Mikhail_Golubev wrote on 03/18/20 at 21:07:55:


I do not remember if it was in a Dragon book from Kalinin???? I had seen unfaithfully some ...h5 against g4. I wonder if it is this variation.

But even 12 moves from each side won't change this position terribly much!

All you're seeing is whether the various computers evaluation function thinks this material balance is = or better for black, which is whatever they programmed into it.

I bet the other comps are just 4 pawns > piece and a little for the passed pawn. Houdini presumably just likes pieces a little bit more

The evaluations of stable positions by computers aren't remotely magic or even especially reliable. Especially not in randomly unbalanced endings like this.
(the concrete tactics that they can spot often are magic of course. But not present here.).

Meanwhile the really obvious human evaluation is surely that white can't ever win this position but can lose it if careless? If you want to know if black has genuine winning chances you'll have to do a careful (hand!) analysis of the various plausible winning plans. Computers never been at all good at all that, and won't ever be.

Edit by moderator: for the time being I let this go, because this is still related to the initial post. But if the subject becomes silicon features I will have to split this thread.
MNb

What is the depth of analysis?
I always go at least to depth 25 ( I have a very fast machine )

Take the engine's evaluation at face value and don't read anything into it - instead pay attention to the way it trends up/down as the game unfolds as that's more likely to tell you something useful. You need to be paying attention to what's going on long-term.

For example with Houdini: there's a Modern Defense variation that favors white, but Houdini thinks black is better. As the game unfolds it takes about 20 moves before it realizes white is better.

Maybe this position is technically equal, but it doesnt matter! black has brilliant winning chances here and the end position is very easy to play, black can be satisfied he has atleast a draw.

Another point is that the majority of white players dont like allowing Rxc3 and Qxa2 and will just play a3 and Nd5 and go into the drawish ending, where in my opinion black slightly better chances.

I doubt a strong player with a 1.e4 repetoire would analyse and learn the theory to get to the end diagrammed position!

A recent of my Dragon games was in this line and when I analysed the victory I discovered something strange with Houdini. This side line of 9 g4 has been given as good for black for decades. All engines I tried so far calls in black advantage except Houdini who calls it equal!
I also discovered an alternative black option.