| |
BGonline.org Forums
Playing against the last two checkers - some BOT info
Posted By: Daniel Murphy In Response To: Playing against the last two checkers (Udo Moerschbach)
Date: Friday, 5 August 2011, at 10:48 p.m.
I mention some analysis here, thus the "BOT info" warning.
Or is there a reason not to trust the bot in this kind of proposition?
Udo, which bot? I have GNUBG. If I force Gnubg 2-ply to analyze all plays, it ranks 17/12 16/11* 6/1*(2) as the 60th best play (!). So I would be skeptical of a Gnubg 2-ply rollout.
Well I asked the bot now and he gives an equity of ~0.230 for my play while your play makes it only ~0.180 (or ~61.5% winning chances vs. ~59%). I only rolled out these 2 plays, maybe another play is still stronger.
Another loose hit play that looks decent is 16/11*/6/1 17/12. I agreed with Stick that loose hits didn't seem best. The builders just seem too much out of place. If a loose hit is best I will be very surprised or in need of more schooling.
Of the plays that hit once, I mentioned 16/11*/6 17/7. But if Blue is not going to attack the ace point now, there's really no rush to bring builders all the way around to bear on the ace point. So any of the plays that bring 3 or 4 builders to the 14 through 7 points should all play well. There's about half a dozen of these. And 16/11*6 13/3 look ok, too. IMHO.
If Blue closes out two checkers and has good spare distribution, Blue should win about 67.8% (says a rollout with spares on 6 5 3). I had estimated that Blue should close out White about 90% of the time. I believe that almost always Blue's closeouts will have excellent spare distribution. 90% of 67.8% is 61.0%. So your bot's results seem reasonably close to what I expected. But I'm not sure whether the results for the three types of plays (point, loose hit, no hit) should be equally reliable.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.