How to use the evaluations given in CCRL games?

[chesskobra]

In CCRL games there are evaluations after each move (after the game leaves the book). I would like to use these evaluations to generate puzzles or interesting positions. Something along the lines of chess-artist.py (which is a script by Ferdinand Mosca on github), but that script uses an installed engine to analyse each position. It would be much quicker if I could use the evaluations that are already there in some game files. I would like to implement something like this, with perhaps other criteria for extracting interesting positions or puzzles. But I would appreciate if others suggested existing tools that consume the eval data in CCRL games.

[Modern Times]

The problem you will strike is that depending on the GUI used, in some games the evals will be from white's pov and other games from the engine's pov.

[chesskobra]

Yes, I noticed this. If that were to be fixed, how would one use the data? Are these evaluations being used by engine developers? How do they use them? Another question is, suppose for example I have

Code: Select all

8. Bxf6 {+0.00/1 0s} gxf6 {+0.00/1 0s} 9. Be2 {-1.15/28 25s} Bb7 {(Bb7) +0.48/29 39s} 

in a game. Does it mean white's 9. Be2 made the position worse than from move 8? If so, then I can look at only white moves, and say that if white's eval increased by large amount after white's move, then black's last move was bad, and if white's eval decreased by a large amount after white's move, then white's previous move was bad.