Author: Robert Hyatt
Date: 11:00:03 11/01/99
Go up one level in this thread
On November 01, 1999 at 12:32:35, Enrique Irazoqui wrote: >On November 01, 1999 at 09:30:30, Robert Hyatt wrote: > >>On November 01, 1999 at 02:32:48, Christophe Theron wrote: >> >>>On October 31, 1999 at 22:11:12, Robert Hyatt wrote: > >[snip] > >>>>We need two tests: A significant crafty vs tiger match. >>> >>> >>>So you say that the match played by Enrique was not significant? >>> >>>I don't understand. You need a hardware advantage, and you had a hardware >>>advantage in Enrique's match. You had the PIII-500, I had the PII-300. >> >> >>I don't have a clue what you are talking about. I have pointed out for _years_ >>that everything I do has been targeted at beating human players, in preventing >>them from using their normal anti-computer strategies. I'm interested in how >>you do against crafty. And how you do against humans as well. Or are you only >>interested in tiger vs computers, as you and Ed stated when you 'merged'??? >> >>As far as enriques test goes, I can't say much either way. I don't know what >>book. Etc. I only take responsibility for "crafty" as it plays on ICC because >>I know what it uses. > >Do you mean that games played by people other than you are not valid? > Did you see where I said that? I said that "I can't say either way. I don't know what book, etc..." >The book, the sound General that comes with F5, wasn't Crafty's, but it was much >better and bigger than Tiger's skeleton book of 7682 positions. I used this >midget book precisely to compensate for Crafty not playing with its own. This >book difference must have played in favor of Crafty by giving more computing >time to it. You can check for yourself that Crafty didn't dislike its positions >after book. So maybe Crafty didn't always play its favorite lines, but neither >did Tiger. One easy question... take a peek at several of the games. What was the eval on the first search out of book? Generally +, generally =, or generally - some value? The book is an integral part of the chess engine. That is why I distribute a book.bin with 'learning' in place, so that after taking a huge PGN file, and turning it in to a reasonably reliable book, I can still cull those outright blunders that always show up. > >I checked for differences in Crafty as a winboard engine and as a native engine >for Fritz, and I could not detect any, except for winboard running about 2% >faster in NPS than Fritz. I checked this with games and test positions: same >moves, same PVs. Below I copy some. > >Enrique I worry more about (when comparing anything else to 'crafty' on ICC): 1. what book? 2. EGTB's used? 3. what interface? 4. one or two computers? 5. what else is running? that is why I said _I_ primarily pay attention to results I get and let everyone else have fun with their 'crafty clones' on the servers. They can tweak with the book, tweak with the eval, tweak with the search extensions, and use what- ever hardware they want... But when I want to test/evaluate something I _always_ run it on my box. _always_... Too many variables otherwise... > > >Badai,B >8/1p6/p4p1p/2p1P2k/5P2/6pP/1P4P1/6K1 w - - 0 1 > >Analysis by Crafty 16.18: > >1.exf6 Kg6 2.f7 Kxf7 > ² (0.63) depth: 1 00:00:00 > µ (-1.15) depth: 3 00:00:00 >1.e6 Kg6 2.f5+ Kg7 3.Kf1 c4 4.Ke2 b5 > µ (-0.75) depth: 3 00:00:00 > µ (-1.12) depth: 19 00:00:46 13687kN >1.f5 > µ (-0.72) depth: 19 00:01:16 22727kN > µ (-0.72) depth: 19 00:01:16 22727kN > >Winboard: > >depth=19 -0.72 1. f5!! >Nodes: 24313481 NPS: 313843 >Time: 00:01:17.47 > >_______________________________________________________ > >Botvinnik,M - Capablanca,J >8/p5kp/1p2Pnp1/3pQ3/2pP4/qnP3N1/6PP/6K1 w - - 0 1 > >Analysis by Crafty 16.18: > >31.e7 Qc1+ 32.Kf2 > -+ (-3.76) depth: 1 00:00:00 >31.Qc7+ Kh8 32.Qb8+ Kg7 33.Qc7+ > = (0.00) depth: 4 00:00:01 > = (0.00) depth: 11 00:00:19 4898kN >31.Nh5+ > ² (0.40) depth: 11 00:01:01 14936kN > ² (0.40) depth: 11 00:01:01 14936kN > >Winboard: > >depth=11 +0.40 1. Nh5+!! >Nodes: 15174959 NPS: 251824 >Time: 00:01:00.26 > > >_______________________________________________________ > >Suetin - Kasparian,G >8/pr3pk1/4p1p1/q1pnP1Br/1pQ1R2P/3R4/PPP2P2/K7 b - - 0 1 > >Analysis by Crafty 16.18: > >1...Kg8 2.Rg4 > ± (1.31) depth: 1 00:00:00 > +- (1.47) depth: 2 00:00:00 >1...Qb5 2.Qb3 Qc6 3.Rg3 Qb5 4.Rc4 Kg8 5.Qd3 > ± (1.25) depth: 2 00:00:00 > ± (1.33) depth: 8 00:00:02 450kN >1...Kg8 2.Rd1 Rd7 3.Qb3 > ± (1.18) depth: 8 00:00:04 901kN >1...Rd7 2.Re2 Kg8 3.Red2 Qc7 4.Qe4 c4 5.Rd4 c3 > ± (1.17) depth: 8 00:00:07 1735kN > ± (1.17) depth: 9 00:00:13 3135kN >1...Qb5 2.Qb3 Qa6 3.Rd1 Rb5 4.Rg4 Kg8 5.Kb1 Ra5 6.f3 > ± (1.05) depth: 9 00:00:18 4575kN > ± (1.00) depth: 10 00:00:24 6253kN >1...Rxg5 2.hxg5 Nb6 3.Qb3 c4 4.Rxc4 Nxc4 5.Qxc4 Rc7 6.Qb3 Qxe5 > ² (0.70) depth: 10 00:00:37 9839kN > ² (0.70) depth: 10 00:00:37 9839kN > >Winboard: > >depth=10 +0.70 1. ... Rxg5 2. hxg5 Nb6 3. Qb3 c4 4. Rxc4 Nxc4 5. Qxc4 Rc7 6. Qb3 >Qxe5 >Nodes: 10686290 NPS: 279965 >Time: 00:00:38.17
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.