I tested with Bob's 4,000 opening positions from EPD file, and it's very reliable, as Bob scrambled it from initial ECO sorting. My test was +/- 4 Elo points error 95% confidence, not some 17 Elo points total error 95% confidence. LOS was very close to 100% for IvanHoe. If you deny that at very fast time controls some IvanHoes beat the crap out of Rybka 4, you deny a simple evidence, presented clearly, a guilt punishable even by religious scholars .Albert Silver wrote:
Actually, I didn't see any proof. First of all you gave zero information on the test conditions, meaning you might have tested without books, or with. The only information you gave showed you played Rybka in a handicapped 32-bit environment.
As to being stronger... In all honesty I am not at all convinced Rybka 4 is stronger than Houdini 1.03a. I think they are roughly equal.
Kai