You want to pick more than 50000 but Lizzie doesn’t let you pick more than 50000? That seems silly, there’s no good reason for it to limit it like that. If you were happy with 50000 with 4 threads, then obviously better would be, e.g. 80000 with 24 threads - more threads runs faster, and then the fasterness lets you do an even bigger numbers in the same time.
What contradicting statements are you referring to? All the places I’m in control of should hopefully be saying the above tradeoff - more playouts is better but fewer threads is better holding playouts fixed (so you pick the threads that best balances playouts per second with the slight weakening due to threads). And not simply saying just that one way or the other is always better. If there’s a place that does, I want to fix it.
Also maybe you shouldn’t worry too much about the “passive” scan-over-the-whole-game analysis anyways. While it can be useful to see where the hot spots are, you probably want to spend much of the review time asking questions interactively.
Unless you’re high-dan-level (and even then!) a large amount of what top bots will suggest will be way over your head if you look at it passively, which is not very useful for learning - just seeing the moves and not understanding why is hardly better than just spectating uncommented pro games or whatever.
But, a lot of it can become more understandable if you interact with the bot. Play out the sequences that are obviously wrong to the bot but that aren’t obviously wrong to you, and see how the bot punishes them and until you understand why a move is a mistake. Interactive analysis massively dominates passive analysis in a lot of cases, for learning. See:
https://lifein19x19.com/viewtopic.php?p=255703#p255703
And 24 threads (or whatever number benchmark found to be optimal) will help with the bot running fast enough for real-time interactive analysis.