H. G. Muller wrote on Mon, Jul 13, 2020 05:47 PM UTC:
You can click on the 'move' header of the piece table to switch that column between displaying the Betza notation for the move, and the piece value. But you should not take the values too seriously. For one, they are estimates for the value the piece would have in a game without piece drops. When pieces can be dropped this tends to push the values closer together, as even the pieces with very low mobility can easily be transported to remote locations where they deliver threats immediately by dropping them. Also, with piece drops the promoted value has more impact, because pieces tend to be dropped in the zone, and then promote very easily.
Even in games without drops some of the well-known piece values are wrong. (E.g. the Archbishop.) Amazingly enough the ratio of the values of the orthodox pieces seems very reasonable. But the algorithm is based on (smartly weighted) average mobility only, and it does not recognize handicaps such as color binding, or worse forms of area binding. So it would likely over-estimate the value of, say, an Alibaba. That the value of a FIDE Pawn it gets is reasonable is probably just a coincidence, because it ignores the latent value due to the promotability, but also does not realize the Pawn is bound to a single file. These errors seem to largely cancel each other.
The Diagram's AI is not intended to be an analysis tool for piece-value determination; its purpose is to act as a not-too-stupid sparring partner for somewhat experienced chess players who just learned the rules of a new variant. Very accurat piece values are not needed for that purpose. That it can show the values at all was mainly for debugging purposes.
You can click on the 'move' header of the piece table to switch that column between displaying the Betza notation for the move, and the piece value. But you should not take the values too seriously. For one, they are estimates for the value the piece would have in a game without piece drops. When pieces can be dropped this tends to push the values closer together, as even the pieces with very low mobility can easily be transported to remote locations where they deliver threats immediately by dropping them. Also, with piece drops the promoted value has more impact, because pieces tend to be dropped in the zone, and then promote very easily.
Even in games without drops some of the well-known piece values are wrong. (E.g. the Archbishop.) Amazingly enough the ratio of the values of the orthodox pieces seems very reasonable. But the algorithm is based on (smartly weighted) average mobility only, and it does not recognize handicaps such as color binding, or worse forms of area binding. So it would likely over-estimate the value of, say, an Alibaba. That the value of a FIDE Pawn it gets is reasonable is probably just a coincidence, because it ignores the latent value due to the promotability, but also does not realize the Pawn is bound to a single file. These errors seem to largely cancel each other.
The Diagram's AI is not intended to be an analysis tool for piece-value determination; its purpose is to act as a not-too-stupid sparring partner for somewhat experienced chess players who just learned the rules of a new variant. Very accurat piece values are not needed for that purpose. That it can show the values at all was mainly for debugging purposes.