-1 >> >> > point, >> >> > but the exact result of the game truncated to the [-50 .. 50] >> interval. >> >> > >> >> > One last rule is that participants may not use more than 4 cores of >> CPU >> >> > power. >> >> > >> >> > Nick kindly agreed to organize and look after the tournament for >> which >> >> > I am grateful. >> >> > Also he is in charge of choosing a ruleset and time settings. >> >> > >> >> > Thank you and let us know if you will participate. >> >> > Łukasz >> >> > _______________________________________________ >> >> > Computer-go mailing list >> >> > [email protected] >> >> > http://dvandva.org/cgi-bin/mailman/listinfo/computer-go >> >> >> >> _______________________________________________ >> >> Computer-go mailing list >> >> [email protected] >> >> http://dvandva.org/cgi-bin/mailman/listinfo/computer-go >> > >> > >> > >> > _______________________________________________ >> > Computer-go mailing list >> > [email protected] >> > http://dvandva.org/cgi-bin/mailman/listinfo/computer-go >> _______________________________________________ >> Computer-go mailing list >> [email protected] >> http://dvandva.org/cgi-bin/mailman/listinfo/computer-go >> > >
_______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
