Not logged inRybka Chess Community Forum
Forum RybkaChess Help Search Register Login
Up Topic Rybka Support & Discussion / Rybka Discussion / RYBKA 4.1 SETTINGS
- - By chess_pr0 (**) [us] Date 2012-04-21 15:38
I KNOW THE SETTINGS FOR RYBKA 4  U MULTIPLY BY 10 TO GET 4.1 SETTINGS.. BUT CAN SOMEBODY EXPLAIN THE BOTTOM 4 BOXES IN SETTINGS WHAT U PUT THERE?

also can u post the experiment 79 settings here too please as well?

SCORE OFFSET MILLIPAWNS = ??
ROOK ENDGAME PENALTY MILLIPAWNS = ??
BISHOPS ARE OPPOSITE COLORED PENALTY MILLIPAWNS = ??
OPPOSITE COLORED BISHOPS ENDGAME PENALTY MILLIPAWNS =??
ROOK ENDGAME SCALING = 100

would be better if someone had a picture of the 4.1 settings box with the experiment 79 settings put in so i can see and make sure im doing right.. thanks in advance..
Parent - - By Uly (Gold) [mx] Date 2012-04-22 00:35

> SCORE OFFSET MILLIPAWNS =


Allows the engine to avoid draws in analysis if it seems them near the root (using a positive number). Or assuming the opponent wants to avoid draw (using a negative number.) Useful when you're analyzing positions that are a sea of 0.00s and you want to know more differences on the moves, and also because in the process Rybka would prune more drawn positions and build a more useful tree.

> ROOK ENDGAME PENALTY MILLIPAWNS = ??


Sometimes, Rybka assigns a winning score for positions that are draw in rook endgames (both sides with one rook and no minor pieces), this can change that by adding penalties to such positions (actually, Vas claims they're not drawish, but are to be avoided because it's harder to beat weaker opponents in them).

>BISHOPS ARE OPPOSITE COLORED PENALTY MILLIPAWNS = ??


The same, but for positions where each side has only one Bishop, and they're on opposite colors.

>OPPOSITE COLORED BISHOPS ENDGAME PENALTY MILLIPAWNS =??


The same, but for endgames.

>ROOK ENDGAME SCALING = 100


Allows you to scale down the score towards 0 in rook endgames.
Parent - - By Weirwindle (***) [us] Date 2012-04-30 02:34
I am not so sure that R4.1's parameters should just be 10x that of R4.
I think Vas did a little more work to the piece weights before the 4.1 release.

After 1000 3'+2" games of R4.1 default vs R4.1 with R4 advanced settings x10
the settings are weakening the engine.
Parent - - By Uly (Gold) [mx] Date 2012-05-01 19:29
Yes, the same can be said about Rybka's time control settings, where Rybka changed her time management but nobody tested for better defaults.
Parent - - By Weirwindle (***) [us] Date 2012-05-02 01:12
A total revamp of Rybka's time management should be considered rather than just finding
better default values.
Parent - By Uly (Gold) [mx] Date 2012-05-02 01:56
But will Vas do it? It was suggested for Rybka 4.1, and Vas didn't do it, chances are he won't revamp it for Rybka 5 either.
Parent - By M ANSARI (*****) [kw] Date 2012-05-02 19:29
There is no doubt that a proper time management scheme for Rybka 4.1 would help dramatically in some setups.  I don't think it would add that much to the ELO of Rybka 4.1 in longer time control games, or in time control games that are at tournament speeds but with Ponder ON ... but in fast time control games with Ponder OFF I can easily see an increase of 50 ELO or even 70 ELO or more.  Doesn't mean that Rybka 4.1 becomes a stronger chess engine, it just means that Rybka 4.1 time management in some setups is way below that of the top engines today.  A lot of that has to do with the fact that Rybka 4.1 is optimized for a cluster setup, where latencies are inherently higher than they are on a PC machine.
Parent - - By Pia (***) [ru] Date 2012-05-09 00:29

>SCORE OFFSET MILLIPAWNS


I'm lovin' it

Mate in 10 (H.Grasemann, 1981)

7b/5Bnr/7k/8/4K1pP/p3N1p1/P3b1P1/8 w - - 0 1


Analysis by Deep Rybka 4.1 w32-2: (default)

1.Ke4-d5 Be2-d1[] 2.Kd5-c4[] Bd1-e2+[] 3.Kc4-d5 Be2-d1[] 4.Kd5-c4[] ......
  =  (0.00)   Depth: 38   00:03:50  57218kN, tb=569

Analysis by Deep Rybka 4.1 w32-1: (score offset = 9999)

1.Ke4-d5 Be2-d1 2.Kd5-c4[] Bd1-e2+ 3.Kc4-b4[] Be2-d1 4.Kb4xa3[] Bd1-e2 5.Ka3-b4[] Be2-d1 6.a2-a4[] Bd1-e2 7.a4-a5[] Be2-d1 8.a5-a6 Bd1-e2 9.a6-a7 Be2-d1 10.a7-a8Q Bd1-e2 11.Qa8-c6+ Ng7-e6 12.Ne3-f5#
  +-  (#12)   Depth: 12   00:00:01  134kN
Parent - - By Barnard (Bronze) Date 2012-05-09 00:38
and even with that,rybka fails to find the mate in 10 :lol:

maybe rybka is soo generous to allow the opponent to 'live' to 2 moves more...and we never thought about that possibility :lol:
Parent - - By Pia (***) [ru] Date 2012-05-09 00:51
No, she making special combo critical strike moves to make opponent lose its balance for next games.
Parent - By Barnard (Bronze) Date 2012-05-09 01:08
:lol:
Up Topic Rybka Support & Discussion / Rybka Discussion / RYBKA 4.1 SETTINGS

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill