The latest Stockfish-12 chess program is incredibly strong.
I installed the latest Stockfish-12 chess program on my desktop PC (which has an Intel Core-i7 4700 CPU) and played it against a couple of other chess programs, and it simply demolished them. Now those ‘other’ chess programs typically quickly destroy me when I play them, so to watch this latest Stockfish-12 play is a pleasure to watch.
I believe that Stockfish-12 is currently (as of November-2020) the strongest chess program, and it currently uses neural networks (NNUE). Some people claim it may be better at playing chess than the Alpha Zero, which defeated an earlier (much weaker) version of Stockfish a few years back - although that is speculation.
My desktop PC running the Stockfish-12 NNEU program is currently running the openSUSE LEAP-15.1 GNU/Linux.
The Stockfish Blog, about Stockfish-12 is here (dated 2-September-2020).
As noted in that blog, Stockfish-12 NNUE is not yet formally released, but one can install the development version, which is stable and as noted, plays incredible chess.
As pointed out in the blog, one can download Stockfish from here: Stockfish Developement versions. One needs to pick the correct version of Stockfish-12 for one’s computer CPU. As noted, my computer is a Core-i7 4700, and I noted from this Intel page that the Core-i7-4770 is in a “Haswell” category (my terminology is not exact here). Further I noted from wiki on BMI (BIT manipulation instruction) that Haswell and new processors support ABM, BMI1 and BMI2.
Hence I knew that I needed to download the “Linux x64 for Haswell CPUs” version.
That came in a ZIP file (stockfish_12_linux_x64_bmi2.ZIP), and I downloaded the ZIP file, unzipped the binary ( " stockfish_20090216_x64_bmi2 " ) , and placed it in my PC in the directory:
Obviously, I could have chosen a different directory name - but the above worked for me.
Download Neural Net
I then when to the site Stockfish Testing Framework - Neural Net downloads and downloaded the Neural Net “nn-cb26f10b1fd9” (released 30-Oct-2020). I note there is a newer ‘stable version’ (highlighted in ‘green’): nn-c3ca321c51c9.nnue (released 7-Nov-2020) , which I may try some time in the future.
After downloanding the NNEU neural net (which is a .nnue file), one needs to place the .nnue file in the same directory as the engine binary, which I noted above. By default, the NNUE is off, but it can be enabled with the Use NNUE (true/false) UCI option. It must be enabled to get the full program chess playing strength.
Configuring XBoard GUI for Stockfish
In GNU/Linux, one needs a GUI front end for the program (where I use ‘xboard’ which is packaged for openSUSE).
For me, configuring Xboard was the ‘trickiest’ part. I launched the program Xboard and went to “Engine” > “Load First New Engine”
which brought this menu and and I edited it:
From the above you can see I edited this as follows:
- Arbitrary assigned it the name “Stockfish12” (I could have used ANY name here)
- In Engine Directory, I entered " /home/oldcpu/chessgame "
- In Engine Command, I entered " /home/oldcpu/chessgame/stockfish_20090216_x64_bmi2 "
- I selected “UCI”
- … see above for remainder of my selections (they may not be optimal)
I then clicked OK.
Next I went to “Engine” > “Edit Engine List” and I saw the following:
I wanted the Neural Network active, so I then edited that line by adding
which gave me this:
and selected ‘commit changes’ followed by OK.
**… continued …