Not the Grail, just a regular one - Bablokos!!! - page 60

New comment
 
faa1947:

Pair trading uses noodles, So what?


That explains your lack of results in trading.
 
Avals:


Take the difference between eurusd/gbpusd and eurgbp and you will see the cointegration. That doesn't mean you can make money on it because of the overhead.

But in most cases the cointegration is temporary (seasonality for example).

Any way you look at it, mean reversion systems for pair trading and statistical arbitrage try to use cointegration


No matter how you look at it, temporary cointegration is not cointegration. Co-integrated series ALWAYS converge.
 
Demi:

Learn the math - the series studied for correlation must be normally distributed. On the question of the difference between normality and stationarity - read what you wrote in the other thread. You were even given an example of non-stationary series with a normal distribution and vice versa.

So where is the calculation?

 
faa1947:

So where is the calculation?


what is it?
 
Demi:

No matter how you look at it, temporal cointegration is not cointegration. Co-integrated series ALWAYS converge.
I don't care if you call it undercointegration as long as it makes money)) It is clear that this mathematical abstraction that rarely occurs in reality in an ideal form. Although it does occur and a couple of examples were given on the previous page.
 
Avals:


Take the difference between eurusd/gbpusd versus eurgbp and you will see the cointegration. This does not mean that you can make money on it because of the overheads.

But in most cases the cointegration is temporary (seasonality for example).

No matter how you look at it, mean reversion systems for pair trading and statistical arbitrage try to use cointegration

Why. Up to 40 pips. I think the problem is different.

We enter by the deviation from the mean in the residual from the cointegration regression. See here.

It turns out that we decide on the position regardless of the kotir.

 
here's hrenfx with Recycle - nothing more than a homemade co-integration search on a sliding window
 
faa1947:

Why. Up to 40 pips. I see a different problem.

The inputs are led by the deviation from the average in the residual from the co-integration regression. See here.

It turns out that the decision on the position is made regardless of the quote.


It can't be 40 pips. Apparently on the spread spread spreading I counted without taking into account the real bid/ask
 
Demi:

Which one?
Co-integration, at least one. After all, you claimed above that there is no such thing. So prove that there isn't. You've done the math here and posted it. Come on, match it.
 
Avals:
here's hrenfx with Recycle - nothing more than a homemade co-integration search on a sliding window

and I thought the calculation was based on spearman's correlation coefficient)))
1...535455565758596061626364656667...650
New comment