Once again, whether OEPM is, or not, holding any price prediction value - this is not putting any dent in NN ability as a trader/ his results are all posted/ and impressive.
I dare to send my comments in this context -
(1) I challenge someone to use 1st day of the month, or the 2nd Fri, rather than the 3rd Fri as a base of these same S/R lines... see for themselves if any difference predicting turns.
Say we define "results" == "% of time the price was rejected by a s/r line calculated based on 1st of the month"... Then, by how much the "results" will differ?
(2) reminder : 1.25% of SPY is +-3 points, right?
When we define "the positives" (the price did turn at the Sx/Rx level), what is the tolerance for considering that turn, re. that line?
(2.a) Is a turn at 2 point delta from the levels, considered 'close enough'? or only 1 point delta?
Needless to say, if distance between neighbor lines is 3 points (1.25%), and the tolerance is 1.5 points - this is not "a system"/is it? There are only positives...
(2.b) what shall be considered "a turn", vs "noise"?
Is a 1 point turn, a check in the box, or do we need 2, or 3 points+ to qualify for a positive?
(2.c) don't forget the negatives... counting similar turns at (2.b) that happened between OEPM levels.
---
So unless you consider some of these points in the "system" evaluation over time, some skeptics might say the claims might be subjectives/rather than factuals
/jet
For these interested by the topic (price prediction), there are time series analysis of stock prices that published here and there, but also more sophisticated ones (multiple discriminants/MDA).
The difference is the first one considers only time as the variable ( e.g. 3rd Fri of the month influences the price turn, at some %-based delta lines).
The MDA considers other variables for finding correlations (Tue/week, FED meeting, etc).
Whatever the analysis - it's a common practice to compare findings to white noise (random values), before claiming any correlation.
Anyways, after all, the interesting bit is the methodology being used (definition of positive/negatives, avoid data over-fitting, etc)
The boyz have spent significant amount of money of them/price studies, and of course one can ignore them/findings as being nonsense.
... as I am already far down sliding on a banana skin ...
For the price predictions considerations related to SPX, many seem to conveniently overlook it's an index calculated on 500 stocks, every 15 sec.
Moreover, there is no bid and ask on the spot, and only futures are being traded.
You have a bunch of computers trading the arbitrage, eagerly waiting for an imbalance to appear between the said stocks (fair price) and what the futures say it's a fair price ...
For the price considerations related to SPY, there is an intraday bid/ask (so the boys do take positions), there are option players, etc - however overnight SPY re-balances the underlying basket and gets in line to SPX.
And again a bunch of players trade the arbitrage, knowing the little trick.
Now, imagine the above mechanisms (OEPM based turns), at some static levels based on past month OE as reference ....
By definition, given the dependency on the basket of stocks SPX, you have 500 stocks that would need to turn, and to stop/or reverse at some level....
And by definition, given the dependency on SPX, SPY will follow course for whatever exceeds one day.
If the 'by definition' is not respected - the boyz get creamed big time, is that simple.
Market Data 
Markets 
AI
