Looking at the crosstable the first thing I noticed was that out of 15 players, three withdrew and did not finish the tournament. Not counting the games of those three players and the two games remaining (which have no bearing on the standings), 142 games were played. Total decisive games…11, or less than 8 percent!! Loinjack accounted for 4 of those wins and the games among the top five finishers were all drawn.
Engines are, of course responsible for this situation. I was reading a message board the other day and one guy was asking about the best computer and software to play chess…his budget? $12,000.
If you want to play correspondence chess these days your choices are pretty much limited to: 1) anonymous opponents with meaningless ratings who are using an engine or 2) opponents whose names you do know, have meaningless ratings and who are using an engine. I prefer the latter and I don’t see why I should pay to enter a tournament under these circumstances, so I use a free site that allows engine use…LSS.
In any case it seems that chess engines are not optimized for correspondence play. Of course, if you expect to get a high rating in modern CC play, you have to start at a high rating because it is extremely difficult to defeat another engine unless you are keenly aware of how engines work.
My understanding is engines prune a lot of moves out of the search tree and sometimes they are good moves but that only would become apparent at deeper search depths.
Under normal circumstances as long as a move does not lose it’s good enough, but for these really highly rated CC players that’s not enough. One problem is that often times an engine move shows an evaluation score and because it has pruned some moves, it doesn’t make any difference whether it searches 2 hours or 2 days, it’s not going to change its evaluation.
In CC these is no such thing as two moves with an equal evaluation score; one move simply must be preferred over another. The result is that if you just let an engine run and run then play its recommendation, guys like Loinjack will beat you. The thing is they somehow manage to get positions engines don’t understand and they have the resources and the patience to keep searching.
One high level CC player noted that you must use tablebases and gave one example where his opponent’s mistake was using the wrong engine for the relevant position. He then went on to explain how, using three engines, he basically strung together a composite of their moves. He added that, ideally, you would want to examine at least ten different alternatives and each one needs to be evaluated thoroughly.
Capablanca complained about the draw death of chess back in his day and proposed a variant in the 1920's. He believed chess would be exhausted in the near future, that games between masters would always end in draws. Fischer made the same claim. He believed the chess openings have been analyzed to the point that games are decided by opening preparation alone and that engines and databases also contributed to the death of chess. Indeed opening research wins games, both in OTB GM play and top level CC play.
Look at the following table:
first 15 world CC championships
white wins 37%,
black wins 24%
draws = 39%
1996-2007 championships
white wins 27%,
black won 12%
draws = 61%
2008-present championships
average percentage of draws exceeds 80 percent.
Upsets. In the first 15 world CC championships, upsets happened in about 10 percent of the games. Between 1996 and 2007 the percentage was down to 4% and since then it’s 1%.
Playing on LSS even at my level shows how difficult it is to win: Wins: 23.7% Losses: 23.0% Draws: 53.3% Compare these to a Senior IM I played a while back: Wins: 20.5% Losses: 19.9% Draws: 59.6%
Now, compare these to Loinjak: Wins: 51.1% Losses: 00.0% Draws: 48.9% He clearly knows something about using chess engines I (and a lot of others) don’t!
I am sure I would beat both ICCF and LSS "champions" decisively in real CC (engine-free) matches. They aren't that good without their computers.
ReplyDeleteWithout engines, I do not doubt it. I looked up FIDE ratings for the players and most did not have one; I did not bother to check for national ratings. The FIDE ratings I found ranged from 1850 to just over 2200…hardly ‘world championship’ strength. When I returned to CC after a 12 year absence it was at IECG (now LSS), was not aware they allowed engines and scored +0 -4 =2 in my first tournament…was not sure if my play had deteriorated or maybe 2100 European players were stronger than their 2100 rated American counterparts…did not realize until after the tournament that I had been playing Fritz5..the joke was on me!
ReplyDeleteMaybe the best correspondence chess players still 'clearly know something about using chess knowledge a lot of others don't ?!
ReplyDeleteHave you checked the games? Are the critical maneuvers and tactics really that easy to spot by strong chess engines?
No engine can have more chess knowledge than the maker. If you know more about chess than the programmers of the best engines then you can beat all opponents who use these engines.
Combine your own chess strategic skills (if you have any) with the merciless calculation precision of the engines to produce strong play. This is what the modern correspondence chess is about.