Author: Ed Schröder
Date: 10:40:50 08/03/98
Go up one level in this thread
>Dear Ed >Could you briefly clarify the points below - just Yes/No will do if you like. >On August 03, 1998 at 08:55:25, Ed Schröder wrote: >[...] >>As for the requested differences between Rebel9 and Rebel10: >>Positions tested : 506 >Please, Ed, what sort of positions? Basically problem-type positions, or >what? A good mixture of positional, tactics and end-game positions. >>Different moves : 48 >>Different times : 349 (interval is -10%, +10%) >I can put several meanings here. >Does it mean that in 349 out of (506-48) positions (or 506 positions), the >time taken to make the move by R10+Anti-GM(Strong)was either >110% or <90% of the time taken by R9? At what depth (or other?) setting? >This 349 sounds like quite a significant number anyway! From the 506 positions 349 of them are slower or faster than 10%. >>Total time Rebel9 : 02:04:57 >>Total time Rebel10 : 03:00:10 >>Total percent : -31% >Does this mean in this test set, at (presumably) a fixed depth setting, >R10+Anti-GM(Strong) was 31% slower than R9? (I know Strong Anti-GM setting is >not as wise as SMART). Yes, fixed depth otherwise a comparison on time would make no sense. The test is done with anti-GM=OFF to make a good comparison with Rebel9. The -31% comes from the new search algorithm. >>Score differences >Are these absolute (i.e. unsigned) differences - surely R10+Anti-GM(Off) can >sometimes give lower evaluations at fixed depth than R9? I am not sure if I understand the question. Anyway anti-GM may produce higher or lower scores if that is what you asked for. >You wrote below: >>These figures are without the anti-GM options >which is why I believe these differences are between R9 and R10+Anti-GM(Off). Yes. >>Score 0.01 - 0.10 124 >>Score 0.10 - 0.25 27 >>Score 0.25 - 0.50 14 >>Score 0.50 - 0.75 7 >>Score 0.75 - 1.00 2 >>Score 1.00 - 2.00 0 >>Score > 2.00 4 >i.e. total 178. Yes. >How many showed 0.00 difference, please? Was it (506-48-178), or (506-178)? 506-178 >Sorry to ask so many questions, but I am sure many are very interested. Never mind, for me the statistic is self-understood as it is the first thing I check after running a test-version. I could have added some explanations myself in the first place. - Ed - >Kind regards >fca
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.