Stockfish 3 PA_GTB

[BorgKing]

Looking forward to testing this, once there's a version for my oldish computer.

[ernest]

Looking forward to testing this, once there's a version for my oldish computer.

You can test the 32-bit version...

[Jeremy Bernstein]

This version becomes easily decoupled from the GUI.
...
The engine was still running in memory, consuming 98% of 4 CPUs, but it was not writing to the output terminal.
I usually make my own version, with at least unit buffering.
Standard buffered I/O does not work well in Windows via either Winboard or UCI apis.

Thanks. Older versions of Stockfish had some code for unbuffering the output (suggested by BB+ on this forum and adopted into the official source at some point). Looks like that's been removed in version 3. I'll replace that code for the next release.

jb

[Jeremy Bernstein]

Here's the 2nd release of the PA_GTB_Gran2 build. The only PA_GTB-related change is the unbuffering of stdout (leaving stdin as-is), which should solve the reported issue with engine output. Otherwise, this version incorporates the latest changes to the official Stockfish master branch. Windows versions for non-SSE/non-POPCNT hardware are included this time.

Enjoy, please report back with any questions or comments.

Jeremy

[Jeremy Bernstein]

@Jeremy Bernstein
Can we have a 3-fold rep=draw in Stockfish 3 PA_GTB?

While searching SF3 does detect 3-fold repetition and provides a draw score, which is used to score/sort PVs. That score isn't used directly at the root, but given the following line:

1.d4 d5 2.Nf3 Nf6 3.Ng1 Ng8 4.Nf3 {0.00: Nf6} Nf6 {+0.42: c4} 5.Ng1 {-0.55: c5} Ng8 {+0.31: c4} 6.Nf3 {0.00: Nf6}

You can see that something is going on after White's 4th move (1st repetition of the position). So Black is offered an eval of 0.00 if he plays Nf6. But surprise, after playing Nf6, the eval jumps to +0.42 if White plays c4 (strangely enough, this position is the 1st repetition of the position after 2...Nf6, so it should also be marked 0.00). Back and forth with real scores until 6.Nf3, when the draw recognition kicks in again.

So something isn't really working 100% there. I honestly don't know what the correct solution here is. Repetition detection doesn't come for free (this thread on approx the same topic might be of interest: https://groups.google.com/forum/?fromgr ... ZdxhyDZ9kJ), so that probably explains the half-hearted implementation here.

Instinctively, I think that a draw score only makes sense when the side to move has the option to make a 2nd repetition (and immediately end the game as a draw), and only if that's the best move, all other factors being equal. That's not how most engines do it, though, and I'm not well-versed in engine development to know why. Maybe Marco or Bob or someone equally knowledgable will chime in and explain it to us.

Jeremy

[User923005]

Here's the 2nd release of the PA_GTB_Gran2 build. The only PA_GTB-related change is the unbuffering of stdout (leaving stdin as-is), which should solve the reported issue with engine output. Otherwise, this version incorporates the latest changes to the official Stockfish master branch. Windows versions for non-SSE/non-POPCNT hardware are included this time.

Enjoy, please report back with any questions or comments.

Jeremy

I will give it a spin.

[User923005]

Seems fine so far (no disconnects in an hour). The behavior is somewhat random, so it will take a while to make sure it is fully stable.

[BorgKing]

I tried it in a test tournament during the night, on a probably losing position from one of my corr games. It won the first game and then stopped working after about 20 moves on the worse side, -4.27. Didn't get any error code, the system box just reported that it had stopped working.

On the three move repetition, I've noticed that most engines seem predisposed to repeating the position, for instance I had another test game a week ago where StFish and Houdini both had the same expected line of 1.Rg5+ Kg7-h7 2.Rh5+ Kg7 3.Ra5 +2.03 instead of simply 1.Ra5 - if you're better it makes sense to not repeat unless it's forced, since then the opponent wouldn't be getting nearer to get more time. But maybe the engine authors think it's better to get more time yourself?

[endings]

both versions you made cause an exception in fritz when accessing tbs. the sse4.2 completed one game but did not need the tablebases. when I get stockfish from ebrok.eu/stockfish it is the middle compile that works on my computer using fritz.

the sse4.2 version form ebrok.eu/stockfish also causes an exception in fritz on my computer..

can you tell me how to compile for modern computers but without sse4.2? I would like to try and compile it myself. I down loaded mingw for 64 and 32 bit but I am new to compiling. I am going to try to compile on windows but if I can't I will install linux. I use to have a Debian machine and compile all the time so I can do it. My problem will be what flags to use or not to use while compiling stockfish for modern computers without sse4.2 ....

Thanks Bob

[Jeremy Bernstein]

both versions you made cause an exception in fritz when accessing tbs. the sse4.2 completed one game but did not need the tablebases. when I get stockfish from ebrok.eu/stockfish it is the middle compile that works on my computer using fritz.

the sse4.2 version form ebrok.eu/stockfish also causes an exception in fritz on my computer..

I don't think it's the SSE support, but rather the new Gaviota TB support in recent versions of Fritz and Chessbase, added to support Houdini. I'll need to do a little probing to see what's going on, and will have it fixed for version 003.

Thanks,
Jeremy