When are inaccuracies acceptable?

In my most recent game, the computer gave me having 35 excellent, 10 good, 8 inaccuracies, 4 mistakes, and 2 blunders (I think). All of them except one inaccuracy came from a piece up position, so is it okay? The blunders were dumb mistakes, but neither really changed that much about the position.

I think playing a few inaccuracies is ok, maybe to complicate the game for your opponent or to simplify a winning position.

As an example, some gambits are considered inaccuracies, also exchanging a queen and two pawns for a queen to have an easy ending while a rook up.

Never

Actually, the only type of inaccuracies that are completely fine are practical moves, where the goal is to remain in control of the position or keep it simple. Practical moves can happen when you're +3 and completely winning, and also when you're +0.3 and just trying to push your advantage, often when your opponent has sacrificed material. 

 

Of course, making positional mistakes when you're up a piece doesn't hurt you at all.

 It's fine unless your opponent would have been able to get compensation for the piece (or was able to win it back).

The only computer analysis I have access to is C.c's post game and I find that the blunders it identifies always result is a substantial change in the evaluation of the game. But inaccuracies, not always. If I play a decent move that I understand how it leads to a win or just a better position, it is 'interesting' to see what alternatives the analysis offers that could have left me with a slightly quicker win or slightly better position but in the end, inaccuracies aren't always game-changing.

When they are really interesting though, is when you or your opponent are completely outplayed, even though one may not have made any 'mistakes' or blunders. When the position just gets incrementally worse over the course of the game, It's that collection of inaccuracies over the game that lead to the loss. I think those games are the most instructive, and the most fun- win or lose.

To clarify my "never" above...

I don't consider something an inaccuracy based solely on a computer eval, though that factors in.

If the computer shows a crazy long computer line as +3 but the human looking move you play is "only" +2 I don't think of that as an inaccuracy.

I'm mainly basing this off of a computer eval since I don't have a human coach, so I'm trying to understand when my electronic "coach" just needs to get off my *ss

The assessment of the online analysis is often very questionable. I remember a game where I finally simplified a won double rook endgame with two paws up to a trivially won pawn endgame with just one pawn up, which was clearly the best practical choice. The engine classified this as a blunder, while the opponent just resigned.