Nonetheless, a creditable performance by the Houdini 6.01 engine showing useful gains against all competitors here. One benefit of using a machine with significant surplus RAM was that LP Hash was used throughout by Houdini 5.01 and 6.01 engines confirming there was insufficient RAM fragmentation to block the 4 Gb LP hash used by both.
Another important consideration is the influence of openings on engine performance. For this tournament I used my own compiled openings based on Human 2400+ games over the 12 months to mid April 2017. There was no engine analysis used to select the openings. Houdini 6.01 did not perform so well in the King's pawn lines and after 38 games was 2 games adverse against Stockfish 8 but it did pull one game back in the Caro Kann opening. The performance should be compared with the non-King's Pawn games section when it performed much better.
This does show how much random selection of openings from an opening book can adversely affect an engine's performance and skew Elo ratings. I am satisfied that most of the openings in my set reflects contemporary human preferences at Master + level and thus these results are broadly in line with what people are playing today.
Machine: dual Intel Xeon E5-2687w + 64 Gb RAM, HT fdisabled, Windows 7 Pro, Chessbase Fritz 14 GUI
Engines: 6 cores + 4 Gb hash each
Time Control: 40 moves in 5 minutes repeating to conclusion.
Syzygy 5 man TB's, GUI uses 5 man Nalimov and some 6 man.
Opening set: Derived from popular 2400+ player games with no engne influence.
Houdini 6.01 Pro vs Stockfish 8 +25 Elo
Houdini 6.01 Pro vs Houdini 5.01 Pro +45 Elo
Houdini 6.01 Pro vs Komodo 11.2.2 +48 Elo
I think Houdini 6.01 Pro could improve on this without the impediment of the Nalimov EGTB slowdown. Future matches may confirm this.
so that the Fritz GUI would only pass the path to an empty location, resulting in no slow-down.
> As a workaround, you could move the 6-man Nalimov to a different location on your HD,
> so that the Fritz GUI would only pass the path to an empty location, resulting in no slow-down.
That is effectively what I have done by creating DummyEGTB directory paths with no data in them to overwite the paths passed to the engines. The GUI paths remain to the true EBTB paths. There would still be a potential issue in an engine tournament if some engines such as Rybka for instance required Nalimov EGTB paths when the problem would return.
I have started a new full tournament with the top 4 engines as of Sept 2017 and there is no noticeable slowdown now with Houdini 6.01 approaching or in the endgame.
> Why not just use Arena.
Never spent any time looking at it. All of my work is done in Chessbase.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill