Tuesday, March 20, 2007

When and how much to analyze?

Analysis, or internally working through variations ("If I go here, and he goes there, and I go here,...") is a skill possessed aplenty by the grandmasters. In this post I examine what seems to be a consensus opinion on when analysis is necessary, and how far into the future one should analyze.

Soltis, in How to choose a chess move devotes Chapters 4 and 5 to the topic. His basic claim is that in positions with lots of potential forcing moves (tactics, checks, and the like) you simply must calculate the continuations for those forcing moves. In quiet positions, you shouldn't spend a lot of time on analysis because it is so hard to predict your opponent's replies anyway.

Here is a frankenquote from Soltis that I sewed together From Chapters 4 and 5 of his book:
In many quiet positions you can go ahead and play the candidate virtually without any calculation because there are no significant replies to worry about [he calls these 'low calc moves']. When the pawn structure is fairly static and enemy counterplay is limited, the calculating quotient declines sharply. The basic guideline is: You can afford to overlook most quiet moves because they're quiet. You must examine all forcing moves because they're forcing. In very sharp positions, low-calc options are rare. The price of failing to look two or three moves into the future can be high.
How many moves ahead should you calculate those branches of the tree with forcing moves? Soltis says (Chapter 5):
The minimum number usually depends on how far into the future one player can continue to make forcing moves. In sharp positions in which your opponent is doing the threatening, you should continue looking until his moves have run out of force. But bear in mind we are talking about a minimum number of moves to look ahead. If you have the clock time to spend, you should analyze the position until you run out of forcing moves--and then look one move further.
What of Kotov? His book Think like a grandmaster is often criticized for imposing unrealistic expectations on players, to think through dozens of variations on every move. However, this is an inaccurate picture of what he actually says. On the question of when to do a deep analysis, we find that Soltis is basically a mirror of Kotov. Kotov says, in Chapter 1:
When a position is closed and lacks direct contact between the opposing forces, then the choice of the best move is normally made based on positional factors and positional considerations predominante. Thinking will be based on general considerations without concrete analysis. When the opening leads to sharp hand-to-hand fighting then you analyze and analyze.

For the record, Soltis suggests alternatives to Kotov not about when to analyze, but in the best way to analyze. For Kotov, you must look at each branch of the tree for each candidate only once, then move on to the next candidate, and so on, to maximize cognitive efficiency. Soltis is more flexible, and discusses the benefits of other methods of analysis that Kotov would call inefficient.

My next post on analysis will describe what various authors have to say about how to get better at analysis. This is what we all want!

Note: the following is an ongoing list of relevant quotes I'm adding subsequent to publication of the original post.

Blumenfeld, Chapter 3 of the otherwise horrible book Attack and Defense:
In situations that are not sharp, where there cannot be any forced variations, your calculations should be confined to a few short lines which serve to bring out the characteristics of the position.
From the commentary on Game 1 in Euwe's Chess master vs chess amateur, an amazing book:
Analysis is the basis of all good chess play, especially when the position becomes tactical.