The Correlation Game
Correlation trading & correlation strategies in the Forex market and inter-market analysis in general
Best correlation from indicators to the price
Well, the result of this study shows : (except perceptrons) - I checked also on USDJPY/M5
- HP_oscillator
- DPO
- FX_Sniper_s_Ergodic_CCI_Trigger
- WPR
- EntropyMath
- PFE
- DeMarker
This implies a new perceptron with 4 neurones (HP_oscillator,DPO, FX_Snipper_Ergodic_CCI_Trigger,WPR)
Nota: If there were an idea to create a double perceptron (Slow and Fast-14), the HP_oscillator provides the same values. DPO w/ 37 is better than higher value like 51 i.e, same for FX_Snipper_Ergodic_Trigger. WPR is more stable for 37 than 51 as period, same for Entropy... So , I made another display with a value for correlation period as 6 for tests
Comments
I saw this Inverse transform Fisher (my improvment put here for friends) belongs to this list
Before playing w/ genetic algorythm, the next step is to create a logic based on the indicators but, selected upon pearson correlation's results
This study is really very interesting stuff, and I like it a lot. I wonder if a spinal implant based on the better performing indicators here is going to perform better.
As for me my empirical experience is that is really hard to know what preprocessing tool is going to work in the future. So in that way I prefer to choose the tool I understand and the tool that is faster to computhe with the genetic optmimizer.
So even if Hodrick Prescott is really the top I may not use it because I do not fully understand the calculations inside of it.b But this is just me.
Yes, I just improved the HP_oscillator, as I made the FisherTransformHP
In the correlation game, I was thinking about the input value, to take instead the value from the Hodrick Prescott, instead of the Close . The value from the Close of a bar is universally recognized today as a good input, but , in my humble opinion, be replaced by the HP_oscillator, because it's a refined Typical_PRICE
Also, now, in the file, I just uploaded in private, the next step is to build a perceptron w/ the selected indicators (choosen for this market phase, pair and timeframe)
and the last step would be to add a genetic alogorithm inside the init phase, to set up the best parameters for the selected indicators.
Of course, the GA will be redesigned after to be applied to a bunch of those correlated to price 's indicators. But I think, we have far reached the limits of MQ4, and the processing inside the Init phase with GA applied to the 8 indicators would take at least half a day of processing
Now, check too the both Spearman and Pearson correlation of those indicators above