20 August 2011

Game Analysis for Improvement in Play

An earlier post described my general conception of how game analysis can be used as a method of improving play.  Essentially, the focus is put on one's own games and the lessons drawn from them can then be followed up on most effectively.

So far, while I've certainly benefited by re-examining the openings, tactics and strategies on display in each game, the most significant impact for me, interestingly, has been on the thinking side.  I am being repeatedly exposed to alternative moves that I either could not find or did not originally evaluate properly.  Since I know my own thinking process, this type of analysis has had a direct consequence regarding how I now consider each position in front of me, opening up more possibilities and not limiting me to my self-identified "playing style" or "natural" inclinations.  While I do believe that the concept of a playing style is valid and observable, I am more inclined to reject it as an excuse for choosing obviously inferior moves that do not take advantage of the position.

More specifically, here is a description of the method I've been using for analysis of past games:
  • They were all initially analyzed, shortly after being played, as a complete game by Fritz on a reasonable time setting (usually 60s/move). This process catches all of the major tactical ideas and offers some insight into alternative moves beyond simply checking for blunders.  
  • The setting for changes in position evaluation (the "threshold" option under complete game analysis) was usually set to 20 centipawns (in other words +/- 0.2 in a numerical positional evaluation).  This is enough of a swing in the position to be worth flagging, while cutting out a lot of what for humans would be meaningless differences.
  • The game is loaded into a database program, with Houdini running as the analysis engine.  I also bring up my current personal openings book database so I can compare it to the game.
  • I review the game move-by-move.  In the opening phase, I look at the choices made, the implications of alternative choices to the main lines (especially moves by my opponent that are not in my openings book), and evaluate the positional characteristics arising from them.  (This is a fancy way of saying whether I like the position or not.)  The engine is used only infrequently in this part of the game analysis.  This phase helps reinforce and refine my opening knowledge and exposes me to new ideas and areas to explore.
  • The main part of the game (i.e. when it no longer follows a database game) is devoted to looking at the ideas contained in game positions and investigating alternative moves.  These alternative moves are considered based on the original Fritz and current Houdini analysis functions.  I have Houdini set to display its current top three alternatives in a position, which offers a good amount of variety.  Alternatives are looked at critically and in order to understand their tactical and strategic ideas, rather than attempting to always find the "best move" and slavishly following engine recommendations.
  • The game score is annotated with more detailed commentary on individual move alternatives and key variations, while thoughts on the overall course of the game and its significant lessons are then captured, in this case in an introductory blog post.
This analysis process takes roughly two hours per game for me.  I consider this a reasonable amount of time to spend, especially since the focus is on improving my play, rather than achieving a mastery of deep annotations.

As a final note, there has been a fair amount written publicly on the use of computer analysis by players looking to improve their level of play, a good portion of it negative.  My experience has been that computer-assisted analysis is quite valuable and productive, with a great deal of return on the time invested.  One still has to do one's own thinking; that said, having engines point out alternatives, which can then be analyzed and understood, has for me resulted in an improved thinking process.

No comments:

Post a Comment