Ray wrote: ↑Sun Mar 17, 2024 5:25 pm
Yes "Rybka" is 1 thread only. But Deep Rybka does not have a UCI parameter for threads either. The parameter is "Max CPUs". Old engine as you say. But with cutechess cli you must surely be able to specify options for individual engines as well as options common for all. If you use it properly then that is fine - it is used by many people just not any in CCRL.
Look at what TCEC or CCC do. Last time I looked they gave the cli scripts they use so you can learn from those.
It's true, Deep Rybka uses the uci CPU MAX parameter not threads. I have added 4 in cutechess interface and am trying a gauntlet 120+1 256 hash book depth=12. Both in the taskmanager and in the program that I use to monitor the use of hardware resources, btop4win, use 4 threads. See Screenshot:
![Image](https://i.ibb.co/rvbjhgB/ccrl.png)
In the management of hashes, since in this test, Deeprybka is the reference engine, I note that it uses 9MB per thread. On the other hand, its more modern opponent does come close to the assigned hash usage, that is, 256mb. But the question I have now, I have looked in the subforums, is the opening book that you recommend. Well, I understand that it must be the same book for all the tests, although I may be wrong about this, but since it is about settling the distance between engines, I understand it this way.
Code: Select all
Rank Name Elo +/- Games Score Draw
0 Deep Rybka 4.1 4 CPU -271 83 69 17.4% 26.1%
1 minic3.41 4 CPU inf nan 6 100.0% 0.0%
2 caissa 1.17 4 CPU inf nan 6 100.0% 0.0%
3 Alexandria 6.0.1 4 CPU 470 nan 8 93.8% 12.5%
4 arasanx 24.1 4 CPU 446 nan 7 92.9% 14.3%
5 Winter_3.0 4 CPU 417 nan 6 91.7% 16.7%
6 Wasp650 4 CPU 417 nan 6 91.7% 16.7%
7 combusken2.0 4 CPU 417 nan 6 91.7% 16.7%
8 nemorino 6.0 4 CPU 280 nan 6 83.3% 33.3%
9 marvin 6.2.0 4 CPU 191 238 6 75.0% 50.0%
10 Vajolet2_2.8 4 CPU 58 110 6 58.3% 83.3%
11 naum 4.6 4 CPU -191 238 6 25.0% 50.0%
SPRT: llr 0 (0.0%), lbound -inf, ubound inf
69 of 440 games finished.
Regarding the scripts, thank you for your recommendation, since I have all the possible ones that appear in the cutechess-cli github and it is possible to use a bin book, you just have to change -pgn opening, etc. to -book Mode. But I will look at TCEC and ask at Openbench, too, who are always very willing to help. Since looking at the interface and the command version, the command version is much better due to its number of configuration options and resource use. Regarding the book, I await your response and Graham has told me about the latest version of Chessgui, as I also understand that I must use what you use. To avoid distortions between test benches. Thank you in advance and good morning from Terrassa, Barcelona. A big hug to all of you.
Jorge Ruiz Centelles
I am not an expert in anything, I do not believe in anything, but everything is possible.