If you want to improve your model - I think continuously updating your model as the day/week progress should help. This makes sense as any prediction should be based on what you know now at the time of the prediction. As time progresses you should be able to confirm if the market is tracking your prediction and make adjustments if not. Simple example: By EOD I know that tomorrow will either be A, B or C. By the Open, I can eliminate C. After the initial 30 minutes, I can tell with a high degree of certainty that A is the play for today and as long as certain conditions are in effect, A it is. I know what needs to happen for scenario A to be negated. At least this is how I do it. Then you have those days where the market seems more like a coin toss and a particular scenario is less certain. Not sure if your model addresses that somehow or if you weigh each particular predicted outcome equally.
Two components were removed to do the 3X3 forecast and it made remarkable improvements. So any of the previous charts before the 3X3 forecast can be disregarded. Also the forecast does not measure magnitude, sometimes it shows the magnitude, other times it will not at all but that does not take away from the information provided. That is very important to know and realize when looking at the graph of the forecast. It is a map of the price path. Any time you start adding conditions you get in to curve fitting and it will not improve your results. The 3x3 forecasts are not optimized at all.
Here is what the 3x3 forecast on the ES Daily would have looked like if was done at the end of September. Keep in mind the 3x3 is displaced by 3 days. Even with those considerations this can followed with a simple moving average or the displaced 3x3 moving average for entries and exits. You have to keep it simple. Also there is no Gann or Astrology Fibonacci Elliot Wave etc., just pure binary mathematics with a non-linear structure.
Gold Daily for the next 3 months...Next weekend after the Christmas Holiday I will begin doing the weekly forecast for the entire year (2021) along with the new Daily forecast for the new year first quarter. I will continue with the intraday on a daily basis.
I go back and forth on placing the polynomial and regression slope with the data. As you know it's just a "fit", but there times that it is extremely accurate and also can provide you information at the inflection points of the peaks and valleys of the polynomial.
Here is one of my favorite markets to position trade...Coffee! I scaled the chart to get hopefully a better view of the data points.
Crude Oil is anything but flat with the sharp decline occurring immediately...This chart could be inverted with the low and reversal at 2100. Since I won't oppose the 3x3 it always keeps me out of trouble.
Crude hits the Daily 3x3 which is at 47.49...the market went as low as 47.43 and so far has reversed during the 2100 30Min bar...That would have been a low risk entry considering the market is at 47.86 as I write this. I am not opposed to taking trades off a 3x3 ass long asI view it as a low risk entry. In this situation the higher time frame came into play.