lichess.org
Donate

how do combinations work in the game analysis?

So when you analyze a game you get a running graph of who's winning.

Let's say you make a move that allows the opponent to fork your rook and king and force a loss of the rook. The opponent misses it, and then you move the rook again and all's back to normal. So this would be a combination at depth "1" and it would show up in the analysis.

Now let's say you make a move that allowed your opponent to put you in check, there's only one move for your king to move to, and on that square as above the opponent can move their knight to fork your king and rook, and force the loss of your rook. The opponent misses this check, and you move your rook and avoid this. This would probably be reflected in the analysis because your valuation would go way down due to the existence of that fork, and then go back up when your opponent misses it, even though it's a combination with depth "2".

My question is, when we're looking at a graph, how intricate, how deep "combinations" get reflected in valuations? 2? 3? 4? 5? 6? 7? 8? How deep are combinations analyzed?

This is particularly interesting to me because in practical terms against 1500-1800 players who I play with, the existence of an 8-move combination wouldn't necessarily put me in a worse position practically, since the opponent would almost certainly miss it.

So how deep do valuations (in the chart that shows who's winning) go?
the "Local computer evaluation
for variation analysis" is 18 plies deep and gives the same evaluation as the analysis provided so id guess it possibly might be these 9 moves deep.
The computer analysis, which crunches and provides data to generate the graphs, runs for a set period of time, going through nodes in the count of millions in a cluster of Stockfish instances, allowing distributed analysis of every position occuring in the game for as deep of depth as it can traverse, which is not 2, or 8, or 18, but as deep as it can go.

Also, combinations of outcomes aren't analyzed, but positions. Stockfish is but a program that crunches through the position and extracts factors from it (such as king safety, mobility...) to give a position rating.

This topic has been archived and can no longer be replied to.