Chessbase has a recent interview with Leonardo Ljubicic who is the 28th World Correspondence Champion in which he talks about his way to the title and reveals how he prepares for his games. He talks about his openings, what is important to play successful correspondence chess, explains how humans can use engines to play better than engines alone and he gives some insight about what humans can do that engines can't.
Ljubicic readily admits that it is impossible to achieve any great measure of success in correspondence play without engines and databases. That said, he claims that humans are still important in two areas: the choice of openings and steering the engine toward (or away) from certain types of positions. He observes that if you want to be successful in top correspondence chess you can only play certain openings because you simply cannot afford a single poor move. That's kind of unfortunate because it means tons of Najdorf Sicilians and Nimzo-Indians, etc. It's no use preparing sharp lines in questionable openings or variations...and that sounds like a never ending string of the same old openings with improvements coming ever deeper into the game.
He also points out that one needs a good general chess knowledge (things like P-structures, good and bad Bs, etc.). In other words, it helps to be at least a master. The idea is, as he put it, if you have enough time and patience you can give the engine more and better suggestions than your opponent. This alone eliminates most of us from ever amounting to a hill of beans in the correspondence chess world.
He used to use Rybka exclusively, but 4 or 5 years ago switched to Stockfish after testing many engines. Today those two are his main ones. He also said he thinks it's important not to switch engines too much because you have to understand and recognize the strengths and weaknesses of the engines you use. A deep understanding the differences is something most of us average players can't do either.
Ljubicic's climb to the world championship required patience and a lot of effort. The championship is played in cycles. Every year a new cycle starts and finals are played every two years. The standard time control is 50 days for 10 moves. You have to slog through three preliminaries (Preliminaries, Semifinals, Candidates) before you reach the finals an on the average, it takes four to six years to qualify for a world championship final. In the Preliminaries opponents are typically 2300+ in CC and in the semi-finals they are over 2400. Of course, if you make it to the finals, you opponents are likely to be rated well over 2500. So, it requires a lot of patience even assuming you are good enough to get that far. Most of the guys who have won the championship either with or without engines don't come back for more.
Ljubicic said it required all his free time and a serious investment in equipment. In his case he purchased an early version of a 4-core PC, acquired all sorts of opening books, databases, and the Nalimov five piece tablebases. He managed to win some CC tournaments and receive his GM title, but it took him five or six years of serious work. Reaching the finals required twice the time he usually took in a tournament.
When asked what he thought engines can't do he stated that they still misjudge positions. In his case, being an OTB rated 2200 player, he was in a position of having to evaluate the output of a much stronger engine. To that end, he emphasized how important knowing where an engine is strong, where it is weak, and which positions it plays well and which positions it does not like. Again, this does not sound like something a casual engine user can do.
He stated that you cannot let the engine do all the work and you have to “give it a position then leave the computer for a couple of hours, and when returning just check what the computer proposes. You will do much better if you watch the thinking process, to try to recognize and understand the strengths and weaknesses of the machine and to guide the analysis.”
Of course, anybody familiar with engines knows that very often two different engines suggest different moves and the score of the suggested moves can be very close. What do you do in those cases? Ljubicic suggests patience! Let the engine think some more and he suggested that because all top engines prune heavily, looking at the next best line is a good idea. Personally, I don't have the patience to let an engine run overnight examining two or three different moves.
As for software he uses the ChessBase GUI, several databases, particularly the ICCF database, MegaDatabase and the Playchess games database to get new ideas. He does NOT use human games because they are “too unreliable.” Using these databases, he forms opening trees, but one thing he does not do is rely on statistics. For example, after the moves 1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 a6 my old Fritz 12 opening book gives a table that looks like this:
You should not rely on the win/loss percentages in selecting which line to play. What he does is analyze all variations carefully and then decides on his move after checking and preparing his analysis. As he said, in CC you cannot rely on an "if he doesn’t see it strategy."
He also follows the latest in chess engine development and acknowledges that Komodo and Stockfish are the best, adding they have almost no weaknesses. Stockfish calculates variations fast, and excels in tactics and attacking, while Komodo is solid in style, and its positional play is second to none. They are very close in strength and are excellent choices for serious correspondence chess.
A lot of this interview may be an advertisement for Chessbase products, but I have no doubt that it IS the software he uses. He stated that the new highly hyped ChessBase features such as Cloud, LiveBook and Let’s Check are probably too “light” for top correspondence chess players, but when coupled with the Sampled Search feature in the new Fritz 15 they can offer insights and he considers the Sampled Search to be ”the biggest invention in computer chess ever.” To be honest, I'm not sure what a “sampled search” is or how you go about using it.
|This is how I play chess|