Tuesday, August 29, 2006

Mystery chess position V: messing a perfectly good endgame up

The position shown on the right with white (me) to move is from on eof my recent games at ChessWorld. White has the material advantage and a promising passed pawn on the g-file. The game started balanced until black made a tactical mistake on the 35th move and I was able to keep the advantage for 20 moves until I finally managed to come up with a series of really bad moves. It started with 55. g6 - passed pawns have to be pushed, right? As usual, rules do have exceptions. There is only one sensible move for white in the shown position. What is the correct move? (Hint: it is not 55. g6 ;-)

Lately, I have been playing mostly against players whose ratings are above mine (in this case, the difference is about 150 points). Therefore, I am loosing most of my games but I hope to improve faster by playing against stronger opponents. In the few games where I manage to obtain an advantage, I tend to make devastating mistakes in the endgame. It seems that there is some mental barrier making me look for "killer moves" instead of ensuring that all my pieces are safe (see also the interesting article "When You're Winning, It's a Whole Different Game" [pdf format] by Dan Heisman at chesscafe.com). Even after 55. g6, however, the game was not lost. Take a look at the full game with annotations for more hints on making bad moves in the endgame.

Friday, August 25, 2006

How important is repetition for acquiring pattern recognition skills?

A bit more than one month ago, I decided to start to do some regular tactics training on the Chess Tactics Server (CTS). Using CTS to improve chess skills is still a hot topic in the chess blogosphere and a small but growing group of chess bloggers, most notably Temposchlucker, Transformation, and General Kaia, put a significant effort into solving tactics problems on the Chess Tactics Server. One attractive feature of CTS is the huge amount of available problems. On one hand, this guarantees variety. On the other hand, it (almost) prevents that problems have to be solved repetitively. Repetition, however, is considered as very important for transferring the knowledge of tactical positions to long-term memory. This raises the question, how useful is solving chess tactics problems with a very low repetition rate at CTS to improve chess skills?

Several scientific studies tried to address the difference between grandmasters and the common patzer. The huge skill differences between grandmasters and even good amateur players can be observed in simultaneous exhibitions. One grandmaster plays seemingly effortless against dozens of amateur chess players and wins most, if not all, of the games even though the grandmaster's thinking time is sometimes only few seconds per move! This immense difference in ability is attributed to the grandmaster pulling information about the positions out of his long-term memory, which is a very fast process, while amateurs have to analyze the position. The efficient use of long-term memory for a complex game like chess requires storage of a substantial amount of positions, or so-called "chunks", in the long-term memory. Many common techniques to memorize information base on spaced repetition: the information is memorized repeatedly until it is finally safely stored in long-term memory.

This method can be also applied to studying chess tactics. A set of tactics problems is solved over and over again until solving them becomes automatic - the solution is recognized on first sight without the need for analyzing the position. There is only one problem with chess: the number of possible chess positions is astronomical. Therefore, the brain has to rely on recognizing the most important features of the position instead of the position of each individual piece on the board. For instance, there are well-known tactical themes such as forks, skewers, etc. An individual position with a knight forking king and rook may be unique and may never repeat in any chessgame. The general characteristics of the position, for example that the forked pieces have to be on squares with the same color, are always the same. Therefore, it is not necessary to repeat individual positions to achieve the skill to play a knight fork. Playing through a large number of different problems where knight forks occur will also do the trick. But which approach is more efficient to develop chess tactics skills?

She short answer is: I don't know. People became chess grandmasters before the availability of databases or even books with large collections of tactical problems. Therefore, repetition of individual problems is definitely not necessary to acquire a high level of tactical skills. Individual problems are only rarely repeated during CTS sessions. Tactical pattern, such as the knight fork, however, are repeated. I speculate that the very low rate of repetition of individual positions together with repeated occurrence of specific patterns such as forks might actually be very effective for learning tactics because the brain is forced to learn recognizing the essential pattern instead of specific complete positions. Proving this, however, would require a long-term study with many players using a very controlled training regime, something that certainly collides with the pronounced individualism of chess players. For now, I can only say that I am making some progress with training on the Chess Tactics Server and I even see (or want to see) some positive influence of that in my games. But more about that in a later post.

Sunday, August 13, 2006

Studying chess endgames: how important are pawnless positions?

I finally started to study chess endgames and decided that it might be best to actually start at the beginning of the book (I am currently using Bruce Pandolfini's "Pandolfini's Endgame Course"). However, every endgame book I looked into starts with studies of elemental checkmates and other positions without pawns. However, positions without pawns rarely occur in chess games. This makes me wonder if it actually makes sense to spend a lot of time analyzing these positions.

