Squaring the circle

I've been pondering this for a while and think I have found a fairly convincing way to resolve the argument (convincing to me at least).

We are regularly told that chess is all (ok 99%) tactics. Look at your games, almost all are decided by a simple tactical oversight by one player or the other. This is certainly true at all amateur club levels, up to the lower titles and perhaps arguably even as high as grandmasters. The blunders are usually a bit more sophisticated at their level but they still count as blunders relative to the players' knowledge and ability. I hardly need mention the ending of the recent World Championship match.

And yet.

Does this really mean that chess is just a matter of moving pieces around randomly until you spot (or make) a blunder? My answer is no, strategy really matters!

In order to investigate this question, I devised a simple experiment which I believe demonstrates this convincingly. I played a trial game against an engine designed to play at a fairly moderate level but (and this is the crux) to never blunder. I did this by using the analysis board on chess.com which automatically displays 3 analysis lines. For the engine side (playing black in this single experiment), I selected the worst move out of those three that did not exceed 50 centipawns loss relative to the best move. That is, if the three proposed moves were all within 0.5 of each other on the engine evaluation, I played the 3rd one. If the 3rd was more than 0.5 below the top move but the 2nd was above this threshold, I played the 2nd. If both the 2nd and 3rd were more than 0.5 below the top move then I played the top move (typically but not always this occurred on a recapture).

[Note that this is not how lower levels of engines are usually designed. They are typically configured to play huge blunders at least occasionally. It's an ongoing challenge to make engines play in a human manner, but that's another story.]

I repositioned the browser window on my screen so I couldn't see the full analysis lines proposed by the engine, just its first move for each of them. And I made my own move for white rather rapidly without being tempted to take a peek at the engine recommendations. It wasn't a very serious experiment, just a trial to see what happened. I'm sure a competent programmer could set up something along these lines more rigorously, but I can't really be bothered with that.

And what happened is that I beat the engine in a fairly straightforward manner. The engine made a large number of fairly aimless moves, never a proper blunder, but moving to and fro allowing me to accumulate positional advantages such as doubling and isolating their pawns until I won a material advantage and forced victory. Note that the engine side would probably have failed any decent cheat detection tests as it was literally always playing one of the top three engine moves. I'm sure I played some worse moves for white, but mostly I played better.

The engine made no substantial tactical errors yet played pretty weakly. Unfortunately I deleted the result without managing to save it (it wasn't a proper game as such, just an analysis) so no further analysis of the game is possible at this point but I reckon it must have played with an average centipawn loss per move of about 30 give or take a bit. By design the loss was capped at 50 for every move, and quite frequently the third move in the analysis list was rather better than this, and of course the centipawn loss was zero on moves when it was forced by circumstances to play the optimal move. 30 is quite a respectable value for a club player around my standard, I'm certainly worse on occasion though hope to do better most of the time (it varies with the nature of the game).

So, my point here is that strategy really does matter! If you play with no strategic insight, just moving stuff around, then even if you manage to avoid all blunders and never drop so much as a whole pawn in a single move, a reasonably competent opponent will still be able to accumulate enough advantages over time that they will convert to a win.

In practice I believe strategy matters more than this experiment can show, because an accumulation of strategic advantages creates a situation where one side has more tactical pitfalls to avoid than the other. The final error that brings defeat may be labelled by the computer as a "blunder" and by following the analysis one can often see how that move leads to an immediate loss of material that could have been avoided. However when you are walking a tightrope finding only move after only move, playing against an opponent who has 3 decent choices at every turn, it's hardly surprising that it's the tightrope-walker who is more likely to fall. If they'd avoided that one blunder, there would probably have been another one in a couple more moves.

So I'll continue to enjoy working on strategic analysis and thoughts during my games, while still recognising that there are large gains to be made by working to reduce tactical blunders and oversights. The answer is not tactics or strategy, but tactics and strategy!