OpenChess

Dynamic Openings Test Suite (DOTS) is a set of selected positions for running engine matches. The purpose of this set is to provide dynamic and interesting games.

DOTS6 has 800 positions.

Download:
https://www.mediafire.com/file/lzqvapxc ... 2.zip/file

The strenght difference between these engine must be minimal.

I found some interesting statistics including DOTS6 here:
https://sites.google.com/site/computers ... testings-2
However, I don't agree with the statement in this article that some sets are weak and some are strong - those with a lot of draws. My set is just an unbalanced one.

dorsz wrote: ↑

Thu Nov 24, 2022 8:32 pm

The strenght difference between these engine must be minimal.

I found some interesting statistics including DOTS6 here:
https://sites.google.com/site/computers ... testings-2
However, I don't agree with the statement in this article that some sets are weak and some are strong - those with a lot of draws. My set is just an unbalanced one.

Hi!

Updated DOTS6 ProteusSF-Sunrise-Stockfish-test and other gauntlets results. Waiting for you updated engines to start TCEC23 Gauntllet

When will it start?

dorsz wrote: ↑

Thu Nov 24, 2022 8:32 pm

However, I don't agree with the statement in this article that some sets are weak and some are strong - those with a lot of draws. My set is just an unbalanced one.

This test was botched :
- without 100 games or more per opening, the margin of error is enormous
- not all the openings of each suite have been played, the displayed elo scores are tottaly random
- too few engines tested, no idea of ​​favorable openings for each engine
- etc.

Yes, but I did not see similar test to this one.

deeds wrote: ↑

Mon Nov 28, 2022 6:47 am

dorsz wrote: ↑

Thu Nov 24, 2022 8:32 pm

However, I don't agree with the statement in this article that some sets are weak and some are strong - those with a lot of draws. My set is just an unbalanced one.

This test was botched :
- without 100 games or more per opening, the margin of error is enormous
- not all the openings of each suite have been played, the displayed elo scores are tottaly random
- too few engines tested, no idea of ​​favorable openings for each engine
- etc.

Isn't this the same person who ran a Duel Book Vs Experience file competition where he used one of your old Eman experience files with a version of StockfishMZ? I think he matched the experience file against several books.

But luckily he learned the error of his ways by making the following observation: "After checking more carefully, it is clear that
Eman's data is not much suitable for StockfishMZ. . . "

Yes it's the same person ! I think you mean
that other botched test where all is wrong :
- don't reveal the source of the eman exp file
- don't check the experience rate of the testset
- no SFMZ's experience data
- don't train against these books
- use a biased score by adding cumulative losses of an "1 against several" odd match
- etc.

As for DOTS6 and the other testsets, there is a desire to denigrate !

Then I understand that an organizer of opening book tournaments, who has been doing this for years, does not want to throw all his work away. Newer CPUs with a lot more threads and accelerating instructions hurt people who stuck with hyperthreading off, numa off, no bmi2, no avx2, etc. Even SYZYGY and NNUE technologies seem to hurt him because he believes on "ponder on" yet.

sarona wrote: ↑

Wed Nov 30, 2022 12:58 am

..he used one of your old Eman experience files with a version of StockfishMZ?...

No, it was the old (2021) Khalid's Eman.exp generated during online games at InfinityChess.