First of all, one might say that pawnless positions are much less complex than positions with pawns. Therefore, it cannot possibly take that much time to go through these introduction(?) chapters. To my surprise, it turned out that even in clearly won positions it can require accurate play to actually execute the win. As Blue Devil Knight also reported recently, examples such as K+B+N vs. K are definitely won but can be tricky to execute even for good players. Therefore, it takes a lot of time to develop an understanding of even the most basic positions, the elementary mates.

This week I decided to "graduate" from the elementary mates and to move on to the battle between heavy pieces. In general, the queen (white) wins against a single rook (black). However, black has drawing chances if she/he manages to keep the rook next to the king. Among the chess positions analyzed by and named after the chess master Francois-Andres Danican Philidor (1726-1795) is also the queen vs. rook endgame position shown on the left. This is a zugzwang position. With black to move and forced to move her/his rook away from the king, white wins quickly. Regardless which move black chooses, white can either checkmate or capture the black rook with a queen fork. For instance, the only legal move for the black king, 1. ... Kh6 allows white to pin and capture the rook with 2. Qf8. Moving the rook by playing, e.g., 1. ... Ra7 enables white to capture it by 2. Qe4+ Kh8 3. Qh1+ Kg8 4. Qg1+ forking black's king and rook. However, it is white's move. White's goal is to get in the same position as shown above, but with black to move. This can be achieved by playing 1. Qe4 Ka8 2. Qa8 Kh7 and after 3. Qe8 the pieces are back on their original positions with black in zugzwang. White wins as described above.

As mentioned above, it took me quite a while to comprehend Philidor's position. One of the reasons for my troubles studying these seemingly simple chess positions is that I am still not able to analyze variations longer than, say, 2 moves. In chess endgame positions, the number of possible moves is typically much lower than in typical middlegame positions. However, it is often necessary to carefully analyze much longer variations to prevent the opponent from escaping or even forcing a draw. As a consequence, I came to the conclusion that studying very basic endgames will probably help me to improve my general ability to analyze positions.

Monday, August 07, 2006

From the news: play chess against the ocean's tides

Wednesday, an unusual blitz chess tournament will happen on the North Sea's ground before the German island Baltrum. The tournament will start at between high and low tide when the water is running off and the rising water of the upcoming tide will add to the pressure on the participants of what is described as the "only lightning chess tournament on the bottom of the sea". See here for details and a photo of last year's event (in German, English translation here).

Wednesday, August 02, 2006

Just in time endgame study part II - nearly there!

As discussed in this previous post, it should be possible to study endgames as they are coming up during correspondence chess games. If "just in time" deliveries help car manufacturers become more efficient, why not doing the same in chess? The slow pace and explicit encouragement of the use of chess books enables just in time studies during correspondence chess games. A week or so ago, I browsed through a used book store and found (surprise!) a small shelf filled with chess books. Not that I need another one - in fact, I promised myself not to buy more chess books before I work through the ones I already own, but well... After extensive browsing I finally grabbed a nearly new copy of Reuben Fine's "Basic Chess Endings". Good decision, because apparently, Fine's work is (was?) the chess endgame "bible", for more about the book see Anthony Saidy's insightful review at Jeremy Silman's web site (By the way, I learnt from the review that even grandmasters consult books during adjourned games. And I always thought grandmasters have everything in their memory...).

The diagram on the left shows the position from one of my recent games on ChessWorld after 45. ... Rh4 with me playing white to move. I figured with as few as two rooks and three pawns left on the board, it might be worth to take a look in "Basic Chess Endings". I have to admit that I was quite surprised to find out how complex rook and pawn endgames are (at least to a beginner like me who never seriously thought about them). Therefore, I decided to simplify it further by playing 46. b5. I was sure that black's response would be ...Rh5 to skewer my K and P and after playing 47. Ke4 Rxb5 48. Rxh7 only one pawn would be left over reducing the analysis in "Basic Chess Endings" to just slightly more than 20 pages of densely written variations. Of course, there are other possible replies for black such as 46. ... h6 but during the post-mortem computer analysis it seemed to me that all variations sooner or later lead to captures leaving only one black pawn on the board. White is then able to obtain a draw if he/she manages to maneuver the king in front of the black pawn. The full game with annotations can be found here.

The most surprising thing that I learned from this just-in-time study was how complex even these very simple endgames can become. On the other hand, I had fun analyzing the position, something I never really imagined since I started to do some basic endgame study. My previous lack of motivation to study endgames is probably due to the fact that I started and got stuck with elementary mates, which seem just much too far from reality to be interesting. This game, however, got me interested in learning more about endgames. Unfortunately, there is still one big limiter: time.