Random Posts

Wednesday, April 8, 2015

My Proposed New Title for Correspondence Players

    My April Chess Life just arrived and there was a long article on the Thoresen Chess Engines Competition (TCEC) that I found quite interesting. For those not familiar with the TCEC competition it is an engine tournament held twice a year between the best engines with the games played on high end computers. For those that understand such things, two 8-core Intel Xeon E5-2689 processors which, according to the article, will set you back over $1,700 are used. So, my quad core Toshiba Satellite (around $500) is pretty useless when it comes to serious analysis such as is required by professional players and serious correspondence players at the GM level. But, I already knew that.  For more information on TCEC please read Christopher Chabris' interview with Martin Thoresen HERE.
     The author of the Chess Life article, IM Erik Kislik, makes the statement that at 3329 Stockfish 6 probably really does deserve a rating that high even though it has never faced any human GM-level opponents in a real tournament. His claim is that Magnus Carlsen playing a 40 minute game with a 40 second increment would score about 1 point in 10 games. There are other factors involved like opening choice that could alter the results a tad, but the author points out Carlsen has said he gets no joy out of playing engines and Kislek hints at the fact it's because Carlsen has probably lost a lot against them.  Kislek adds that there are still a few situations where humans can offer some helpful input, but those instances must be rare.
     Kislek gave some recent examples of games played in the TCEC between Komodo 8 and Stockfish 6 (Komodo 8 won by 3 games) and just out of curiosity, I checked out one of the more interesting positions on my laptop just to see what the result might be.
     In this position which Kislek describes as “incredibly complicated,” because “it is far from obvious how to deal with white's Pawn advantage in the center” he says both engines got it wrong because human intuition might have given a clear cut solution.

     So, what did my puny laptop come up with? First, the actual game continuation: 1.e4 Re8!? Kislek says the plan is ...dxe4 and ...Nbd7 so white has to try and consolidate his center by pushing the e-Pawn. 2.e5 Kislek says this is the obvious try because it is necessary for white to get the biggest center he can. 2...Nfd7 3.d4 Na6 (?!) Kislek says this move is "pretty inexlpicable since the obvious human attempt at striking back at the center with 3...f6 has no flaw."
     Now, from the starting position I let both engines run on three cores and the results were:

Stockfish 6 Main Line after 10 minutes: 1.e4 Re8 2.exd5 cxd5 3.h3 Be6 4.Nd4 Nc6 5.Nxe6 fxe6 6.Bf4 Evaluation 0.08

Komodo 8 Main Line after 10 minutes: 1.e4 Nbd7 2.exd5 Nxd5 3.Nxd5 cxd5 4.h3 Bh5 5.Re1 Rc8 6.Be3 Evaluation: 0.21

     Clearly, on my laptop the two best engines in the world don't come up with the same moves as the same engines do on dedicated computers costing three times as much, so I have about as good a chance competing at the highest levels on places like ICCF and Lechenicher SchachServer as I do playing over the board against Carlsen, Anand and others of that ilk.
     As one poster of advice on playing engine assisted chess points out, unless you just want to dabble in it or don't care about results or ratings, it's not for you. Maybe they ought to create some new correspondence chess classes: Beginnng Dabbler, Intermediate Dabbler and Serious Dabbler. Above us Dabblers are the titled players...those guys with a couple of thousand dollars to spend on hardware/software and a LOT of patience.

No comments:

Post a Comment