This is a nice idea but does not exploit the information inside the correlation of [eval N ....eval N-M] set.Rebel wrote: One idea I have experimented is counting "clear advantages" and then add a bonus to score based on the number and weights.
IOW in your idea there is no reference to the evaluation at previous ID loops and a way to correlate with current evaluation at ply N. You introduce additional chess knowledge while I would like something chess knowledge independent and based only on the series of root search scores we get while running the iterative deepening loop.