Author: Sune Fischer
Date: 18:55:51 12/28/03
Go up one level in this thread
On December 28, 2003 at 21:29:07, Federico Corigliano wrote: >Hi Sune > >Here are my results for the two FRC positions that you posted some days ago: > >In this FRC position: >1r1k1r2/ppp1bppp/2np3q/4p3/b1B3Q1/N2PP1Pn/PPPB1P1P/1R1K1R2 w KQkq - 0 1 >you results are: >perft 1 42 >perft 2 1533 >perft 3 58858 >perft 4 2208844 >perft 5 82619281 >perft 6 2336234512 > >My results are similar until depth 5, I get 82619285, only a difference of 4 >moves. >Here is my divided perft 5: > >c4b3 -> 2253975 c4b5 -> 2006928 c4a6 -> 2020382 c4d5 -> 2583005 >c4e6 -> 2446340 c4f7 -> 2264156 g4f3 -> 2127573 g4e2 -> 1548674 >g4h3 -> 1545140 g4f4 -> 2013791 g4e4 -> 2211628 g4d4 -> 2102126 >g4h4 -> 1973755 g4f5 -> 2073936 g4e6 -> 1785245 g4d7 -> 32388 >g4c8 -> 58214 g4g5 -> 1721744 g4g6 -> 1964836 g4g7 -> 1956370 >g4h5 -> 2088914 a3b5 -> 2389764 d3d4 -> 2423248 e3e4 -> 2013702 >b2b3 -> 2188632 b2b4 -> 2010410 d2c1 -> 1550346 d2e1 -> 1899175 >d2c3 -> 2295277 d2b4 -> 2192149 d2a5 -> 2111405 f2f3 -> 1967134 >f2f4 -> 2071002 b1a1 -> 2109142 b1c1 -> 1916246 d1c1 -> 2102930 >d1e1 -> 2114506 d1e2 -> 2165102 d1c1 -> 2037027 f1e1 -> 1999123 >f1g1 -> 2187382 f1h1 -> 2096463 > >Maybe you can implement this command in Frenzee and tell me your results. Is >easy to make, you generate the moves list and do and perft (depth-1) for each >move. Here are my numbers, but it's not as easy to compare due to the move ordering :) -->: perft 5 Bc4xf7 ~ 2264156 [0.33 secs] Qg4xh3 ~ 1545140 [0.21 secs] Qg4xg7 ~ 1956370 [0.25 secs] Kd1c1 ~ 2037027 [0.23 secs] b2b4 ~ 2010410 [0.24 secs] f2f4 ~ 2071002 [0.25 secs] b2b3 ~ 2188632 [0.25 secs] f2f3 ~ 1967134 [0.22 secs] d3d4 ~ 2423248 [0.26 secs] e3e4 ~ 2013702 [0.22 secs] Na3b5 ~ 2389764 [0.23 secs] Bd2c1 ~ 1550346 [0.15 secs] Bd2e1 ~ 1899175 [0.18 secs] Bd2c3 ~ 2295277 [0.21 secs] Bd2b4 ~ 2192149 [0.20 secs] Bd2a5 ~ 2111405 [0.19 secs] Bc4b3 ~ 2253975 [0.20 secs] Bc4b5 ~ 2006928 [0.18 secs] Bc4d5 ~ 2583005 [0.20 secs] Bc4a6 ~ 2020382 [0.19 secs] Bc4e6 ~ 2446339 [0.19 secs] Rb1a1 ~ 2109142 [0.15 secs] Rb1c1 ~ 1916246 [0.12 secs] Rf1e1 ~ 1999123 [0.12 secs] Rf1g1 ~ 2187382 [0.12 secs] Rf1h1 ~ 2096463 [0.11 secs] Qg4e2 ~ 1548674 [0.07 secs] Qg4f3 ~ 2127573 [0.12 secs] Qg4d4 ~ 2102126 [0.12 secs] Qg4e4 ~ 2211628 [0.09 secs] Qg4f4 ~ 2013791 [0.07 secs] Qg4h4 ~ 1973755 [0.07 secs] Qg4f5 ~ 2073936 [0.04 secs] Qg4g5 ~ 1721744 [0.06 secs] Qg4h5 ~ 2088914 [0.06 secs] Qg4e6 ~ 1785244 [0.05 secs] Qg4g6 ~ 1964836 [0.07 secs] Qg4d7 ~ 32388 [0.01 secs] Qg4c8 ~ 58214 [0.00 secs] Kd1c1 ~ 2102930 [0.05 secs] Kd1e1 ~ 2114506 [0.05 secs] Kd1e2 ~ 2165100 [0.07 secs] Just a thought, maybe this is better suited for email (I fear it could become a lenghty FRC debug session :)? >In this other FRC position: >Rr3kr1/2p2ppp/2nbp1n1/3p4/1pB1q1b1/1P2P1N1/1PPP1PPP/1NB1QKR1 b Kkq - 0 1 >your results are: > >perft 1 47 >perft 2 1335 >perft 3 60626 >perft 4 1754084 >perft 5 79306332 >perft 6 3163057546 > >and I get the same. I can't make perft 6 in my poor PC, so I don't have results. Took about 30 seconds with a hash :) >I'm now converting Chispa from C to C++, and the most difficult work was create >a CPosition class. Now that it's done, Chispa's move generation is about 7-10% >slower, but the code is a lot more cleaner. >I will do tomorrow a full test with a IsOKthePosition() function that checks all >the aspects of the position. It helped me with the normal chess move generator. Yes I'd strongly recommend one of those, it has helped me fix a lot of bugs. -S. >Regards, >Federico
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.