Looking for a draw tool ...

[Frank Quisinsky]

Hi there,

is there perhaps a tool available that can do the following:

Basis: *.pgn database

Give the number of draws to:

- 3-fold
- 50-moves rule
- stalemate

Example:
*.pgn database with 1.000 games, 600 are draws.

Output:
- 600x draws
- 125x 50-moves rule
- 25x stalemate
- 450x 3-fold

Best
Frank

[Ras]

There is pgn-extract which can filter a PGN database with --stalemate, --fifty, and --repetition. See:
https://www.cs.kent.ac.uk/people/staff/djb/pgn-extract/
https://www.cs.kent.ac.uk/people/staff/ ... /help.html

[Frank Quisinsky]

Thanks Rasmus!
Not thought about this program.

Best
Frank

[Colin-G]

Load the pgn file into Scid.
Use the search function to search for text in the PGN contains text box.
Entered text is case sensitive in my Scid version.
e.g. in the picture below, Stalemate is found in 14 games out of 6000 games.

[Frank Quisinsky]

Hi Colin-G,

yes, Scid is indeed a very strong database.

I am looking more for the reason to do the following:

Engine A, 3.000 games ... how many % of draw games ended with 50-moves-rules?
Engine B, 1.500 games ... how many % of draw games ended with 50-moves rules?

Contempt used?, Contempt parameter in UCI options yes / now

This have the following reason:
For the future I will kill all engines (not have a contempt parameter and avoid 50-moves rule) for the reason:
Save electricity with games and longer time-controls, without resign-mode.

Yesterday in the evening:
Dragon 2.6 NN (Komodo) - Stockfish 15 NN
Game over 330 moves ... games over 300 moves replayed for my rating system.

The second game goes over 200 moves.
Constellation:
KR vs. KR
And both engines avoid the 50-move rule thinking hours for the next move ... with 4-pieces in clear draw constellation!

Games with longer time-controls are a "Torture".

I can fully understand if others used time-controls like 3+1 or game in 5 minutes or playing with resing=on
In most of cases I have here no fun in looking games between TOP-41 engines with longer time-controls and resign=off.

Yesterday I started a ponder=on test between TOP-41.
The same, a lot of engines have today problems with the time-managment with ponder=on.

Unfortunately:
Many programmers do really a lot to make computer chess more and more uninteresting for + 2-3 Elo.

The move average per game goes with each new engine update higher and higher.
Can be see in my rating list ... at the moment for around 98.000 games = 90,03 moves.
More as 10 engines are responsible.

Can be see also in my KI-Ratings (*.jpg):
https://www.amateurschach.de/t-ki/common/ki-ratings.jpg

In times today a miniature in computer chess is around "80" moves!

I dislike the way many programmers goes and have no problem to write about it.

Other programmers found a way to keep the move-average low!

This annoys me if different programmers are not thinking on useless consumption of resources!!!
Exactly this one must stand on the first position!

Best
Frank

Tomorrow I will looking in pgn-extract.