Houdini 6 vs Stockfish dev 170917 on AMD Ryzen 1800x

[Hugo]

Setup Komodo with contempt 0

After second run and 100 games +8 -10 =82 so a tiny small win for Stockfish170917
A new match against Komodo 11.2.2 is now underway
In previously played match Komodo11.2.2 scored -17 +8 =75 against Stockfish 170917

I will share a pgn link with all the games afterwards

Completely different story here for Houdini 6 with same testbook +16 -7 =27 after the first run of 50 games (previous match first run was +4 -4 =42 against SF170917)

Code: Select all

Houdini 6 - Komodo 11.2.2, Blitz 3m+2s  2017
                                  
1   Houdini 6 Pro x64-popc   +63  +16/=27/-7 59.00%   29.5/50
2   Komodo 11.2.2 64-bit     -63  +7/=27/-16 41.00%   20.5/50

Second run against Komodo11.2.2 has finished with 13+ 5- 32 for Houdini 6
So after 100 games a clear win +29 -12 =59 is 58,5% for Houdini 6

Overall winner over 300 games in these three matches combined is Houdini 6
Houdini 6 +37 -22 =141 -> 107,5 out of 200 games
SF 170917 +27 -16 =157 -> 105,5 out of 200 games
Komodo 11.2.2 +20 -46 =134 -> 87 out of 200 games

I have yet started a new match Houdini 6 vs Stockfish 170917 with TC 15m10s

(games link will follow later)

[Nordlandia]

Considering H6 is stronger than K11.2.2, maybe it's advisable to set contempt at -3, which encourage Komodo to seek draw.

For Komodo ->

Positive contempt is good against weaker engines and negative contempt against stronger engines.

[stavros]

I am happy for Robert he beats a whole team of Stockfish developers

How has he beaten them if the score is equal ??
Anyway, thanks for the Results.
It shows that SDV will be ahead in a month or 2, while Houdini 7 will be released next year.
So......draw your own conclusions if one should buy H 6 or not .

Latest stockfish is still stronger. And well... it's only a matter of time before the team goes on another roll.

not to mention asmfish

[beram]

after 38 games at TC 15m10s still equality +5 -5 =28 (but although 5x longer TC two more wins yet)
avg game lenght 50-55minutes so will take take about two more days for full 100 games
I doubt there will be much difference in outcome with first match at TC3m2s
where Stockfish won +10 -8 =82, but lets see what happens

[beram]

Setup Komodo with contempt 0

After second run and 100 games +8 -10 =82 so a tiny small win for Stockfish170917
...
Second run against Komodo11.2.2 has finished with 13+ 5- 32 for Houdini 6
So after 100 games a clear win +29 -12 =59 is 58,5% for Houdini 6

Overall winner over 300 games in these three matches combined is Houdini 6
Houdini 6 +37 -22 =141 -> 107,5 out of 200 games
SF 170917 +27 -16 =157 -> 105,5 out of 200 games
Komodo 11.2.2 +20 -46 =134 -> 87 out of 200 games

..

Hi Clemens, doesnt matter that much I think
Hurnavich plays a match with TC 5m4s 8cores HERT openings between Komodo 11.2.2 with contempt 0 and Houdini 6 now http://rybkaforum.net/cgi-bin/rybkaforu ... 32286;pg=2
After about 50 games +13-5 is (also) 58% for Houdini 6

[Nordlandia]

K11.2.2 contempt of 0 is probable incorrect against H6.

Modest value of -2 or -3 can alter final score with more drawn games.

[Houdini]

It would be nice if engines received an estimate of the opponent's Elo at the start of the game so that they could adjust parameters accordingly .
It would add an extra dimension to the competition.

[fauzi]

can you please share the pgn?

[Jeroen]

Isn't it possible to estimate the quality of the opponent by evaluating its moves during the game?

An example: if the opponent makes a few inaccurate moves in the first 20 moves of the game, adjust the contempt factor etc.

[Nordlandia]

Isn't it possible to estimate the quality of the opponent by evaluating its moves during the game?

An example: if the opponent makes a few inaccurate moves in the first 20 moves of the game, adjust the contempt factor etc.

Good idea!