OpenChess

Anybody know what these are ??

Are there Houdini derivatives already ??

_op seems to be the stronger of the 2 , maybe ??

I'm testing _op against regular Houdini 1.5 now ......... the evals look wild , but , bottom line is after only 22 games _op is + 6 ..........

Houdini is not open source (though it should be), so I doubt they are 'derivatives'.

Where did you get them?

Peter

From what I gathered, it's not a derivative, but rather "correction" by (Jury Osipov?) of what is believed to be Houdini's eval obfuscation in attempt to hide its RobboLito origins.

JackStraw wrote:Anybody know what these are ??

Are there Houdini derivatives already ??

_op seems to be the stronger of the 2 , maybe ??

I'm testing _op against regular Houdini 1.5 now ......... the evals look wild , but , bottom line is after only 22 games _op is + 6 ..........

Looks like someone patched it. Ran it in a debugger, found the offsets where eval was occurring, and changed the constants being added/multiplied against the 'internal' values. This is just normal hacker stuff, like if you want to get around copy.protection. What it all means, assuming it's legit, is another story. However, from Houdini 1.0 to Houdini 1.03, the eval stopped matching Ippolit's so exactly. Probably due to this sort of interface obfuscation. Again, assuming that this is legit, which I can't easily verify. One could compare the three binaries in a hex editor and find out what changed where, though, and work backward from there with gdb...

Jb

JackStraw wrote:Anybody know what these are ??

Are there Houdini derivatives already ??

_op seems to be the stronger of the 2 , maybe ??

I'm testing _op against regular Houdini 1.5 now ......... the evals look wild , but , bottom line is after only 22 games _op is + 6 ..........

These versions have been obtained by a binary hack of the offical Houdini that modifies the output.
Be aware that the evaluation shown by these hacked Houdini executables will no longer take into account certain elements relating to endgame tablebases, contempt and the stage of the game.
Use these versions at your own risk/peril.

Robert

Robert Houdart wrote: These versions have been obtained by a binary hack of the offical Houdini that modifies the output.
Be aware that the evaluation shown by these hacked Houdini executables will no longer take into account certain elements relating to endgame tablebases, contempt and the stage of the game.
Use these versions at your own risk/peril.

Robert

Excellent "explanation", Robert!
You keep good defense.

Can you explain it better comrade Osipov ? What kind of versions are these, and how strong are they? Stronger than the original Houdini ?

From what I gathered, it's not a derivative, but rather "correction" by (Jury Osipov?) of what is believed to be Houdini's eval obfuscation in attempt to hide its RobboLito origins.

Not sure what to make of these patched versions:

"Real eval"

"Real eval" times 2

Which means in English ?

kingliveson wrote:

From what I gathered, it's not a derivative, but rather "correction" by (Jury Osipov?) of what is believed to be Houdini's eval obfuscation in attempt to hide its RobboLito origins.

Not sure what to make of these patched versions:

Would be interesting to compare the output to that of Houdini 1.0, as well.