Author: Gerd Isenberg
Date: 13:43:31 08/04/04
Go up one level in this thread
On August 04, 2004 at 15:27:46, José Carlos wrote: > In a test game, this position happened: > >[D]3r4/6k1/P5p1/1PP2p2/2Qnrb1q/1B5P/5PP1/R2R3K w - - 3 40 > > White, an old version of Shredder, was seing a7 here for some moves before, >with a plus score. At this point, at depth 11, it fails low on a7 (Qxf2 is very >strong then) and looks for another move. It chooses c6 with a plus score. Time >runs out and... c6?? > My impression is that Shredder was extending a7 (pawn to 7th), so it was >searching it a ply deeper than c6. But the refutation is the same! > A possible solution for this is not allowing extensions for the moves in the >root position. Another solution: to force another iteration after the fail low >in the root (or the idea of Sune, IIRC, of doing a depth+1 null window search >for the new root move, but only after a fail low in the root). Or even forcing >the extensions for the non PV moves at the root if the PV move was extended and >failed low. > Any experiences on the matter? > > José C. Hi José, puhh what a monster! Three connected passers and king attacks. Really a position where too agressive extensions near the root and path dependencies fire back. Considering so many passers and active pieces as well as both king under attack, one should amply reduce the extensions, specially pushing passers. IsiChess has to solve several fail lows but flips between a7, Rf1, Rxd4 (still positive) Qc3, Qf1, a7 again and finally settles with Qc3, after more than 3 minutes, -0.31 at depth 14. Looks like a nice tuning position - thanks. Gerd
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.