Author: scott farrell
Date: 09:32:09 11/30/02
Go up one level in this thread
On November 30, 2002 at 11:55:34, Russell Reagan wrote: >On November 30, 2002 at 10:39:45, Robert Hyatt wrote: > >>Simple. When you start an iteration, you should use some sort of aspiration >>window, rather than setting alpha=-infinity and beta=+infinity. In the case >>of Crafty, if the score for iteration N-1 was +.30, then I might start the >>search at iteration N with alpha=0, beta=+60 to try to bracket the expected >>score. This makes the search tree smaller as you can prune lines that lose >>or win material quickly. However, if you really do suddenly see a path to >>win a pawn, that score will be >= +.60, which means the search returns beta >>rather than a true score. You have to re-set beta to something larger and >>search again to get the true score. That is a "fail high". > >Let me see if I understand. Using this output from Monsoon: > >>10. +3.02 00:00:26.78 5824368 PV= Nf6 1. Bd6 Bc4 2. Rf2 e4 3. Bc2 >> > e3 4. Rg2 Qxf3 5. Re1 <+1.00> >>11+ +3.77 00:02:16.62 28726569 Nf4! ++ >>11. +5.19 00:04:27.40 58248289 PV= Nf4 1. gxf4 Rd2 2. Bf2 [threat] >> > Qxf3 3. Qxa5 Bd5 4. Qxd5 [threat] >> > Qxd5 5. Bxf5 Rf8 6. c4 Qxc4 [Q] >> > <+3.00> > >What happened was Monsoon used +3.02 from the search to depth 10, then it did an >aspiration search using something like (say) +2.27 to +3.77 (which is the score >of ply-1, +3.02, -/+0.75), and it got a fail-high as you describe it, and >immediately reported that score, without having a real PV. Then it re-searches >with a larger window and gets the real PV, and displays it with the real score >of +5.19. Do I understand this correctly? just tihnking about the output of Monsoon more: If they did search +2.27 to +3.77 - then Nf6 would probably have received a score, and be recorded before Nf4, due to move ordering. I think somehow they discarded Nf6 quickly based on the bounds. What if they first searched 3.76-3.77 - just looking for a fail high, that would exclude Nf6 and include Nf4 quickly. But it could have been 3.02-3.77, and Nf6 score dropped below 3.02. Scott
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.