A bit of help for CCRL

[amanjpro]

I don't really understand the point with criticizing Sylwy 100 games match over and over in many threads.

Clearly 100 games is some info, it gives a feeling about the engine playing strength, and clearly also Sylwy is aware of SSS and uses a vocabulary in his post that does not seem confusing.

For instance, I just run this

Code: Select all

Score of zahak-linux-amd64-7.1 vs minic_dev: 3 - 85 - 12 [0.090]
Elo difference: -401.9 +/- 102.9, LOS: 0.0 %, DrawRatio: 12.0 %

and this let me think Zahak 7.1 is in the 2900Elo range. I guess this is quite true ? Of course I don't know that with a +/-5Elo margin, but this is an estimate.

If I repeat with a 2500 rated engine, I will get around +400 and the same range probably.

As everyone here is aware of the intrinsic quality of such a result, I really think it is fine, and even, why not, usefull.

Thanks Sylwy for those tests, always good to see engines play together.

Slywly doesn't say it is in 2900 range, he says it is "2950 in my single test" or "2925 in my single test" or whatever... that is the confusing part

[xr_a_y]

Predictions after my very short tests:
-Coiled_07_x64 around 3052 Elo points;

In recent post like the one for Coiled, the vocabulary seems very ok me: "predictions", "around", "short test".

Maybe a new addition to totally clarify the situation would be a Elo estimate with margin, like the one I gave: 2900+/-100 for instance.