H5 vs SF8 on 40mvs in 60 min

[Hugo]

hi all

Houdini 5 standard vs Stockfish 8.
both engines 6 cpu, 4GB hashtables
ponder off, no tablebases
testsuite 50 possitions, 100 games
40 moves in 60 min repeated
regards, Clemensd Keck

[arunsoorya1309]

So SF scales betters than H5. ... H5 in TCEC is getting a pounding

[beram]

hi all

Houdini 5 standard vs Stockfish 8.
both engines 6 cpu, 4GB hashtables
ponder off, no tablebases
testsuite 50 possitions, 100 games
40 moves in 60 min repeated
regards, Clemensd Keck

Thx Clemens,
Surprisingly good result for H5 if I may say so
btw can you supply the (won) games

grts Bram

[arunsoorya1309]

No tablebases may be makes a difference ?

[Modern Times]

Not surprised by that result. The new Houdini is very strong.

[arunsoorya1309]

i am not sure if its so strong ... in graham tournament i saw it lost to K and also in TCEC its getting a beating

[Modern Times]

The TCEC result is irrelevant to me, I don't have 44 core hardware. I'm more interested in how it does on more normal hardware.

[arunsoorya1309]

fair enough

[Lyudmil Tsvetkov]

So SF scales betters than H5. ... H5 in TCEC is getting a pounding

That is not the reason for the pounding, but simply that the TCEC opening positions played so far hugely favour SF: many open games, many Sicilians, many positions requiring deeper search with concrete threats after a long book, etc.

Once the book changes, the result will change too, but I am not certain the book will change. Clemens' result is very close to the truth: H5 is almost on par with SF, has better eval for most positions, however, as noticed, some search bugs and lesser depth are a major downside.

[Hugo]

Thx Clemens,
Surprisingly good result for H5 if I may say so
btw can you supply the (won) games

grts Bram

here are the full 100 games
http://clemens-keck.de/ReplayZone/H5-S8-long.pgn

best regards
Clemens Keck