Since you’re here...

We hope you will consider supporting us today. We need your support to continue to exist, because good entries are more and more work time. Every reader contribution, however big or small, is so valuable. Support "Chess Engines Diary" even a small amount– and it only takes a minute. Thank you.
============================== My email: jotes@go2.pl



Chess engine: Stockfish 19111809

on


Stockfish - UCI chess engine, interesting compiled by Joost VandeVondele

 Use exploration rate for reductions 

 This patch measures how frequently search is exploring new configurations. This is done be computing a running average of ttHit. The ttHitAverage rate is somewhat low (e.g. 30% for startpos) in the normal case, while it can be very high if no progress is made (e.g. 90% for the fortress I used for testing). 

 This information can be used to influence search. In this patch, by adjusting reductions if the rate > 50%. A first version (using a low ttHitAverageResolution and this 50% threshold) passed testing: 

 STC 
LLR: 2.96 (-2.94,2.94) [-1.50,4.50] 
Total: 26425 W: 5837 L: 5650 D: 14938 Elo +2.46

LTC 
LLR: 2.96 (-2.94,2.94) [0.00,3.50] 
Total: 32313 W: 5392 L: 5128 D: 21793 Elo +2.84

However, as discussed in pull request 2414, using a larger ttHitAverageResolution gives a better approximation of the underlying distributions. This needs a slight adjustment for the threshold as the new distributions are shifted a bit compared to the older ones, and this threshold seemingly is sensitive (we used 0.53125 here). 

This final version also passed testing, and is used for the patch: 

 STC 
LLR: 2.95 (-2.94,2.94) [-1.50,4.50] 
Total: 16025 W: 3555 L: 3399 D: 9071 Elo +3.38

LTC 
LLR: 2.96 (-2.94,2.94) [0.00,3.50] 
Total: 37576 W: 6277 L: 5998 D: 25301 Elo +2.58



Rating JCER=3257

 Stockfish 19111809 - download


Comments

Post a Comment