The engine score doesn’t tell the full story

It is common nowadays to have some chess engine running while analyzing a game, an opening line or even following a live GM event. We all know how strong chess engines are. As an example, the very popular Stockfish is around 800 rating points higher than Magnus Carlsen, the best chess player of all times. The difference between Stockfish and Magnus is similar to the difference between Magnus and a random club level player. It makes sense to have almost complete faith in the numbers provided by the engine when evaluating the position from a game in progress.

Indeed, the numbers are right. But that’s only if we assume perfect play, something that we, as humans, are not capable of. The “patzer with the engine running” might see a score of +3 and assume that white is easily winning, then laugh when the grandmaster playing that position fails to win it (or even loses). The most important thing to understand is that the score does not refer to the current position on the board. The +3 evaluation is what the engine thinks about the position reached after a certain number of moves, if both follow what the engine deems to be the correct path (best moves for both sides). Strong chess players are very good at assessing quiet positions, where not much is happening. They can fully understand a pawn up position where the other side has no significant counterplay, and would win that game most of the times, even if the current engine score is just +1. But the same strong chess players can have a very hard time winning a position where the score is a lot higher, because they fail to find the very narrow path leading to that big advantage. This should be easy to understand if we take it to the extreme. Imagine a position where white is a full queen down, but there is a forced mate in 8 moves. If the player doesn’t see the (potentially difficult/hidden) path to checkmate, he would regard his current position as dead lost (queen down).

So … whenever I hear someone say things like “I lost a +5 position”, my question is “what kind of +5?” Was it a rook up in an endgame? Then it’s really sad, the player did something terribly wrong. Huge time trouble, horrible technique, rushing … could be a number of things. But if that +5 score only flashed for a brief moment and was predicated on finding a particular tactic, it can happen to anyone.

I will take one of my games as an example. I think it shows quite clearly the difference between a static and a dynamic advantage, and how the engine score given at various points of the game does not tell the full story.

Lessons to learn?

• In very dynamic positions, the engine score can paint a picture that has little to do with what the players see and feel during the actual game
• “Understanding” is overestimated in master games. In general, all master level players have a good grasp of the fundamental notions of positional play. But some masters are 2100, and others are 2800.
• The ability to calculate concrete variations quickly and without missing important candidate moves was, is, and always will be the most important chess skill. It’s a point made by my favorite book as well.
• Good time management is essential. The game is often times decided by a tactical sequence. It has the bad habit of occuring when players have only a few minutes left on their clocks.