CCRL Discussion Board

Possibly enough of an Elo gain for it to reclaim 5th place in this round-robin:

https://ccrl.chessdom.com/ccrl/Chess324/index.html

I'll run it once a fix is out.

So did you achieve 70% ?

Gabor Szots wrote: ↑Sat Sep 16, 2023 7:49 am

Graham Banks wrote: ↑Sat Sep 16, 2023 5:39 am DLL files missing.

They run OK here without additional DLL's.

Won't run here, missing DLL message. I suspect it is OK for you Gabor because you probably have GCC installed on your machine ?

The last couple of weeks I ran a tournament using the SPCC 3-move Chess 324 opening book Chess324_xxl_big_+090_+119 which has 70355 lines. (evals per KomodoDragon 2.6) Download from here: https://www.sp-cc.de/anti-draw-openings.htm The tournament is at the link below: https://ccrl.chessdom.com/ccrl/...

None of us have Macs. Chess is dominated by Windows and Linux machines, but potentially one of us may be able to compile for Windows (which is what all of us happen to use)

All of the testers here are Windows users, and the usual procedure is for authors to provide windows exes for download.

Is that 136 games ? If so, error margins are probably not far off +/- 50 Elo.

Shredder GUI is obviously coded to deal with this, when it creates the .eng file it silently substitutes the "/" automatically with an underscore.

Caissa 1.9 (AVX2_BMI2).eng

In the file ChsGUI_Eng.TXT [Caissa 1.9 (AVX2-BMI2)] MyName=Caissa 1.9 (AVX2-BMI2) OriginalName=Caissa 1.9 (AVX2-BMI2) EngineFile=C:\ChessEngines\Caissa 1.9\caissa-1.9-avx2-bmi2.exe EngineAuthor=Michal Witanowski CommandLine= EngineCode=571763226312736 157288800 128 0 0 0 0 0 0 0 0 0 EngineInts=0 1 2...

Solved it by manually editing the ChessGUI config files.

Eng07 (caissa-1.9-avx2-bmi2) <- AdjustTokenPrivileges failed, error code: 1300 Cancelling StartEngineTimerEvent() for caissa-1.9-avx2-bmi2 at 0000000189899531. ChessGUI thinks the failure to allocate large pages is a critical error and then aborts ? However, I still think that "/" in the ...

OK I can duplicate this in ChessGUI as well (although I haven't used that for years now)

Perhaps it does not like the "/" or the brackets in the UCI engine name, that is all I can suggest.

bastiball wrote: ↑Mon Jun 05, 2023 12:53 pm Try run some tournament with a handicap

I'm not really interested in that. But other ideas for something a bit different are welcome . I'm still keen on a DFRC tournament of some sort, but I need to put some work in to get a subset of those openings to work with.

Very low draw rate as expected with unbalanced openings. Indeed, but probably a lot of rubbish games. I don't think so, these are the least-unbalanced in that opening set, nothing extreme, and remember they are all from human games. Thing is with NNUE engines, the evals are no longer centipawns in ...

Graham Banks wrote: ↑Mon Jun 05, 2023 1:20 am No real surprises.

No, I don't see any big surprises. Yes there is a very low draw rate, but the resulting rankings and ratings of the engines seem similar, all within the error margins with a small number of games.

Very low draw rate as expected with unbalanced openings.

Rankings of engines are similar to the standard CCRL blitz list. Any differences could just be statistical margins of error.

Games can be downloaded here for anyone interested:

http://ccrl.chessdom.com/public/UHO2022 ... ndrobin.7z

If you prefer Ordo ratings: 1 Stockfish 20230531 : 2494.2 592.5 780 76 2 Komodo Dragon 3.2 : 2446.5 548.5 780 70 3 Berserk 11.1 : 2384.8 485.5 780 62 4 Ethereal 14.00 : 2341.4 438.0 780 56 5 RubiChess 20230410 : 2319.9 414.0 780 53 6 Koivisto 9 : 2319.5 413.5 780 53 7 Revenge 3.0 : 2286.7 376.5 780 ...

Head to head statistics: 1) Stockfish 20230531 780 (+409,=367,-4), 76.0 % vs. : games ( +, =, -), (%) : Komodo Dragon 3.2 : 60 ( 25, 31, 4), 67.5 : Berserk 11.1 : 60 ( 27, 33, 0), 72.5 : Ethereal 14.00 : 60 ( 31, 29, 0), 75.8 : RubiChess 20230410 : 60 ( 32, 28, 0), 76.7 : Koivisto 9 : 60 ( 32, 28, ...

BayesElo ratings, default parameters: Rank Name Elo + - games score draws 1 Stockfish 20230531 157 20 19 780 76% 47% 2 Komodo Dragon 3.2 120 19 19 780 70% 50% 3 Berserk 11.1 70 18 18 780 62% 53% 4 Ethereal 14.00 35 18 18 780 56% 54% 5 RubiChess 20230410 17 18 18 780 53% 54% 6 Koivisto 9 17 18 18 780...

I decided to try a one-off experiment, a tourmament using the mildly unbalanced 6-move UHO_2022_6mvs_+110_+119.pgn compiled by Stefan Pohl. Testing conditions: Format: Round robin tournament with the top 14 engines from CCRL single CPU lists GUI: Cutechess with 500ms overstep margin, concurrency 12 ...

Interesting, I don't think there are that many python engines out there.

Possibly because there are not enough games for a reliable rating.

You mean games played in April, not the April ratings list ?

You are going to have to provide Windows exes. None of our testers use Linux, and none of us are experts in compiling code.

I have a couple extra questions that I have not been able to find answers for, and which I think wouldn't hurt to put on public record. i) When running BayesElo, what values are used for anchoring the rating lists? ii) Was anchoring to FIDE rating lists performed, and if so, what historical notes d...

Could be, but when I'm not running chess that boost is worth having.