First MEA rating list

[Rebel]

Which engine are you using and which options would you like to add besides MultiPV and I will try.

I am asking because MEA doesn't always respond correctly on --eoption

[peter]

Which engine are you using and which options would you like to add besides MultiPV and I will try.

I am asking because MEA doesn't always respond correctly on --eoption

There are several, e.g. Ethereal needs it's NNUE to be set by

Code: Select all

 --eoption "EvalFile=ethy.v1425.std.nn" 

Lc0:

Code: Select all

 --eoption "NNCacheSize=8000000" 

(for longer TC, 8G is the max. of my 3070ti- GPU)
ShashChess:

Code: Select all

 --eoption "High Tal=true" --eoption "Variety=Psychological" 

That's examples of syntax, these UCI options work for me with MEA , but not with NICE (at least not those with Lc0 and ShahChess, Ethereal I haven't tried so far) regards

[fsanders]

I want to give another feedback to NICE:
The "debug.txt" files do not correct cout points here an example

Pos 502
[ 2025-03-22 11:44:50,737 - INFO ] EPD: 8/1P2k1b1/3pp2p/6pP/3pP1N1/3P4/q4PP1/6K1 w - - c0 "b8=Q=10";
best move : b8=Q
time : 7
points : 0

This has to do with double equal sign I think.

[fsanders]

Code: Select all

	 1	stockfish-17.1	  3940	00:08:26.5	 226   480	00:00:46.7	 4800  82.1%   1000	  128	 1	   0
	 2	stockfish-17.1	   189	00:04:10.0	  13	20	00:00:00.0	  200  94.5%   1000	  128	 1	   0
	 3	stockfish-17.1	   188	00:04:10.0	  10	20	00:00:00.0	  200  94.0%   1000	  128	 1	   0
	 4	stockfish-17.1	   182	00:04:10.0	  13	20	00:00:00.0	  200  91.0%   1000	  128	 1	   0
	 5	stockfish-17.1	   177	00:04:10.0	   9	20	00:00:00.0	  200  88.5%   1000	  128	 1	   0
	 6	stockfish-17.1	   175	00:04:10.0	  12	20	00:00:00.0	  200  87.5%   1000	  128	 1	   0
	 7	stockfish-17.1	   172	00:04:10.0	  10	20	00:00:00.0	  200  86.0%   1000	  128	 1	   0
	 8	stockfish-17.1	   171	00:04:10.0	  14	20	00:00:00.0	  200  85.5%   1000	  128	 1	   0
	 9	stockfish-17.1	   169	00:04:10.0	   7	20	00:00:00.0	  200  84.5%   1000	  128	 1	   0
	10	stockfish-17.1	   169	00:04:10.0	  11	20	00:00:00.0	  200  84.5%   1000	  128	 1	   0
	11	stockfish-17.1	   168	00:04:10.0	  10	20	00:00:00.0	  200  84.0%   1000	  128	 1	   0
	12	stockfish-17.1	   168	00:04:10.0	   9	20	00:00:00.0	  200  84.0%   1000	  128	 1	   0
	13	stockfish-17.1	   162	00:04:10.0	   9	20	00:00:00.0	  200  81.0%   1000	  128	 1	   0
	14	stockfish-17.1	   160	00:04:10.0	   9	20	00:00:00.0	  200  80.0%   1000	  128	 1	   0
	15	stockfish-17.1	   159	00:04:10.0	   7	20	00:00:00.0	  200  79.5%   1000	  128	 1	   0
	16	stockfish-17.1	   158	00:04:10.0	  10	20	00:00:00.0	  200  79.0%   1000	  128	 1	   0
	17	stockfish-17.1	   158	00:04:10.0	   6	20	00:00:00.0	  200  79.0%   1000	  128	 1	   0
	18	stockfish-17.1	   156	00:04:10.0	   9	20	00:00:00.0	  200  78.0%   1000	  128	 1	   0
	19	stockfish-17.1	   152	00:04:10.0	   9	20	00:00:00.0	  200  76.0%   1000	  128	 1	   0
	20	stockfish-17.1	   152	00:04:10.0	  10	20	00:00:00.0	  200  76.0%   1000	  128	 1	   0
	21	stockfish-17.1	   150	00:04:10.0	   8	20	00:00:00.0	  200  75.0%   1000	  128	 1	   0
	22	stockfish-17.1	   143	00:04:10.0	   9	20	00:00:00.0	  200  71.5%   1000	  128	 1	   0
	23	stockfish-17.1	   143	00:04:10.0	   9	20	00:00:00.0	  200  71.5%   1000	  128	 1	   0
	24	stockfish-17.1	   140	00:04:10.0	   6	20	00:00:00.0	  200  70.0%   1000	  128	 1	   0
	25	stockfish-17.1	   136	00:04:10.0	   7	20	00:00:00.0	  200  68.0%   1000	  128	 1	   0

I think there is something wrong with NICE counting points. I did the following test:
1) Takeing the first 480 positions of mea-stripped.epd
2) Analysed this 480 positions with Stockfish 17.1 using NICE by splitting it into 24 threads
3) Added the result to mea-results.csv

Now adding the points of the 24 parts should give the same result as the overall result.
The overall result is given as 3940.
Adding all the parts gives 3897.
Why there is a difference I could not find out, but I can confirm that the 24 numbers are the same as given in the 24 log files.

[fsanders]

Just for my testing, will there be a new version of NICE, or is it as it is?