Sunday, August 12, 2007


Obey the Tactics

From my recent Knights Errant post:
Chess is not more than tactics. Chess is tactics. If we could calculate tactics perfectly, instantly to any depth there would not be any talk of "strategy." Strategy is the application of an accumulated common experience as a (poor) substitute for perfect tactical skill. The same is true of opening theory.

Quoting GM Yuri Averbakh from Chess Tactics for Advanced Players (a wonderful book on tactics):
The grandmaster Teichmann once aptly remarked that chess was 90 per cent tactics. Each one of us -- from beginer to world champion -- knows how true this is.

From Wikipedia:
Chess Tactics
Chess computers are considered superhuman at tactics and rather spotty at strategy. The fact that computers can dominate the best humans suggests that chess is primarily a tactical game.

Strategy versus tactics is similar to the difference between Heuristics and Algorithms.

Checkers has been solved:
Solving Checkers

Chess may never be solved in the same sense as checkers but it has become apparent in the last twenty years that computers doing "just" brute force calculation can beat the best humans. Calculation is a tool we use to evaluate moves and sequences of moves. It is a tool that computers wield better than humans. Humans look at a very small subset of the moves and sequences that computers consider. Humans prune this move tree using knowledge of chess, general principles, guidelines and strategy to find their way through the maze of possibilities. Humans must do that because we are not capable of doing what the computers do -- calculate every possibility to some significant depth. If humans were capable of calculating all of that they would not rely on those general principles, guidelines or strategy. Strategy is a heuristic while calculations are an algorithm.

It is even possible that future chess programs will incorporate more and more chess knowledge and strategy as they reach the limits of brute force. If so, they will be doing it for the same reason human players do it -- there are limits to how much they can calculate.

In practice, most chess games between non-masters turn on simple tactics. In master games they turn on tactics, less simple.

Is my point that General Chess Knowledge or Strategy or Endgame knowledge or Opening Theory does not matter? No, not at all. But tactics rule. If the general principal says "capture towards the center" but the tactics say otherwise in that specific position, obey the tactics.


I think you are technically right but, as you suggest, in practice this doesn't help all that much. Since we'll never be able to crunch through variations like a computer, we are left considering factors that will be correlated with tactics in the future (e.g., outposts, mobility, etc). In quiet positions, especially, it is impossible to think through the game tree with any thoroughness, so we end up going on 'intuition' and general principles.

Incidentally, computers already use lots of strategy in their evaluation function. It is pretty easy to code strategic ideas into the mix (e.g., maximize mobility all else being equal, don't create holes all else being equal), and this is now pretty standard.
Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?

Subscribe to Posts [Atom]