A series of mutual blunders

Aug 22, 2009, 1:02 AM |

Since my prior brief obsession with chess in the mid-1980s, at least two important things have changed. First, computers (even inexpensive ones) are really, really good at chess. Second, my ability to calculate has decreased, and not merely for lack of practice. I'm trying to use both of these changes to improve my play.

Back in the 1980s I bought a Fidelity Excellence chess computer. I remember that at first I had a hard time beating it on level 2, but after a month or so of playing several games per day--and training myself to slow down as I consider my moves--I was able to beat it more than half the time on level 3. That was a clear sign of progress. I didn't know why I was progressing. I couldn't articulate what I was learning. But my unconscious was doing its usual marvelous job of making sense of lots of data.

Around that time, I played a few games at the Portland, ME chess club. In all but one of those games, I was trounced soundly. At some point in each game, I would notice that my position was hopeless. What was harder to see, even with my opponent's help during post-game analysis, was just where my play had gone wrong. Given how badly I was trounced, it was clear that I was unable to assess the positions well myself. But my opponents were usually unable to articulate, even to their own satisfaction, how I had failed. The few who did offer an explanation described my problem in terms of tactics. I wasn't able to abstract those specifics into something more general that I could apply in other games.

I wasn't learning from my failures, and I didn't know how to fix that. My interest faded, and 25 years passed.

A month ago, when I joined chess.com, I visited local book stores looking for help. Right away I noticed that I could not follow books that were mostly annotations. I'm not able (for now) to visualize more than a move or two beyond the latest diagram.

Realizing that, I selected books that focused relatively more on principles and relatively less on deep analysis. Jeremy Silman's The Amateur's Mind and Silman's Complete Endgame Course appealed to me immediately--lots of text, lots of diagrams, and analysis that I'm able to follow reasonably well even without a chessboard. I also bought John Nunn's Understanding Chess Move By Move, John Watson's Mastering the Chess Openings Volume 1 (1.e4 openings) and Volume 2 (1.d4 openings), and Efstratios Grivas's Practical Endgame Play. Though these books are significantly more analytical than Silman's, each includes description that is both clear enough for me to understand (most of the time), and frequent enough that I can follow both the games and the text.

I'm skimming these books for the general principles that I can see most easily. I know that by skimming I will learn only the surface of what these books offer, but what I'm seeking right now, and all I'm able to absorb right now, are general principles that I can apply immediately to the games I'm playing. I can study more deeply later, once I've learned from applying the surface stuff.

I've now completed 20 games on chess.com. I've submitted each game to chess.com for analysis, and I've also used a number of other chess engines to analyze the games. From this analysis I've learned one thing clearly: Each of my games has been a series of mutual blunders. When one player blunders, the other generally returns the favor, often immediately, and sometimes with a bonus. "Oh, you think your move was a clam? That was nothing. Watch this!"

In case you think I'm exaggerating, here is a graph of chess.com's analysis of one of my games: A series of blunders

The x axis shows the move number. The y axis is the score assigned by chess.com's computer analysis. Each unit is roughly equivalent to one pawn. A positive score indicates an advantage for white (which was me, in this case), a negative score for black. In several cases my opponent made a huge blunder, and I blundered in retort. On move 36, I generously overcompensated for my opponent's blunder, then through move 39 graciously proceeded to yield the rest of my advantage. Things turned in my direction for the rest of the game, but Good Gumby, these reversals of fortune are bad for my aging heart.

Almost all of my games look more or less like that, whether I win or lose.

For each blunder, the chess engines recommend moves that would, I presume, have been better than mine. Every now and then, I can immediately see why their move was better than mine, such as when five different engines unanimously pointed out the checkmate I missed. Yeah, even I can see that mating my opponent was better than what I did.

But most of the time, especially when the blunder is smaller than the big spikes you see in the graph, I can't immediately see what's better about the chess engine's recommendation. Alas, the engines are usually even less able than my human opponents to tell me just what was so bad about my move. A chess engine's "explanation" consists of a sequence of 17 subsequent moves, which (as I've mentioned a dozen or so times in this post) makes my eyes glaze over. What's a patzer to do?

I noticed something. Though different chess engines often disagree about what move would have been best, they all agree that my move was gawdoffal. And even when I can't make head nor tail of their reasoning, there is something I can do: Check the position to see whether I'm following the principles I'm learning from the books.

Here's an example, a position from a game from the King's Indian Tournament (the comments are from chess.com's computer analysis):

The chess.com computer considered my next move, 15.Ng5, to be an innacuracy. Not a blunder, but not the best move. It recommends instead 14.c5. Why? This is one I think I understand. In The Amateur's Mind, Silman says to identify the imbalances in the position, then exploit and expand the imbalances in your favor. I this position, I owned more space queenside. The computer's recommended line (14.c5 and subsequent moves) expand white's queenside space even further, just as Silman recommends. My move (and subsequent moves) launched a kingside attack, which I was unable to sustain. When the attack inevitably collapsed, all of my opponent's pieces were aimed directly at my king. Doom.

Even in other cases, where the computer's assessment is not immediately clear to me, I'm finding that if I look at each move that the computer considered to be a blunder, I can usually find something significant that I overlooked in the position. And usually I find that I violated Silman's recommendation to exploit and favorable imbalances.

So unlike my mid-1980s opponents, computers can point out exactly where I went wrong. And unlike those opponents, Silman and others can articulate, with simple, clear explanations that I can understand, the kinds of principles that my blunders likely violated. The combination seems to be working well for me. When I can match up the computer's highly contextualized, specific advice ("You goofed right here") with the chess authors' more abstract and general advice, I can learn better from my mistakes.