Once again, arbitrage, pair trading. - page 9

New comment
 
Maxim Dmitrievsky:

I've done the same... I just wish I could find a way to determine the significance of each attribute and form an optimal portfolio on the fly, discarding the bad pairs.


There is no such thing as an optimal portfolio, you just have to accept it. And there are no bad pairs. No analysis, no mat-formulas can guess or find the "preponderance" in our favour.

 
Anatolii Zainchkovskii:

There's no such thing as optimum! And there are no bad pairs. No analysis, no matformulas can guess or find the "edge" in our favour.


Yes, today it's optimal and tomorrow it's goodbye.)

 

)) true )) what is more interesting is another feature of highly correlated forex instruments ) the higher the correlation (you can take 25-30 bars for calculation) the stronger the tie will be broken ) What to do with it I think it does not take much intelligence ......

 
Maxim Dmitrievsky:

Yeah, today it's optimal and tomorrow it's goodbye.)

Oh...

I agree 100%

The quoteer will learn and say "Adieu, it didn't work out, sir, it didn't work out ...".

 

I'll do a neural network later, it's interesting to see what happens and compare )

 
Maxim Dmitrievsky:

I'll use a neural network later, it will be interesting to see what happens and compare)


You'd better use the neural network to do a Fourier phase search of the trading system, and if you can find a pattern .....

 
Anatolii Zainchkovskii:

You better use a neural network to do a Fourier phase search of the trading system, and if you can find a pattern .....


there is a better Fourier - empirical mode decomposition

I already did it that way but in a slightly different way - only one pair traded depending on several others... the results are so-so at a first approximation

 

By the way, on the one-minute chart you can already see where and which to trade....

 
Maxim Dmitrievsky:

I will do it later with a neural network, it will be interesting to see what happens and compare )

And please look now audchf / nzdjpy

 
Maxim Dmitrievsky:

there's a better Fourier - empirical mode decomposition


even better ) only I'll probably upset you early on and it won't do anything)

12345678910111213141516...20
New comment