ernest wrote:Do you agree, and which version did you use?
OK, I found it: Latest compile from KLO Date 2010.11.20
Quote:
IvanHoe 999950t Beta Analysis/Game Mode combined in one plus Intel and Microsoft. Windows only -- Linux (Intel and GCC) build available upon request.
Which version are you testing: Intel, Microsoft? Default, popcnt?
Is there something to be checked in the "Engine parameters" in order to choose Analysis or Game Mode?
Intel(R) Core(TM)2 Quad Q9550 2.83GHz4x @3.5 GHz 4,096 MB Memory
Microsoft Windows XP 64 Bit Professional Service Pack 2 (Build 3790)
Fritz Benchmark:
Speed: 20.90
KNS: 10032
GUI: CB Rybka 3
Book: Kevin’s Huge Ctg-10 moves
Hash: 256
RB and TB: ON
Ponder: OFF
Blitz:10' 0
Intel(R) Core(TM) i7 4CPU 960 3.20GHz8x @ 4.005 GHz with 5.99 MB Memory
Microsoft Windows XP 64 Bit Professional Service Pack 2 (Build 3790)
Fritz Benchmark:
Speed: 24.98
KNS: 11988
GUI: CB Rybka 3
Book: HS-10 moves
Hash: 256
RB and TB: ON
Ponder: OFF
Blitz:10' 0
Intel(R) Core(TM)2 Quad Q9550 2.83GHz4x @3.5 GHz 4,096 MB Memory
Microsoft Windows XP 64 Bit Professional Service Pack 2 (Build 3790)
Fritz Benchmark:
Speed: 20.90
KNS: 10032
GUI: CB Rybka 3
Book: Rybka 4-10 moves
Hash: 256
RB and TB: ON
Ponder: OFF
Blitz:10' 0
The last sample game shows how IvanHoe and Rybka differ in evaluation (I intentionally do not say "understanding") of rook endgames. The first game shows what happens when White plays 1. h3?! The fourth game is almost a Father-like game, with the blocked structure. The fifth game shows the high/lows of RobboBases, with a strange reduction (Rxe6?!) into a "won" position that takes longer to win than doing it directly. With the tenth game (IvanHoe's win as Black), I don't think it signifies any real endgame superiority, but just that it so happened the value of passed pawns [at various stages] worked out to Black's favour in this case.
BB+ wrote:The last sample game shows how IvanHoe and Rybka differ in evaluation (I intentionally do not say "understanding") of rook endgames. The first game shows what happens when White plays 1. h3?! The fourth game is almost a Father-like game, with the blocked structure. The fifth game shows the high/lows of RobboBases, with a strange reduction (Rxe6?!) into a "won" position that takes longer to win than doing it directly. With the tenth game (IvanHoe's win as Black), I don't think it signifies any real endgame superiority, but just that it so happened the value of passed pawns [at various stages] worked out to Black's favour in this case.
Interesting analysis! As for the 1.h3, it comes from a test suite i designed. Some positions are purposely unbalanced. After a little over a thousand games, it might be safe to say Rybka understand it better than IvanHoe series.
The position below is one that IvanHoe tends to outplay Rybka though both won as white in the sample posted.