I think Larry Kaufman is wrong.

Larry Kaufman is a computer chess expert. However, in the April issue of Chesslife in the article "What Your Engine is Trying to Tell You" he makes several assertions I disagree with.

1) "a score of 1.00 is now defined as a position where white is expected to win 50% of games from that position, with black drawing or winning the other half"

I am not disagreeing with that being the definition, but with the evaluation being correct. Specifically in closed openings at the beginning of the game.

2) He then uses that as the basis to show why the drop in popularity of certain openings at the top level, i.e. Philidor, Pirc, Kings Indian etc.

That may be the reason why, but I'm not sure the reasoning is correct.

3) Finally he states "Many test have shown that if Black wastes even a single tempo at the start, he is dangerously close to losing, i.e. 1. e4, Black 'passes' (illegal, I know), 2. d4 is around +.90"

This was the comment that caught my interest, because I thought I had looked at this exact position with engines in the past and came to the opposite conclusion.

Luckily with todays engines and software all this theorizing can be swept aside.

With Lucas Chess (R) (I believe that is the version that includes the neural net files) I set up a quick test.

1) I created the starting position of d4 e4, basically giving White a free move as per his example. And true to Larry's prediction most top engines gave an evaluation of around +.90.

2) I choose the 3 top engines, Stockfish 16.1, Komodo Dragon 1, lc0 and had each engine play itself for 2 games with a time limit of 10 minutes each side from that position.

The predicted results per the article should have been roughly 3 wins for white. The actual results were as follows.

Stockfish - 2 draws

Komodo Dragon 1 - 2 draws

lc0 - 1 win for white 1 draw

*it should be noted even in the win for white, the evaluation had dropped down to a drawn evaluation (.16) at one point.

So, small test admittedly, but 1 win 5 draws. Not exactly 50%. And that was with the worst possibly defense (the "I pass" defense).

With this in mind I don't see how I could agree with some of his conclusions (i.e. on the pirc, philidor, kings indian etc.)

PS. A couple of notes on the test. The most common defense chosen by the engines was c6 d5 kind of structure. The draws varied, from an exchange down and an extra pawn, to dead drawn king and pawn games. lc0 played the most human like, but was also the only one to lose a game.

PSS. I am a Larry Kaufman fan, just really disagreed with this article

@1

"if Black wastes even a single tempo at the start, he is dangerously close to losing"
++ Black can afford to lose a tempo, e.g. 1 e4 d5 2 exd5 Qxd5 3 Nc3 Qd8.
White can afford to lose 2 tempi, e.g. 1 a3 e5 2 h3, or 1 Nf3 d5 2 Ng1 e5.

"a time limit of 10 minutes each side" ++ That is too short: a blitz game.

"kings indian" this seems right:
https://www.iccf.com/game?id=948179https://www.iccf.com/game?id=948250

I may test a longer time control, but guessing it would increase the draw percentage?

@3

"guessing it would increase the draw percentage?"
++ Yes. See Figure 2

I missed the first part of his article (dont have a Chesslife subscription) so I may be misunderstanding what he is saying.

You know what they say about opinions...

@5

"dangerously close to losing" means: still drawing, but difficult to draw in a real game