So far I had no success when NICE indicated a possible tuning improvement. But I must add these indications were very small, 0.1% or 0.2% improvements and testing them with cute-chess didn't gave the wished positive result, small regressions to equal at best.
That changed when I decided to test the [Passed Pawn = 100] parameter. Much to my surprise the lower I set the value the more the NICE score started to climb. A cute-chess match of 10,000 games confirmed a 5 elo gain.
http://rebel13.nl/mea/endgame-4000ms.html
NICE vs CUTE
Moderator: Ras
-
Rebel
- Posts: 7491
- Joined: Thu Aug 18, 2011 12:04 pm
- Full name: Ed Schröder
First fruit is in
90% of coding is debugging, the other 10% is writing bugs.