For those who have (are) seriously engaged in co-movement analysis of financial instruments (> 2) - page 7

 
lea:


The facts are there. You need a synthetic instrument to make some economic sense (the sum of all pairs is what? and if I add up the prices of 40000 instruments?)

And the facts are in the most prominent places. Instead of thinking it's better to take it for granted, right, sanyooooook?

Take two oils, make up a synthetic, narrow channel, everything eats the spread, there is a brokerage company that has zero spread on oil, but the commission, make an affiliate and let them pay you back part of the commission, then the system is in the black.

 
Reshetov:

I don't know why we need a synthetic that forms a sideways trend with a channel?


You run a channel strategy and you're in the black. Is it bad if you know that the synthetic will not go far outside this channel and then sell at the upper boundary and buy at the lower one, the main thing is that the channel should be at least equal to the total spread, so you won't be in the black.
 
sanyooooook:
If you run a channel strategy, you're in the black. It would be bad if you knew that the blue print would not go beyond this channel and then you sell at the upper limit and buy at the lower one,
But what good is it if these bounds do not exceed the spread or commission, and even if they exceed us by two pips, while we are turning, the price will go to the loss (because it will not go out of the channel, and therefore it will only go in the direction of the loss). I'm not even talking about slippage. The price of one chart does not always have enough time to trade, but here the entire portfolio must be handled with the dexterity of a pipser.
 
Reshetov:
And what good is it if these limits do not exceed the spread or commission, and even if they do exceed it by a couple of pips, then while we are turning, the price will go to a loss (because it will not go outside the channel, so it will only go to a loss). I'm not even talking about slippage. The price of one chart does not always have enough time to trade, but here the entire portfolio must be handled with the dexterity of a pipser.
If you use a terminal other than MT4 and connect through PLAZA2, the speed increases significantly.
 
Mathemat:

The topic (precisely at >2) is unbounded.

It is simply a generalisation of the two-dimensional case. And this generalisation is generally available in the dummy case:

In the screenshot (below) a dummy, which is designed for multidimensional cases, but applied in this case only to two FI. You can see how it agrees with the correlation. So to speak of a mathematically correct generalisation of correlation to the multivariate case I take it completely.

Correlations are fine. It would be better to deal with cointegration theoretically (the cointegrating vector is elementary for three pairs, tied in a ring).


There is no cointegrating vector for majors (no rigid functional connections, like "rings" for example). This begs the question, is it possible to do without it?
Of course we can! The main thing for us is to keep the synthetic property (horizontal channel) at least slightly to the right of the interval of its construction - on Out of Sample.
The question arises, why it should be saved? We can build a synthetic under any curve, but it doesn't mean that arbitrarily chosen curve will also continue on OOS.

And that's where the main idea of building a synthetic comes in. If a synthetic is built using market relationships (the relationship between all FIs), then the probability of preserving the properties of the synthetic is maximized.

That is, how does a real market differ from a pseudo-market - all FIs are random wanderings? By the fact that in a real market there are correlations, while in a random market there are no correlations. It is this property of the real market that the synthetician has to exploit. And that is the basis of the claim that the probability of the synthetic's properties being preserved will be as high as possible.

It is another matter that the above craft for the multivariate case is a generalisation of the two-dimensional case (correlation). But correlation is not the only correct method to characterise the degree of a linear relationship. This method is based on the method of least squares (OLS). ANC is a very convenient and popular method, but it is not the only one...
 

All you are talking about now is an attempt to exploit standard arbitrage. I didn't raise the question of it, I raised the question of statarbitrage:

Главное - попытаться "правильно расшатать" коинтегрирующий вектор. Вот куда его расшатывать, с какими критериями, - это вопрос вопросов.

Here's a chance to work steadily.

P.S. hrenfx, just now saw your post. I am aware of it.

And this is where the main point of building a synthetic comes in. If a synthetic is built using market interconnections (interconnections between all FIs), then the probability of the synthetic retaining its properties is maximal.

I.e. how does the real market differ from the pseudo-market - all the FI are random walks? By the fact that in a real market there are correlations, while in a random market there are no correlations. It is this property of the real market that the synthetician has to exploit. And that is the basis of the claim that the probability of the synthetic's properties being preserved will be as high as possible.

Yep, close to what I'm about to do myself; that's why I'm considering the ring.

 
If there is a ring, it is at once a fake (FI names are not used in any way):

And if you look at the Variance parameter, which is almost zero (and from the previous post you can see the relationship Variance = 1 - "Correlation"), it is clear (almost maximum correlation) that we cannot earn anything from it. The channel is too narrow(Variance - its variance), the overhead (spread + commission) will eat everything up. And this is understandable as with a "ring" we have pure arbitrage. Which, as we know from some studies, has almost no place in the FOREX market.

Therefore, we safely throw out all the crosses, as they do not carry any additional information, and use only the majors.

 
sanyooooook:

Take two oils, make up a synthetic, narrow channel, everything eats up the spread, there is a brokerage company that has zero spread on oil, but the commission, make an affiliate and let them pay you back part of the commission, then the system is in the black.



As for the affiliate - it is more profitable to conduct seminars on trading :)

It is possible to create a synthetic that fluctuates more than the total cost. But for such strategies the speed of receiving data and order execution is important.

 
lea:


About the affiliate - it is more profitable to run trading seminars :)

It is possible to create a synthetic that fluctuates more than the total cost. But for such strategies, the speed of receiving data and executing orders is important.

No one is arguing that speed is important in this business, but I don't understand about the affiliate, justify it.
 
lea:

It is possible to create a synthetic that fluctuates more than the total cost. But for such strategies, the speed of receiving data and executing orders is important.

Justify.