There are few existing programs able to extract depths (average depth) from pgn files.
I need a tool able to do the same but with score. Compatible with popular pgns like these from cutechess or Arena.
I mean not only average score but average score versus opponents also (it's important).
So if we have gauntlet with A vs B/C/D/E engines, output would look like:
An average score of A vs B "X.X"
An average score of A vs C "X.X"
An average score of A vs D "X.X"
An average score of A vs E "X.X"
An average score of B vs A "X.X"
An average score of C vs A "X.X"
An average score of D vs A "X.X"
An average score of E vs A "X.X"
An average score of A "X.X"
An average score of B "X.X"
An average score of C "X.X"
An average score of D "X.X"
An average score of E "X.X"
Is there such a tool? I searched a lot but didn't find anything like this, even created some spreadsheets with ability to read pgn and compute score but it works only with fixed nr of moves and consumes too much time.
Im looking for a tool
Moderator: Ras
-
Dann Corbit
- Posts: 12814
- Joined: Wed Mar 08, 2006 8:57 pm
- Location: Redmond, WA USA
Re: Im looking for a tool
It is a surprisingly difficult challenge. The problem is that there is no standard definition for the format of the score or even the sign of the score. It can be the score according to white's perspective or it can be the score according to the side on move.
Every chess GUI prints the data in a different way, and even with the same chess GUI, the output changes in format from year to year and month to month.
All that having been said, my friend Les Fernandez wrote a tool set called CUT (Chess Utility Toolbox) that has among its functions turning analyzed PGN games into decorated EPD records. It is not always successful because some engines will have non standard output (e.g. reporting approaching mates with enormous scores above 32767).
If you are interested, I could ask him for permission to give you a copy.
The way that Les approached the problem was to use an expanding collection of regex expressions. When a new format is found that is popular, a new regex is added to the collection.
Every chess GUI prints the data in a different way, and even with the same chess GUI, the output changes in format from year to year and month to month.
All that having been said, my friend Les Fernandez wrote a tool set called CUT (Chess Utility Toolbox) that has among its functions turning analyzed PGN games into decorated EPD records. It is not always successful because some engines will have non standard output (e.g. reporting approaching mates with enormous scores above 32767).
If you are interested, I could ask him for permission to give you a copy.
The way that Les approached the problem was to use an expanding collection of regex expressions. When a new format is found that is popular, a new regex is added to the collection.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
-
majkelnowaq
- Posts: 28
- Joined: Fri Aug 10, 2018 4:07 pm
- Full name: D.S.
Re: Im looking for a tool
We can assume that's not a case. Using Stockfish-like engines with default options (related to score) is enough for me.Dann Corbit wrote: ↑Sun May 24, 2020 7:31 pm (...)The problem is that there is no standard definition for the format of the score or even the sign of the score. It can be the score according to white's perspective or it can be the score according to the side on move.(...)
It is not always successful because some engines will have non standard output (e.g. reporting approaching mates with enormous scores above 32767).
(...)
It's not a problem as well. For my spreadsheets I found a way to unify output, no matter is it form cute or arena, result is the same. Im using one of function from pgn-extraxt and convert pgn files to Chessmaster format, it erasing headings and other unnecessary data (there is one condition - pgn shouldn't has mainlines included).Dann Corbit wrote: ↑Sun May 24, 2020 7:31 pm Every chess GUI prints the data in a different way, and even with the same chess GUI, the output changes in format from year to year and month to month.
Converted game looks like (fixed nr of moves - 20):
WHITE: 1stockfish_20042208_x64_modern
BLACK: C (00)
1. e4 ... ; +0.96/18 1
... e5 ; -0.07/15 0
2. Nf3 ... ; +0.81/18 1
... Nc6 ; -0.27/16 0
3. Bb5 ... ; +0.82/18 1
... a6 ; -0.23/16 0
4. Bxc6 ... ; +0.88/19 1
... dxc6 ; -0.22/16 0
5. O-O ... ; +0.81/17 0
... Qd6 ; -0.37/16 0
6. d3 ... ; +0.71/18 0
... Be6 ; -0.37/16 0
7. Ng5 ... ; +0.77/17 0
... Be7 ; -0.19/18 0
8. Nxe6 ... ; +0.85/18 0
... Qxe6 ; -0.16/18 0
9. Nd2 ... ; +0.76/19 0
... Nf6 ; -0.07/16 0
10. Qe2 ... ; +0.88/18 0
... O-O-O ; -0.17/16 0
11. Nc4 ... ; +0.75/19 0
... h5 ; -0.02/15 0
12. Qe1 ... ; +0.76/18 0
... Kb8 ; -0.36/16 0
13. a4 ... ; +0.98/19 0
... h4 ; -0.56/18 0
14. h3 ... ; +1.00/21 0
... Nh5 ; -0.42/18 0
15. b4 ... ; +1.07/20 0
... Nf4 ; -0.17/18 0
16. b5 ... ; +1.11/20 0
... Qg6 ; -0.93/18 0
17. Bxf4 ... ; +1.01/21 0
... exf4 ; -0.80/19 0
18. Kh1 ... ; +0.89/20 0
... f3 ; -0.58/18 0
19. Rg1 ... ; +0.88/21 0
... Qf6 ; -0.24/18 0
20. bxa6 ... ; +0.91/20 0
... b6 ; -0.61/17 0 Arena Adjudication
; draw
If this tool could convert all moves and annotate them not only with score from game but also with engine name (standing behind this one move) it would be some solution. So Im interested if you could send to me link by pm with this tool, then I ll try it to see if it ll be able to do equations I need.Dann Corbit wrote: ↑Sun May 24, 2020 7:31 pm All that having been said, my friend Les Fernandez wrote a tool set called CUT (Chess Utility Toolbox) that has among its functions turning analyzed PGN games into decorated EPD records. It is not always successful because some engines will have non standard output (e.g. reporting approaching mates with enormous scores above 32767).
If you are interested, I could ask him for permission to give you a copy.
The way that Les approached the problem was to use an expanding collection of regex expressions. When a new format is found that is popular, a new regex is added to the collection.