CCRL Discussion Board

i) How many simultaneous games do CCRL testers typically run? ii) If it is more than one, is benchmark calibration performed when the machine running a CPU load in line with what is expected during the tournament? In Kirill's post he suggests rebooting to get a clean slate before benchmarking. But ...

Yes, as I said, different tester, different hardware, different book, different GUI and even a different compile of Stockfish 15.1. But the large error margins are surely the dominant factor. +/- 20 Elo or so on each engine with the low number of games. Stockfish 15.1 on one thread was tested: Intel...

ssakgul wrote: ↑Sat Dec 17, 2022 5:53 pm What does mean SF 15.1 64-bit > SF 15.1 64-bit 4CPU?

Different tester, different hardware, different book, different GUI. And large error margins with the small number of games, so it is impossible to say. Totally inconclusive.

There is also Taktix.

In Windows you just double-click the Stockfish 10 exe and type "bench", Adapted time control for 40/15 would be/: T minutes for 40 moves where T = Your time / 2054 x 15 (for 40/15) From this post I'm assuming that 2054 is the time produced on real hardware for the "standard" mac...

The ratings and the rankings are always subject to the statistical error margins, so even when engines are separated by 10,20,30 Elo you really don't know unless you play many thousands of games. Which isn't possible at this time control with limited resources.

People still use Wndows 7 ??

Frankly if I were you I would not expend any effort to support that. Unfortunately when you use old hardware and old operating systems you have to expect that over time some newer software might not work.

I've sent a PM with a download link. Let me know when you've successfully downloaded and un-zipped it.

Why we don't offer it is that a minority of the games do not contain this information in the source pgns, possibly due to a GUI configuration error on the part of the tester, or some other reason. If you were prepared to sort through the games to identify those with and without the extended informat...

gepardo wrote: ↑Sun Feb 06, 2022 12:03 pm
The executables are built for Windows and GNU/Linux. They are 64-bit only, as I didn't optimize the engine to perform well on 32-bit machines. I hope it won't be an obstacle for testing.

Hi there, no problem, everyone has x64 capable hardware and operating systems now.

Ordo ratings (Ethereal fixed at 3500) # PLAYER : RATING POINTS PLAYED (%) 1 Stockfish 060122 64-bit : 3652.7 658.0 900 73 2 Komodo Dragon 2.6 64-bit : 3619.9 620.0 900 69 3 Ethereal 13.50 64-bit : 3500.0 465.0 900 52 4 Berserk 8.5.1 64-bit : 3499.3 464.0 900 52 5 SlowChess 2.8 64-bit : 3490.8 452.5 ...

Hi all, hope your holidays have been nice and you're keeping warm if you're in the North. I'd like to submit my engine today for inclusion in CCRL. It's written in Rust, uses the UCI protocol, is open source, and its repository can be found here: https://github.com/Heiaha/Weiawaga Graham says that ...

I recently ran this tournament Ryzen 9 5900X CuteChess GUI Time control 4'+2" All engines single CPU 5-men Syzygy tablebases Hash 256MB Book 8moves_v3.pgn PGN http://ccrl.chessdom.com/public/4+2_tournament.7z I thought I'd build myself a small database for some possible experiments later. . .

We did manage to implement this. We put the RTX 2080 engines in the 8CPU lists as well since they are competing in the same space, but another option is to have a separate list for them.

Yes it is very rare that we test private engines. When someone sees an engine on our list, they should be able to get it - either freely downloadable or able to be purchased in case of commercial like Dragon, Fat Fritz, Hiarcs, Shredder etc. I'm puzzled as to why we have Stoofvlees on the list actua...

Then the answer is easy, don't use an https URL. However we should look at upgrading if possible.

I'm not getting any errors, it is working normally here

I don't think anyone has bought it.

The Stockfish results are all over the place in the various lists. A couple of them do in fact agree with the fishtest results. There are so many variables - time control, threads and book choice for example. In our case I believe different hardware can also impact the results especially today with ...

Can you please make another one or two posts until it gets past the point where your posts require approval.

Gabor Szots wrote: ↑Fri Jul 09, 2021 12:14 pm
There must be some memory leak which makes the allocated memory grow with each move, at move 70 it has already surpassed 1GB.

Ouch !

dkappe wrote: ↑Mon Jul 05, 2021 5:20 am Still telling me I need a moderator.

Can you try again, and if that doesn't work we'll have to ask the forum administrator what is wrong.

Is there a discussion somewhere about calibrating for lc0/gpu? We don't currently calibrate for the GPU. Time control for games is simply set based on the CPU speed. That works so far because no two testers have had the same GPU. It would be a problem if two testers had the same GPU in machines wit...

They seem fine to me.

Broken how exactly ?