Some main differences in mogo between 9x9 and 19x19: - we have no pattern database for the tree in 9x9 (because we have not done it yet, but perhaps it would be efficient), and go-expertise (rules used for introducing a bias in the tree) is nearly useless in 9x9.
- I have read that Rave values are supposed to be especially efficient in 19x19, but I guess it is essentially efficient in 19x19 when you have no go-expertise - in my humble opinion, if you have patterns or go-expertise, it is in 9x9 that rave values are most efficient (I'm not sure of that, it's just a feeling when comparing our results) - the positive effect of randomization (if any) is much smaller in 9x9 - the thresholds for self-atari are different - the relative weight of amaf-values and standard values is different - the formulas for the bias depending on the distance to the last move are different - the formulas for the time per move are also different (more complicated in 19x19, it's not only parameters) I probably forget plenty of differences :-) Can you put crazystone back on 19x19 so I can see if it is just a fluke > against fuego? > > I added locality to the light playouts - play near last or second to last > move, and some code to handle long ladders in playouts. I don't think this > is anything unusual. > > Both should help 19x19, but I don't know why 9x9 would be damaged. > > I also changed code around playing self atari in seki. Perhaps there is > bug, since seki is more important in 9x9. > > The biggest problem that I have with monte carlo is the amount of testing. > I made four or five changes and I'm testing them together now. To see the > effect of each change independently would take a couple of weeks of > testing, > since it takes several days on 19x19 to get an accurate rating. > > Regards, > > > David > > > > -----Original Message----- > > From: [EMAIL PROTECTED] [mailto:computer-go- > > [EMAIL PROTECTED] On Behalf Of Rémi Coulom > > Sent: Tuesday, September 09, 2008 9:20 AM > > To: computer-go > > Subject: Re: [computer-go] 9x9 to 19x19 scaling strangeness > > > > David Fotland wrote: > > > I made a change over the weekend, which looks like it makes 9x9 150 > > ELO > > > weaker and 19x19 over 200 ELO stronger. > > > > > > Very strange. > > > > > > David > > > > > > _______________________________________________ > > > computer-go mailing list > > > computer-go@computer-go.org > > > http://www.computer-go.org/mailman/listinfo/computer-go/ > > > > > Now you must tell us what the change was :-) > > > > Rémi > > _______________________________________________ > > 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/ > -- ========================================================= Olivier Teytaud (TAO-inria) [EMAIL PROTECTED] Tel (33)169154231 / Fax (33)169156586 Equipe TAO (Inria-Futurs), LRI, UMR 8623(CNRS - Universite Paris-Sud), bat 490 Universite Paris-Sud 91405 Orsay Cedex France
_______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/