Rybka's magic settings...I think I 've found them

[jix]

Lasker's piece values for the early part of the game. I was not expecting much from the settings, but they've worked.


rook pawn: ½
knight pawn: 1¼
bishop pawn: 1½
central pawn: 2
knight: 4½
queen bishop: 4½
king bishop: 5
queen rook: 6
king rook: 7
queen: 11

[jix]

Current tournament.....so far, so good. I should be done with the last 2 tournaments tomorrow morning, at which time I will upload all the games for your viewing pleasure.

Code: Select all

RvS Microsoft  2010

	
1 Deep Rybka 4 x64	        +12/=0/-0	100.00%	12.0/12
2 Stockfish 1.7.1 JA 64bit	+0/=0/-12	0.00%		0.0/12

[Albert Silver]

Current tournament.....so far, so good. I should be done with the last 2 tournaments tomorrow morning, at which time I will upload all the games for your viewing pleasure.

Code: Select all

RvS Microsoft  2010

	
1 Deep Rybka 4 x64	        +12/=0/-0	100.00%	12.0/12
2 Stockfish 1.7.1 JA 64bit	+0/=0/-12	0.00%		0.0/12

I hate to break this to you, but there is something wrong with your setup. If you look at the game scores you published, you will notice Houdini getting the same ply-depth as Rybka, when it should be getting at least 3 plies more. The same goes for Stockfish BTW.

So either ponder is in fact on, and Rybka is stealing all of Houdini's resources, or the settings within the engines, is causing only Rybka to be using 4 cores and the opponents are only using one.

[yanquis1972]

to me the idea is the king bishop is a better weapon against the king since it naturally develops against it & can attack f7 and h7 squares.

[yanquis1972]

im afraid albert is right. the reported stats here are pretty remarkably warped. i'm not even sure how it's possible to get #s that messed up. looks like rybka is on 1 cpu and houdini is on a half of one, yet you're running 4CPU versions on a quad. very odd.

[Event "RVH-4"]
[Site "Microsoft"]
[Date "2010.06.10"]
[Round "11.1"]
[White "Deep Rybka 4 x64"]
[Black "Houdini 1.01 x64 4_CPU"]
[Result "1-0"]
[ECO "B91"]
[Annotator "0.53;0.21"]
[PlyCount "117"]
[EventDate "2010.06.10"]
[EventType "tourn"]
[TimeControl "180"]

{Intel(R) Core(TM)2 Quad CPU Q9550 @ 2.83GHz 4394 MHz W=15.0 plies; 145kN/
s; 1,567 TBAs; Engine2.ctg B=14.2 plies; 427kN/s; Engine2.ctg}

[M ANSARI]

Make sure that you don't have ponder ON as then allocation of CPU sources are not guaranteed to be given to both sides equally. Can you give your exact setup for the match or tourney.

[yanquis1972]

fwiw, i ran a silver suite hyperbullet match with these settings against houdini and got nothing (i think after staying 20-30 elo ahead for awhile it ended up -8 when i stopped). not putting it out there as truth, but unless someone else comes up with similar results i'm assuming this is just a strong personality.

[Mark Mason]

As M.Ansari points out be careful here if this test has been carried out using ponder ON on a single PC. As we all know this is wrong as whilst engine A is thinking and preparing to move, engine B will also compete for the same processing capacity as it is pondering. This can lead to one engine getting use of all 4 cores with the opponent getting less CPU (sometimes only 1 or 2) which could explain the extreme results he got in Rybka's favour. For engine vs engine testing on a single PC you must either use ponder OFF or restrict each engine to only use half the total number of CPUs.

[LaurenceChen]

If you have a dual socket MOBO, that is 2 CPUs on the same motherboard, will Ponder ON after the 2 chess engines fighting for resources, since there is another CPU available for the other chess engine. Could someone confirm this, since I am thinking of purchasing a 2 socket MOBO.

[jix]

Ponder was off, but for some reason when the child process was set to "Normal" Houdini would not search as deeply as it should.

Im running a 10 minute tournament right now, and thus far the results again are in Rybka's favor. I will post the result after 25/100 games, but thus far its +3/2/0 after 5 games.