Far from perfect

[Henk]

During each analysis an engine encounters many positions that require long sequences of quiet moves. Because they cannot cope with these positions computer analysis can not be trusted.

[Dann Corbit]

During each analysis an engine encounters many positions that require long sequences of quiet moves. Because they cannot cope with these positions computer analysis can not be trusted.

During each analysis, a human encounters many positions that expose the player to a series of microscopic inaccuracies. In total, they can lead to disaster as they accumulate. Because they cannot cope with these positions, human analysis cannot be trusted.

[Henk]

Brute force/exhaustive search until a shallow depth is best for then you know exactly which positions they missed otherwise we get fooled by these amazing search depths.

[Lyudmil Tsvetkov]

During each analysis an engine encounters many positions that require long sequences of quiet moves. Because they cannot cope with these positions computer analysis can not be trusted.

During each analysis, a human encounters many positions that expose the player to a series of microscopic inaccuracies. In total, they can lead to disaster as they accumulate. Because they cannot cope with these positions, human analysis cannot be trusted.

but a human, coupled with an engine, certainly will provide far better analysis than any engine.

truth is, computers play some positions better, and humans other positions better.

[Lyudmil Tsvetkov]

I am very happy with engine play in simple pawn endgames.

In the last 5 years, I have not won a single pawn endgame against a top engine, while that used to be frequently the case in earlier days.

[corres]

[quote="Henk"]
Brute force/exhaustive search until a shallow depth is best for then you know exactly which positions they missed otherwise we get fooled by these amazing search depths.
[/quote]
It was not a long time ago when the common view was in the case of alpha-beta search that the first twenty plies depth have real value. I think ought to the large scale optimisation of parameters and the much larger test games to tune them nowadays about the first 25 plies depth have real value.
Naturally it depends on the given engine too.

[JJJ]

I am very happy with engine play in simple pawn endgames.

In the last 5 years, I have not won a single pawn endgame against a top engine, while that used to be frequently the case in earlier days.

Will you play again against Stockfish in correspondance here ?

[Lyudmil Tsvetkov]

I am very happy with engine play in simple pawn endgames.

In the last 5 years, I have not won a single pawn endgame against a top engine, while that used to be frequently the case in earlier days.

Will you play again against Stockfish in correspondance here ?

too weak, that SF, it has first to win TCEC final.

[Dann Corbit]

During each analysis an engine encounters many positions that require long sequences of quiet moves. Because they cannot cope with these positions computer analysis can not be trusted.

During each analysis, a human encounters many positions that expose the player to a series of microscopic inaccuracies. In total, they can lead to disaster as they accumulate. Because they cannot cope with these positions, human analysis cannot be trusted.

but a human, coupled with an engine, certainly will provide far better analysis than any engine.

truth is, computers play some positions better, and humans other positions better.

And there is some randomness in the analysis of both.