TT, I don't mind. Can we agree on the assumptions to be taken though ...
Let's take a reference, the figure you just posted:
--------
-> Q1. Where do we consider the turns as hitting the level/vs anywhere? The definition of "close enough"
- in the picture there are some turns around pivot (or "close" enough)
- but at least that many above and below pivots/ and between pivot and next level (above the tolerance level of "close enough").
Visually I would say there are at least twice as many turns outside of the pivot level above and below the close-enough, and not in the close-enough vecinity of the next level up/down...
-> Q2. What is a turn? The size of it, and its duration in time ...
- counting turns depend on what we consider a "turn".
0.25, 0.5, 1 point swing? or any single wiggle visible? The bigger the "turn" size for qualification - the more noise eliminated.
When the jump between levels is 3 points - what should be a turn/price rejection.
We need the size of the turn on the vertical, and the time (how many bars below/above that level. E.g. on a 1min chart - is a price swing inside a single bar considered a positive? )
-> Q3. Which time frame shall we consider?
- I don't know the time frame you consider here - but a 1 min chart will look different than a 15/30, or 240 min chart where you only have an aggregated candle crossing the said levels and no turn "visible".
- so which granularity shall we consider, for counting the positive/negatives?
Any comments on the above?
Market Data 
Markets 
AI
