Can the elo fluctuations depending on the move randomize?

[hsntgm]

It is sad to see the project slow down.I just try to help by offering possibilities within my knowledge.

Bestmove: Picks the move with the most visits. randomize or tempdecay options allow picking lower moves for variety.

If the reason is corrupted training data (from ID 253 to ID 287) could one of these reasons be a move randomizer? I don't know how works this parameter in background or is it tuneable with weights (e.g. cutting the unnecessary move trees and rewrite the most logicals one) in code or not but it seems affecting training data deeply.Picking lower moves for variety brings trillions of new possibilities and growing unnecessary move trees maybe affect the quality of training data and needed number of games with self play badly.