Setting up enough time and sensitivity for a chess analysis

If you let the chess engine run too long, it just takes a lot of time. Except in simple or forced situations, it is not likely that both sides will play exactly as the suggested line more than only a few moves forward. So although the suggested moves are interesting to go through, there is no reason to run the analysis for too long on all moves.

In the same way, if the analysis is too shallow, the advice given, even on the first move itself, can be wrong. For most situations, minimum around 20-22 moves is a good depth for getting a suggestion from the chess engine to use. In complicated setups, this may take quite a time to go through for all moves. Thus I normally do the initial analysis based on time. With my computer, using around 7 CPUs, 30 seconds is enough for many moves, but I tend to want to set it to 120 seconds. If I can leave the machine on for a long time (going to bed, being at work), 150 seconds is a good setting. It most often produces deeper analysis then I feel necessary, but since I am not waiting while it works, it is anyway a good setting.

The sensitivity threshold should be low rather than high

The sensitivity setting (threshold in Chessbase and Chessok programs ) determines how often the chess engine will create a suggested alternative approach. The program thinks of it as the threshold for detecting a blunder, but that is not how I use it. My interest is that when the analysis is done, I can easily see where alternative moves existed, which they were, and then be able to understand why they could have been better. When I start looking at the game, I do not want to sit there and wait for an analysis on every move I find interesting. Thus I want the alternatives written down in the initial analysis. I use 20 (which means 0.2 pawns difference in evaluation). Chessok has 30 as default, Chessbase has 60 as default. In Chessbase, you have to manually make a tick mark in the dialog for evaluations to be stored after blunder-check is run, which is the whole idea with this step.

Let the chess engine evaluate both sides

It should be possible to ignore doing this. But I am in the belief that the chess engine does a better evaluation if this is turned on, so I have it on.