Stockfish evaluation

[feistyram]

I understand Stockfish has had a new evaluation function since 15.1

"An evaluation of +1 is now no longer tied to the value of one pawn, but to the likelihood of winning the game. With a +1 evaluation, Stockfish has now a 50% chance of winning the game against an equally strong opponent. This convention scales down evaluations a bit compared to Stockfish 15 and allows for consistent evaluations in the future."

I find this quite confusing. I find it easier to think in terms of centipawns.

Are all other top engines (that are not Stockfish derivatives) producing evaluations based on centi-pawns still? I my swtch to one of those for my personal analysis

[darmar]

I understand Stockfish has had a new evaluation function since 15.1

"An evaluation of +1 is now no longer tied to the value of one pawn, but to the likelihood of winning the game. With a +1 evaluation, Stockfish has now a 50% chance of winning the game against an equally strong opponent. This convention scales down evaluations a bit compared to Stockfish 15 and allows for consistent evaluations in the future."

I find this quite confusing. I find it easier to think in terms of centipawns.

Are all other top engines (that are not Stockfish derivatives) producing evaluations based on centi-pawns still? I my swtch to one of those for my personal analysis

No, only Stockfish. You can try Komodo Dragon but forget centipawn, it uses different metrics

[feistyram]

There's so many free engines, I don't see the point in paying for Komodo.

[Rowen]

I understand Stockfish has had a new evaluation function since 15.1

"An evaluation of +1 is now no longer tied to the value of one pawn, but to the likelihood of winning the game. With a +1 evaluation, Stockfish has now a 50% chance of winning the game against an equally strong opponent. This convention scales down evaluations a bit compared to Stockfish 15 and allows for consistent evaluations in the future."

I find this quite confusing. I find it easier to think in terms of centipawns.

Are all other top engines (that are not Stockfish derivatives) producing evaluations based on centi-pawns still? I my swtch to one of those for my personal analysis

I was the same when I started to analyse with stockfish. Incidentally I remember seeing a stockfish graph for expected game results with the new evaluation's, an evaluation of 1 had an expected win chance of 50% AND 50% chance of draw giving a 75% expected outcome. An evaluation of 2 had virtually a 100% expected outcome. So I remember those two evaluations and their expected outcomes when analysing. . By the way dragon 1 can be downloaded for free on their site.

[Uri Blass]

I understand Stockfish has had a new evaluation function since 15.1

"An evaluation of +1 is now no longer tied to the value of one pawn, but to the likelihood of winning the game. With a +1 evaluation, Stockfish has now a 50% chance of winning the game against an equally strong opponent. This convention scales down evaluations a bit compared to Stockfish 15 and allows for consistent evaluations in the future."

I find this quite confusing. I find it easier to think in terms of centipawns.

Are all other top engines (that are not Stockfish derivatives) producing evaluations based on centi-pawns still? I my swtch to one of those for my personal analysis

I do not find it confusing.
Humans also do not think in terms of centi-pawns in games.
If you are 2 pawns up and the opponent has some initiative you do not think about value of initiative in centi-pawns.

[Rowen]

I understand Stockfish has had a new evaluation function since 15.1

"An evaluation of +1 is now no longer tied to the value of one pawn, but to the likelihood of winning the game. With a +1 evaluation, Stockfish has now a 50% chance of winning the game against an equally strong opponent. This convention scales down evaluations a bit compared to Stockfish 15 and allows for consistent evaluations in the future."

I find this quite confusing. I find it easier to think in terms of centipawns.

Are all other top engines (that are not Stockfish derivatives) producing evaluations based on centi-pawns still? I my swtch to one of those for my personal analysis

I do not find it confusing.
Humans also do not think in terms of centi-pawns in games.
If you are 2 pawns up and the opponent has some initiative you do not think about value of initiative in centi-pawns.

I guess with centi-pawns evaluations it seemed simple, if the engine gave a valuation of plus 1 for white you knew that the engine 'thought' that, white had a winning position equivalent to 1 pawn extra, irrespective of what the advantage constituted. Yes it was vague and perhaps arbitrary and subjective but at least the meaning was simple.

Now with the new evaluations the user will have to interpret the evaluations with respect to likelihood of winning, and know what +!, +2 meant etc. Perhaps its just a learning experience, once you know it is fine. It was centi-pawns for a long time. Now many still use centi-pawns and some dont.