Is there some recent depth change your mind research?

[Uri Blass]

I know that there was a research about probability to change your mind at different depth with old programs but not with programs like stockfish.

I suggest the following idea:

1)Take stockfish and a big database of games and allow stockfish to analyze for 1 second every position in the games(suppose games have 1,000,000 positions).

2)save an epd file with the following details
a)position
b)depth that stockfish got
c)move that stockfish suggest
d)last depth that stockfish changed its mind(1 in case stockfish never did it)

It may be interesting to see the distribuition of 2 and 4

3)Take only the positions that stockfish could analyze to the biggest depth that is not the maximal depth(let say 10000 positions) and analyze them again with stockfish (1 minute per position) to see if stockfish change its mind in these positions

One interesting question is what is the maximal depth that practically stockfish change its mind and what is the probability that stockfish changes its mind as a function of depth.

It may be also interesting if we find practically cases that stockfish does a mistake inspite getting the maximal depth(I know that stockfish cannot solve some studies inspite of getting the maximal depth but I do not know if it happens in games).

[Dann Corbit]

I have a lot of data like this.
I have hundreds of thousands of positions for which I have saved the last ten plies of analysis.
I store a supposedly best entry in my main table, and all the other entries go into a table called AlternateEvals.