Improving your own play can be a very difficult process, especially when you try and do it without the assistance of another stronger player. Of course software can help, but using the information that the software provides can be a difficult task.
A quick scan of USENET (internet news groups) shows that most people have no concept of what the analysis process entails. The typical post goes something like this: "What is the best software to analyze my games?" Some one will come back with the usual reply which is "Fritz". They'll then go on to describe the process of importing games into the program and setting analysis parameters. End of story.
Indeed, a common misconception about computer analysis is that all you need to do is to plug your games into some chess program (be it Fritz, Chessmaster, or Chess Assistant), and let it analyze the games for a while. Come back a few hours later, and your analysis is complete. Great! But what has actually been accomplished here?
Many people don't realize it, but real chess improvement only manifests itself in the human analysis that takes place after the engine has finished its work. The technique I am about to outline is very simple in concept, and should be employed as part of the game analysis process, which is discussed more fully in this article. Essentially, it involves tracking and classifying mistakes that you make. Classifying these mistakes makes it easier to see faulty thinking patterns. However it is not quite as easy as it sounds, since determining the root cause of a mistake can be quite difficult. Using CA to keep track of these errors turns out to be the easy part of the process.
The easiest way that I've found to keep track of my mistakes is to create a classifier. Typically, I have one or two databases that I use to keep track of my own games. Over a long period of time, I've gradually constructed a classifier that catalogs the mistakes I make. To make your own mistake tracking classifier, you should go through the following steps:
More Detail on the Process
Determining the cause(s) of a mistake
During this part of the analytical process it helps to remember why you played (or did not play) a particular move. Thus, the analysis process should be conducted relatively soon after you've played the game you are analyzing. First, let's assume that you've either conducted an automatic engine analysis of the position in question, or are attempting a manual analysis. To explain how this is done, let's take a look at the position to the right.
Black has just played ...Bd7, and it is now white's move. Clearly white has some choices for his response. b5 can be rejected out of hand since it provides a wonderful square for the knight c5. Several moves previously, I had seen that black would play Bd7 eventually, and had planned to play Qa6. However, when the game reached this point, I had spent a little more time thinking than my opponent, and quickly played Qb3?!. After examining the computer analysis for this move (which indicated that Qa6 was best, fastening on a nice weakness, and attacking the knight on b7 as well), I went through my recollections of the game, and reconstructed my thoughts. When I really looked at my reasons for playing this move, I was forced to realize that I have a tendency to retreat pieces when they are attacked. And maybe I have a tendency to respond defensively to an attack, rather than offensively. In other words, there are several problems here. One is a failure to notice and make use of a weak square (a6). Another is not noticing some counterattacking potential when confronted with a threat. The third problem was an inherent tendency to retreat a piece when attacked.
I categorize failure to make use of a weak square as a positional error. Not noticing counterattacking potential I place more in the psychological category, since it relates to an overly defensive frame of mind. The tendency to retreat when attacked, I place in the realm of calculation errors. So this position was annotated with three classifier nodes.
Once you've put words to the mistake, you need to categorize it (as I did above). In my case, I've settled upon a scheme that breaks mistakes down into the categories shown in the figure below. Note that my classification structure is three levels deep (i.e. there are no more classifier nodes past this point). I've left off a number of nodes, but the diagram should be enough to give you an idea of what I'm talking about.
You should use whatever structure suits you best. The above diagram is meant only to give you some ideas. My mistake classifier is actually quite complicated, with many tens of folders. It did not make a good screenshot, so I used the above simplified diagram to give you the general structure. If you want more basic information on classifiers, see this article.
Adding a node to the classifier
Adding a classifier node is fairly easy. Simply open your mistakes classifier (assuming it has been created). Then click on the "+" icon in the button bar. This will add a new classifier node, that you can then give a descriptive name to. The name should be one that describes the mistake in question (once again, see the picture above).
Once you've created the node, you'll want to assign the appropriate search criteria to it. You can do this very quickly by right-clicking over the classifier, and selecting Folder data->Self search criteria->As Annotation. What this does is tell the classifier to look for moves that are annotated with a pointer to this particular classifier.
Annotate the move with the node/folder
When you are annotating a game, and come to a position where you made a mistake, you should assign the appropriate classifier node to the preceding move. You do this by going into the annotation dialog (accessible via the ctrl-a key combination), and selecting the tab called "User Folders". Then click on the "Add" button.
What you are doing is telling CA that is part of the classifier you defined in the previous step. Once the classifier browser window comes up, you need to click on appropriate classifier, and then click "Ok".
You'll then see that the appropriate entry has appeared in the annotations tab. Make sure that you save all the changes you've made to the database.
I should also add the number immediately to the right of a classifier node tells you how many games occurred with that particular type of error. However, this number is not automatically updated by CA. If you want to make sure that this number is current, you will need to select the particular classifier(s) you want to update, and execute the "Repeat search" command from the right-click context menu. Note that you can use "Select all" command (followed by "Repeat Search"), which is also in this menu, to update all the classifiers at once. Incidentally, if you keep repeating searches for the classifier nodes as new games are annotated, you'll see a running tally of the number of times a particular type of mistake was made. This number is very important, since it gives you an indication of what types of errors you commit most frequently.
Reviewing your mistakes
All you need to do is open the classifier containing your errors. Then go to the folder containing the type of errors you want to review. Double-clicking on the folder will bring up a list of games where that error occurred. However, I've found that the annotation process is in itself helpful. You'll notice that as you create this classifier structure, certain patterns will emerge. For example, another problem I've noted in my games is a failure to consider moving pieces back to their home squares. This pattern became evident after I'd analyzed nearly 200 games.
While this article mostly concerns the mechanics associated with tracking (and becoming more aware) of your own deficiencies, the real key lies in the self improvement and examination process. It is important to occasionally go back and take stock of where your weaknesses lie, and review those positions that were troublesome for you in the past. The technique that I've outlined in this article is an important part of that process.