Stockfish 9 is OUT!

[Nordlandia]

Asmfish 9 is out or equivalent ->

https://github.com/lantonov/asmFish/tree/executables

[Ron Langeveld]

The nasty contempt 'bug' effectively makes Stockfish 9 completely useless for my purposes. Luckily there is asmfish. Will have a look if this port has more common sense.

Ron

[syzygy]

Asmfish 9 is out or equivalent ->

https://github.com/lantonov/asmFish/tree/executables

That is quick!

Unfortunately the base and popcnt Linux versions segfault immediately...

edit: there was no real update

[Ovyron]

(I think Rybka has this. And I thank Uri for this confirmation.)

Well, Rybka dropped Contempt altogether at some point. Nowadays, a "Score Offset" setting remains, for latest Rybka (Fritz 16), and though it can be used to avoid draws, it's not the same thing, and actually, Contempt and Score Offset could be used complementary if an engine implements both (in the case of Stockfish, you could use Score Offset to have it choose the same moves of Contempt 20, while still showing more sensible scores.)

For an actual Contempt for Rybka you have to actually use some settings like those from Mindbreaker, which work well (among a lot of other things... this is the advantage of allowing the user to modify each piece's value from each side, too bad Fritz's strength is so low as to make this irrelevant.)

Yes, even Komodo's or Houdini's (or Rybka's) solution would be better than the current implementation.

Houdini's implementation is not trivial, Stockfish would need to detect what side has the advantage, and apply the contempt to this side. It solves the problem of switching sides (a side has the advantage regardless of what point of view you're using), but introduces a new one where scores are erratic and might jump from huge white advantage to huge black advantage from one ply to the next.

I also don't know what Houdini does in 0.00 positions, apparently its detector manages to know what side is defending those positions, which is something very useful I haven't seen any other engine doing (all others will show 0.00s, even Stockfish with Contempt=40, so Houdini is the only one useful, that at least tells you what side is fighting for the draw), we don't want Stockfish to use Contempt=0 on those, because high Contempt is the most useful in the positions you want to look for an advantage, so again, programming a detector that knows what side should have a high Contempt in 0.00 positions isn't trivial.

[syzygy]

Houdini's implementation is not trivial, Stockfish would need to detect what side has the advantage, and apply the contempt to this side. It solves the problem of switching sides (a side has the advantage regardless of what point of view you're using), but introduces a new one where scores are erratic and might jump from huge white advantage to huge black advantage from one ply to the next.

According to the online manual, Houdini's "Analysis Contempt" simply forces contempt to be from white's pov:

Analysis Contempt (checkbox)

Activate Contempt for position analysis.

It is usually not recommended to activate the contempt for analyzing positions.
When contempt is active for analysis, the score of the analysis will be optimistic (over-evaluated) for White if the Contempt value is positive, and optimistic for Black if the Contempt value is negative.

http://www.cruxis.com/chess/manual/index.html

This makes sense.

The difference with Komodo's "White Contempt" is that the latter also forces contempt to be from white's pov in game play.

[syzygy]

The nasty contempt 'bug' effectively makes Stockfish 9 completely useless for my purposes. Luckily there is asmfish. Will have a look if this port has more common sense.

Ron

Until an updated asmFish is out, you might want to try Cfish:
https://github.com/syzygy1/Cfish/releases

Checking "Analysis Contempt" forces contempt to be from white's point of view.

If you compile the latest version, the "Analysis Contempt" checkbox becomes a combo box with options Off/White/Black.

[Nordlandia]

Until an updated asmFish is out, you might want to try Cfish:
https://github.com/syzygy1/Cfish/releases

One my machine dedicated specifically for chess uses an Q9650 CPU from Q3 2008. It offer no popcnt instructions.

Is it possible to provide Cfish_base.exe compiles`?

bmi2.exe | Haswell
popcnt.exe | modern
base.exe | Oldest

[syzygy]

Is it possible to provide Cfish_base.exe compiles`?

Please see if it works:
https://github.com/syzygy1/Cfish/releas ... indows.exe

[Nordlandia]

Is it possible to provide Cfish_base.exe compiles`?

Please see if it works:
https://github.com/syzygy1/Cfish/releas ... indows.exe

Working great and still faster than SF9

[Ovyron]

According to the online manual, Houdini's "Analysis Contempt" simply forces contempt to be from white's pov

According to "actually firing up the engine and using Analysis Contempt", this is not the case.

Reminds me of my experience in the MultiPV thread, many people kept claiming that Stockfish behaved in a way it didn't. Can't people go and actually check how the engine behaves instead of telling me that it behaves like it doesn't or that some manual says it behaves differently?