which engine is stronger?

[towforce]

Is it really a weakness to take a 50-50 gamble on a win or loss, rather than preferring a certain draw?

Code: Select all

if (probability(success) = 0.5)
	"not weakness";

else if (in_Unsound_Moves_Get_Killed_World)
	"weakness"

Increasingly, we are living in "Unsound Moves Get Killed World" - and not just in chess!

[Lazy_Frank]

... The 200/700/100 on the other hand shows that in at least 10% of the games X2 plays a losing move...

Given information do not says which openings (if any) engines play's. Can be "objective busted" and both engines just converts all of them.

[hgm]

True, I assumed balanced openings.

[jkominek]

We see several examples Stockfish 15 not losing a single game and yet it gets negative rating points.

I would characterize it differently. BayesElo and Ordo optimize whole-tournament rating assignments. Looking at the image you cut & pasted from CCRL showing Stockfish 15 matches, in all but two encounters (Shash 22, Dragon 3), every engine is an Elo donor. Some donate more than expected, some less. As I interpret it, the performance column is taken relative to the rating difference expectation. There will always be some on the negative side, and some on the positive side of the balance.

I confess I cannot duplicate the CCRL Perf column by hand. Example: the rating difference to Berserk 9 is 86 Elo. From the standard equation this predicts a performance of 62.13%. Yet, feeling in form, Stockfish mopped the floor with Berserk and scored 73.21%. This corresponds to a rating delta of 174.7 ... 175-86 = 89. But, they show the performance as only +53. So, I dunno.