Saturday, April 19, 2014

An Opening Test for Engines

     The website Chess Openings offered this challenge: fire up the engines and see when, or if, they could find the brilliant 11.Rc1 as was discovered by former ICCF World Champion Aleksandr Dronov who played it against Jeurgen Buecker in the 2007 championship.

 
     I gave each engine about 5 minutes using 3 cores and 1024 Mb of RAM and all they all displayed 11.Nef3 as their first choice and none displayed Dronov’s 11.Rc1 in their top four selections.  Here are their evaluations and ply depth:
StockfishDD64SSE4.2: +0.30, depth 26
Deep Rybka 4.32: +0.30, depth 18
Houdini 2.0c 64c: +0.16, depth 22
Critter 1.64 64bit: +0.23, depth 21
     When I entered the move manually the engines displayed the following evaluations: StockfishDD64SSE4.2: +0.05
Deep Rybka 4.32: +0.15
Houdini 2.0c 64c: 0.00
Critter 1.64 64bit: 0.00
     I decided to try using IDeA in Aquarium 2014 using Stockfish set at 20 seconds, depth 17 to see what results I would get. After about 2.5 hours here are the results:

 
     As you can see, the engine has not gotten very deep into the position, but 11.Rc1 is pretty far down the list. So why is this important? Because these days correspondence chess as played on ICCF or LSS where engine use is allowed (you’ll probably lose all your games if you don’t use one) analysis has to be very deep and detailed, especially openings. Openings represent about 50% of your results in these events so excellent opening preparation is necessary.
     I prefer to play on LSS in the “Rapid” events because 1) I enjoy tinkering with engines and 2) I don’t have the patience to fiddle with the kind of opening preparation and analysis that is required to win consistently as explained HERE.
     The point is that to play successfully on sites like LSS and ICCF there is a whole lot more work involved than just letting an engine run for a minute or two then playing whatever its first choice is. Dronov is only rated 2043 with FIDE and I would be curious to know how such a low-rated player discovered 11.Rc1 even using engines. Perhaps it could be done if you had a really good opening book, a couple of powerful computers and more patience than I have accumulated in a lifetime, but I prefer to stick to LSS Rapids and hope my farrago leads to success.

3 comments:

  1. Maybe you should use a computer with 4 or more cores and 16 or 32 GB memory so you can analyze to 40-50 ply depth like these ICCF players are doing. 30 or less ply is "yesterday's news".

    ReplyDelete
  2. Correct! Top level CC these days requires, in addition to a lot of patience and meticulous opening preparation, that you invest in a computer of mass destruction. My piddling little laptop with 4 cores 4.00 GB of RAM lacks the punch to be competitive. Best Buy has what I need for over $1000.

    ReplyDelete
  3. Re 'Dronov is only rated 2043 with FIDE', the FIDE rating card says, 'B-Year: 1980'. I think it's a different Dronov. Wikipedia says the three time World Correspondence Champion was born in 1946.

    ReplyDelete