Monday, September 23, 2019

Computer Aided Analysis


  What follows is my thoughts on this subject. I do not claim that this information is in any way perfect. I share them with the hope of being of some help to others. If you have something to share leave a comment. I'd love to hear from you. 

  Computers can out play us. It's a fact. Can they also teach us? Yes and there are a lot of programs and websites that do just that. One that comes to mind is Chessable.com. One area where computers might seem to fall short is in the area of game analysis. You put your game into an analysis engine and it spits out a very machine focused review of the game. The explanations if any make no sense to the human mind. Garry Kasparov has been at the forefront of explaining how computers and Artificial Intelligence in particular should be seen as a tool. A learning tool. Maybe a partner in the quest for knowledge.  
  
  Software designers are working on improving this communication gap. Websites like Decode Chess are making incredible strides in this area. The game report and analysis at Chess.com have undergone a design change toward this goal.  

Still for now at least I think we will have to settle for taking the computers advice and trying to make sense of it ourselves. In this article I'm going to share how I do that. I hope someone might find value in this sharing.

So let's begin. At first I thought I would try to pick one of my better games to write about, Then I realized that was wrong. I try to analyze all my games. The good, the bad, and the ugly. We especially want to analyze our losses. So here it is. My process...

First the game without annotation. I run through the game and get a quick impression of what happened. If you want to follow through my process just play through the moves below and see what thoughts come to mind as you do.



Spoiler alert here are my first thought (memories) of what transpired in this game.

  1. My opponent though rated in the 1500's violated multiple opening axioms. Like don't bring out your Queen early, and Knight's on the rim are dim. Yet I lost this game.
  2. My opponent developed like a new player. Looking for a "cheapo" Qf7#.
  3. I must have blundered in the middle game.
  4. I remember only playing on so long on the chance of a stalemate. 
Next I load the game into a database program. Here I'm using SCID vs PC. (See my 3 part article about setting up SCID vs PC) You can use about any chess database program. Chessbase of course. I have a large database of games (over 7 million) that I load in the "Tree" window. Then I start stepping through the moves. When the tree runs out of moves I mark that move as a Novelty. To see the graphics larger right click and select open in new tab. Then open that tab and look it over.

So in this game the Novelty happened at 6. ... Nge7. Is this where I blundered?


  No! A deep look by Stockfish shows that black is about equal or maybe has a very slight advantage.    
Next I start Stockfish running and slowly step through the game. Looking for the first move when the advantage shifts from Black to White. That occurs at 23. ... Nb4. Where bxc4 is better. I'm interested in fixing my game at the earliest error. So I'll return to this later. For now I continue stepping through the moves. At 25. ... Nd3 I find the big blunder. White gains a decisive advantage. No need to look further.


Next I load the game into Decode Chess.com. For my method their decode recommendations come way to late in the game. So I select my two key plays. It becomes clear to me now. The move 23. ... Nb4 allows the White Queen to take c5 and b5! On 26. Qxb5 the Knight and Bishop are attacked.


So 25. ... Nd3 appears to be the big blunder, but the snowball started rolling at 23. ... Nb4.



So I need to be more careful about letting pawns fall, and letting a Queen invade my space. I'd love to hear about your process or your opinion of mine. Soon I hope to publish a article about using a game database to find your strengths and weaknesses. Consider following this blog for future updates. Thank you.