Engine Experiment

    Labor Day was officially established as a national holiday in the United States on June 28, 1894, when President Grover Cleveland signed a law passed by Congress making the first Monday in September a legal holiday to honor American workers. The first Labor Day celebration, however, had already taken place in New York City on September 5, 1882, organized by the Central Labor Union. 

    Now that the long weekend (for those who are still working) is over) and things are back to normal I have been monkeying around analyzing a game Tal lost to Filip in 1962 that has turned out to be more complicated than expected, playing a bit online where I had one interesting game that I might publish and I have been looking at an engine named Alexandria. 

    While visiting the CCLR engine rating list I noticed the open source neural network engine Alexandria 8.0.0 64-bit 4CPU was ranked 7th only 28 points below Stockfish 17.1. 

     I was unable to locate a site where version 8 was available and while searching for it visited Chess Engine Diary on Facebook and found the link to download version 8. Clicking on it took me to a porn site! Quickly backing out and running my antivirus program turned up no threats. 

    What I did find was a safe download of an older (about two years) version, Alexandria 5.1.4 on an interesting site called CHESSENGERIA. On my laptop Alexandria version 5 was defeated in 4 minute blitz games by Stockfish by a score of +0 -4 =1, but I found something interesting in the process. The evaluation of Alexandria notes that it “does not wait for...unfavorable developments”, but is willing to take risks in an attempt to weasel out of an unfavorable situation. 

    As a test, I used a position from the previously mentioned Tal-Filip game because that’s exactly what Filip did. He was in a bad situation and took a risk and it worked. Specifically, I wanted to compare the two engines’ analysis of the position after Tal’s 24.gxf5 to see how they compared. 

 

    Both engines give white a winning advantage of about 3.5 Pawns after 24...Re7. However, I was excited to see that Alexandria initially looked at Filip’s sacrifice of the exchange (24...Rxe5) which it finally evaluated at about 6 Pawns in white’s favor compared to Stockfish’s 5 Pawns. As a result, the risk was too great and so Alexandria finally settled on 24...Re7 as preferred by Stockfish. 

    Obviously, Alexandria is weaker than Stockfish and so there is not much point in using it for analysis purposes after the fact. What it might be useful for is uncovering an opening surprise when preparing for over the board play.