Engine missevaluation?

Good day everyone,

I have a question regarding the game above.

The move 54. f8=Q is perceived by the engine as a blunder, and it drops the evaluation of the position to +0.38, even though, as far as I can analyse, it is completely winning for white; am I missing something here?

(do avoid looking too deeply into other parts of the game for your own sake  )

(p.s. the line 56.Kg7? is what's suggested by the engine)

54. f8=Q ... is completely winning for White. What engine is telling you it's not?

Chess.com's analysis engine agrees 54. f8=Q is completely winning for White.

chess.com engine analysis

the move shifts the eval to +0.38, even though every continuation suggested by the engine is completely winning. I am baffled.

The Engine analysis is correct to my understanding. after Rxf8 Kxf8 kh1  Ke7 h2, wherever white mives except for rook stalemates. after white moves rook out of g-file, Kg7 threatens promotion, it would be a draw.

You are getting an inaccurate assessment because the analysis depth is set to only 10, which is a fairly severe handicap. You should increase the analysis depth for better results.

The picture I posted is Chess.com's analysis engine at a higher depth. The limited 18-ply Chess.com post-game analysis function also concludes the move is completely winning. 

Is the king+pawn endgame winning?

I do not think the position is a draw. White will eventually move the Rook to the h-file and sacrifice it when Black tries to promote the pawn. White’s King will then win the race to support the promotion of the d3 pawn. 

I see, it does seem rather obvious now.

Thank you for the answer

The way the game finished, the white rook was sacrificed for the promoted queen on h1 and the black king was too slow to return.