Why do I get two different ratings for same position (López Opening: Morphy Defense, Caro Variatio) on chess.com?

Question

I've been trying to get back into chess a bit, and noticed the option to practice certain positions - but I've been puzzled by the rating when playing a "Friendly" game (which gives you feedback for your moves).

During practice, the move that's considered "excellent" is considered an "inaccuracy" in the exact same game against a computer.

I'm not sure about chess notation, I'll try to supply pictures.

The position I reach as black is this one, trying to figure out my next best move:

Ruy López Opening: Morphy Defense, Caro Variation, 5.Bb3

If I practice this position (here), Bishop to b5 is rated as "good"

But if I do the exact same move in my own game, it's rated as "inaccuracy"

The ratings also seem fairly different (+0.99 vs +0.39). What am I missing here? Is this move good, or inaccurate? Why do I get two fairly different responses for the exact same game position - does it rate differently according to the opponents Elo rating? Or am I missing something obvious?

Answer 1

This is very likely due to a different engine search depth. The same effect occurs on lichess, when you request an engine analysis after the game directly vs. when you copy it into a lichess study first. Games in studies are analyzed deeper and engine evaluations can differ. Chess.com typically uses pretty low depth (e.g. 18). When analyzing moves live on chess.com, I have noticed that annotations also sometimes change when replaying the move, suggesting that the position has been calculated deeper, necessitating a change in evaluation.