Author: Dave Gomboc
Date: 02:11:14 10/26/99
Go up one level in this thread
On October 25, 1999 at 14:31:21, Scott Gasch wrote: >On October 25, 1999 at 12:39:01, Jeff Kenton wrote: > >[...] >>Given that checkers has fewer choices at any move but tends to require deeper >>look ahead, which of the programming look ahead strategies do you consider to be most suitable? > >I think that a standard alpha-beta search would do very well on checkers -- it >would be like an endgame position in chess... a fast search can routinely get >over 15 ply deep with such a low branching factor. In some endgame positions in >chess (i.e. fine70) it's easy to get over 30 ply deep... I guess an "endgame" in >checkers would be even better. > >On a related note: didn't someone prove that checkers was a closed game? >Whoever moves first wins if they play perfectly? If this is indeed the case, >you might do no search at all and instead but a big hard drive and put the >tablebase on it. Then you can play the "best" move from all positions...? >However I am not sure if this is true or just me being confused. > >Scott Checkers is most likely a draw, not a win for the first (or second) player. Jonathan has about 10 gigs of endgame data. This doesn't sound like much now. It was significantly more impressive several years ago when it was collected! In fact, it fits into _main_memory_ on one of the computers we have at the U of A now. :-) Dave
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.