Cubeman wrote:Please help me to understand if Ippolit is actually based on Rybka? So far I have not heard any compelling evidence, only Vas making a statement along those lines.I believe that Houdini is based on Ippolit, but am unsure if that also implies based on Rybka.
From the similarity tool:
Key:
1) Fire 1.01 (time: 27 ms scale: 1.0)
2) Houdini 1.00 (time: 29 ms scale: 1.0)
3) RobboLito 0.085g3 (time: 27 ms scale: 1.0)
4) Rybka 3 (time: 33 ms scale: 1.0)
5) Stockfish 2.11 (time: 31 ms scale: 1.0)
1 2 3 4 5
1. ----- 71.56 73.80 59.04 47.71
2. 71.56 ----- 71.67 59.37 48.34
3. 73.80 71.67 ----- 58.86 47.80
4. 59.04 59.37 58.86 ----- 49.96
5. 47.71 48.34 47.80 49.96 -----
I personally believe the percentages would be higher except for the fact that the self-simi;arity percentages Rybka 3, the Robbo/IvanHoe engines, and Houdini are only ~75%, unlike most engines that have self-similarity percentages of 95+%.
The similarity numbers between Rybka 3 and the others (Robbo, Fire, Houdini) is enough to raise my suspicion. The likelihood of these numbers, when compared to an estimate for the entire population of engines, is quite low. Mark Watkins has stated that whoever made Ippolito had a good amount of familiarity with Rybka 3. I believe Chris Conkie and others have shown that those engines show similar characteristics to Rybka 3. It is hard to have much doubt that Ippolito was born from Rybka 3. I guess the only questions are in the details.
Note to Kai Laskos:
I was wrong about the low self-similarity numbers being related to the positions. It does show up with other positions.