Author: James T. Walker
Date: 20:21:47 04/09/02
The T-notes at chessbase suggest that giving Fritz 7 a large amount of hash memory may effectively "Labotomize" Fritz at short time controls. The T-note suggest that you should use the formula to determine the optimum value of hash tables. While I believe this was true of Fritz 5, I don't think this applies to Fritz 7. Here is some data to back up my opinion: All games played on 2 Athlon 1.4G machines using auto232. Games were played at Game/2 minutes. Fritz 7/256Meg hash vs Fritz 7/4Meg hash: Score 98-82 (This shows a +30 Elo for 256M hash.) Fritz 7/256Meg hash vs Fritz 7/10Meg hash. Score 136.5-149.5 (This shows about a 16/17 point Elo advantage for 10 Meg) Fritz 7/256Meg hash vs Fritz 7 16 Meg hash. Score 54.5-49.5 (This shows about 14 point Elo advantage for the 256 Meg hash. None of these stats appear unusual for two identical programs playing on two identical computers with the only difference being the amout of hash memory. It's possible the 4 Meg hash was too small for the time control. I see nothing to indicate that 256 Meg hash is "too big" and causes a Fritz labotomy. I conceed that 10 meg at this time control may be optimum but many games would be needed to prove it since the Elo difference is very small. As usual all comments/insults are welcome. Jim P.S. Warning: I just heard a guy on TV say "If you're too open minded, your brains may fall out."
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.