Your network blocks the Lichess assets!

lichess.org
Donate

Ghost in the Engine

I don't think the last thing about the computer switching to "checkers" is not completely foreign to human chess. You might even consciously recognize it in classical chess. When do you decide to end your blunder check? When there are no more interesting checks captures or queen promotions to look at most likely. In speed chess we probably do the same thing just more subconsciously. Maybe we skip it when we are playing on instinct. However, in almost all cases when a chess player "blunder checks" generally you are taught to stop (given sufficient time) to look at all the CCiT's. Also I am curious how the identification of a threat limits the move search like algorithmically. Sure analogously if we see a threat for the opponent our move must stop it or do at least as much, but we know what specifically what we are trying to stop, ex "Nc7 fork". However, the algorithm as I understand it only identifies a threat exists from the opponent and how significant is it however that doesn't tell me anything about which moves address or create exceeding danger. Interesting article.

I don't think the last thing about the computer switching to "checkers" is not completely foreign to human chess. You might even consciously recognize it in classical chess. When do you decide to end your blunder check? When there are no more interesting checks captures or queen promotions to look at most likely. In speed chess we probably do the same thing just more subconsciously. Maybe we skip it when we are playing on instinct. However, in almost all cases when a chess player "blunder checks" generally you are taught to stop (given sufficient time) to look at all the CCiT's. Also I am curious how the identification of a threat limits the move search like algorithmically. Sure analogously if we see a threat for the opponent our move must stop it or do at least as much, but we know what specifically what we are trying to stop, ex "Nc7 fork". However, the algorithm as I understand it only identifies a threat exists from the opponent and how significant is it however that doesn't tell me anything about which moves address or create exceeding danger. Interesting article.

Null Move Pruning (NMP) is an interesting concept and I can definitely see how it could work well.

However it also completely fails in a position with mutual zugzwang where it prunes winning lines and keeps losing lines.

Personally I mostly focused on evaluation of position rather than selection of which positions to evaluate. At one stage I wondered if fact engine calculates evaluation using criteria like number of squares a piece can reach along with positional considerations like weak squares, doubled pawns etc that all have a weighting that if you knew it you could in theory count the engine eval at the board and play move that would boost engine eval most using those metrics. Unfortunately I don't think it worked as while it might be OK positionally it also neglected tactics.

Null Move Pruning (NMP) is an interesting concept and I can definitely see how it could work well. However it also completely fails in a position with mutual zugzwang where it prunes winning lines and keeps losing lines. Personally I mostly focused on evaluation of position rather than selection of which positions to evaluate. At one stage I wondered if fact engine calculates evaluation using criteria like number of squares a piece can reach along with positional considerations like weak squares, doubled pawns etc that all have a weighting that if you knew it you could in theory count the engine eval at the board and play move that would boost engine eval most using those metrics. Unfortunately I don't think it worked as while it might be OK positionally it also neglected tactics.

Oh man this is hard to understand. I don't get your point. What a waste of time.

Oh man this is hard to understand. I don't get your point. What a waste of time.