Do not believe the engine

Sometimes the engine gets it wrong. In a game I played yesterday, Stockfish insists that Black has a decisive advantage. But, this is a textbook draw. 

White will exchange rook for Black’s queen, leading to a pawn ending where Black’s bishop is on the wrong color square. White’s king will maintain control of the promotion square for Black’s remaining pawn.

Yes, there are a few types of positions where hundreds of years of human experience and learning is still valuable.

The computer evidently knows it is a draw.  That's probably why it doesn't take the h pawn.

I wouldn’t wander with my king. There are ways the bishop and king can cut off White’s king from the light corner.

In the game I premoved a king shuffle between h1 and g2 until Black’s king reached h3. Then h1 and g1.

Too easy. I backed the game up to my opponent's blunder and played that out against Stockfish. Notice that Stockfish does not even try, unlike my opponent, who although above 2000 in rapid, may not have understood the position.

It matters how I use the engine on this site. When I analyze the game with the engine running, even on this site, Stockfish sees that it is a draw. When I use the analysis feature, it leads to the result indicated in the OP.

In the autoanalysis offered by the site, Stockfish recognizes equality here:


But, it criticized my subsequent play as I hastened for the textbook draw that is clearly present in the position. The arrow marks the point of the above diagram.


One lesson here is to limit the confidence one places in chessdotcom's game analysis.

I don't know the engine settings you used (because you didn't mention them).  Stockfish 15 on depth 20 showed me -0.87. When I clicked Run Cloud Analysis it adjusted to 0.00 in just a couple of seconds and remained there.

Stockfish can be wrong at lower depths, but I don't think Stockfish is insisting that black has a decisive advantage.

Look at the image. The setting is shown.

I use the site's default because all my serious analysis is done offline in ChessBase.

Computers have changed chess, but there are still some things they don't know!

Computers have shaped the game of chess and can play better than any human ever has. Yet, computers aren't perfect and there are still some positions that you can understand better than your engine.

https://www.chess.com/lessons/when-computers-go-wrong

don't think you'll have this problem with tablebases. But chess.com's engine doesn't come with those I believe.

this question is for the better players. 

how does an engine evaluation determine how you might move in the future? i ask because the evaluations have a numerical value for the current position (admittedly, i have no understanding of 'depth'). surely those above 2000 level have seen many of these positions and it is static. any advantage or disadvantage an engine "sees" can be changed significantly in a few moves, no?

It does not. It analyzes all possible moves* and evaluates millions of future positions.

"Depth" is the number of half-moves the engine calculates ahead.

When you wrote "surely those above 2000 level have seen many of these positions and it is static", I don't understand that sentence at all.

*Well, this is not really true, but almost.

the static position is where the pieces are on the board. there is no context regarding player strength. this is also apt regarding my statement about 2000+ players, like you. i would assume you, and others like you, have seen many positions on the board and can see many moves ahead. your asterisk seems to make my point-that analysis is based on past results and possible moves via engines while it is humans that are playing.

i also have no idea what "half-moves" are. 

1. e4 e5 is a move

1.e4 is a half move.

Not a lot of engines do come with 8 man tablebases.

Like any software, chess engines can have bugs or issues that may affect their performance. These bugs can range from minor issues that do not significantly impact the engine's play to more significant problems that may cause the engine to make mistakes or produce incorrect analysis.

One example of a bug that has affected chess engines in the past is the "50-move rule bug," which occurred when the engine incorrectly counted the number of moves made in a game and declared a draw due to the 50-move rule being reached, even though the rule had not actually been satisfied. Another example is the "threefold repetition bug," which caused the engine to incorrectly recognize a position as a threefold repetition, leading to a draw being declared when it was not warranted.

It's important to keep in mind that chess engines are not perfect and can make mistakes, just like human players. While it's generally a good idea to rely on chess engines for analysis and guidance, it's also important to use your own judgement and to be aware of the limitations of the engine.

In recent years, newer chess engines have been developed that are faster and more powerful than older engines like Stockfish. These newer engines often use advanced techniques such as machine learning and neural networks to improve their analytical abilities and enhance their performance. While Stockfish is still a strong engine, it has indeed been surpassed by some of these newer engines in terms of speed and analytical ability.

I believe Lichess' software had a three-fold repetition bug at one stage. I distinctly remember a draw being declared after my opponent moved his king back and forth between two squares in order to force an (erroneous) draw a few years back.