Engines for Human OTB analysis

[Rowen]

Some sources seem to suggest that engines such as Stockfish due to their style, strength and counter intuitive play are not ideal to analyse a human/human game? if this is so which engines are best? Would it also follow that engines best suited for human analysis would be best to play/train against?

[Fritz 0]

Personally, I understand Komodo pre NNUE versions better than Stockfish. Of course, don't know how much is this relevant from a 1900 ELO FIDE player.

[Rowen]

Personally, I understand Komodo pre NNUE versions better than Stockfish. Of course, don't know how much is this relevant from a 1900 ELO FIDE player.

Well I am just 1650 Fide. Its not just the strength of Stockfish etc, its their 'unhuman' conclusions. I guess i have been influenced by reading about the Alexander (shashchess) engine. While analysing some of my games I can get quite varied analysis by different engines, particularly with / without NNUE. But not just about strength, Perhaps its that old issue about getting engines that think in a human way.

[gordonr]

I always use multi-PV analysis (maybe top 5 lines or so) and pick the move that I understand best that still has a decent evaluation. Often e.g. the 3rd best move makes most sense to me while not that inferior to the best move

- if the top move is indeed much better than the rest, you may want to explore and compare the lines to see if you can find out why
- for smaller differences in eval, don't get too hung up on engine evals. There are many human playable positions that top engines don't like
- it is indeed a tricky balance between trying to learn a bit from the engine when possible, while being prepared to say "I don't understand that line and I never will... let's analyse another move closer to my ability"