Author: Gerd Isenberg
Date: 11:34:33 10/13/03
Go up one level in this thread
On October 13, 2003 at 09:07:42, Anthony Cozzie wrote: >On October 13, 2003 at 08:59:00, macaroni wrote: > >>I'm sure if i do that, the new version will do dramatically better...but does >>this in fact add so much extra strength? and do I extend +1 on EVERY check? >>maybe I should run a 'fixed node' game, just to see if it's my incheck thats >>slow. Does my pawn extension system seem reasonable? >>thanks > >Essentially everyone extends +1 on check (either giving check or moving out of >check). > >_every_ check. No, i not. Only single move replies exceed one ply per se. I use fractional extensions and a lot of code and heuristics to determine the amount to extend. I extend also by x/N plies if attacking some squares around the king and a lot of other things. It's even a kind of delimited on how much extensions are already done in this path in comparison with total extends i this iteration so far, and with respect depth to search. Big and probably "wrong" extensions in conjunction with "bad" move ordering may lead to rather worse result in engine/engine matches. What most likely happens, the "bad"-extender sees the mate against it a bit earlier ;-) Gerd > >Here is Zappa analyzing the position from the Lasker Bishop Sacrifice post. > >With Check Extension: Bxh7 in 1.45 seconds, Mate score in 2.3 seconds, solved in >10 seconds. >1. Qc2 g6 2. c4 f6 > = (0.01) Depth: 4/11 00:00:00.15 50230kN >1. Qh5 g6 2. Qf3 Nc4 3. Bd4 Rxe3 4. Bxc4 > = (0.09) Depth: 5/13 00:00:00.43 139088kN >1. Bxh7 Kxh7 2. Qh5 Kg8 3. Bxg7 f5 4. Bd4 Qxd4 5. cxd4 Rxe3 6. Qg6 Kh8 7. fxe3 > = (5.39) Depth: 6/17 00:00:01.45 549222kN >1. Bxh7 Kh8 2. Qh5 Bg4 3. Bxg7 Kxg7 4. Qg5 Kh8 5. Qf6 Kxh7 6. Qh4 Bh5 7. Qxh5 >Kg7 8. Rg3 Kf6 9. Qf3 > = MAT9 Depth: 7/19 00:00:02.28 1049625kN >1. Bxh7 Kh8 2. Qh5 Bg4 3. Bxg7 Kxg7 4. Qg5 Kh8 5. Qf6 Kxh7 6. Qh4 Bh5 7. Qxh5 >Kg7 8. Rg3 Kf6 9. Qf3 > = MAT9 Depth: 8/18 00:00:03.40 1842380kN >1. Bxh7 Kh8 2. Qh5 Bg4 3. Bxg7 Kxg7 4. Qg5 Kh8 5. Qh6 Qxe3 6. Rxe3 Rxe3 7. Bg6 >Kg8 8. Qh7 > = MAT8 Depth: 9/18 00:00:05.75 3622526kN > >Without check extension: Bxh7 in 9 seconds, and still no mate score after 20 >seconds. > >1. Qh5 g6 2. Qf3 Nc4 3. Bd4 Rxe3 4. Bxc4 > = (0.09) Depth: 6/9 00:00:00.34 142375kN >1. Bd4 Qd6 2. Be5 Qc5 > = (0.00) Depth: 7/10 00:00:01.28 606191kN >1. Qh5 g6 2. Qh6 Rxe5 3. Rxe5 Qxc3 4. Re7 Be6 5. R1xe6 Qxd3 6. Re5 > = (0.79) Depth: 8/15 00:00:04.14 2187381kN >1. Bxh7 Kxh7 2. Qh5 Kg8 3. Bxg7 f6 4. Qh8 Kf7 5. Bxf8 Rxe3 6. Qg7 Ke8 7. Rxe3 > = (4.03) Depth: 9/13 00:00:08.78 4862209kN >1. Bxh7 Kh8 2. Qh5 Bg4 3. Qh4 Bh5 4. Qxh5 Qxe3 5. Rxe3 Nc4 6. Bf5 > = (8.44) Depth: 10/16 00:00:11.18 6514072kN >1. Bxh7 Kh8 2. Qh5 Bg4 3. Qh4 Qe7 4. Bxg7 Kxg7 5. Qxg4 Kh8 6. Rxe7 Rxe7 7. Rxe7 >Kxh7 > = (9.90) Depth: 11/18 00:00:12.95 7845541kN > >So Zappa is 4 times faster at solving this particular position with the check >extension, and that result is really pretty typical. > >anthony
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.