Pair trading and multicurrency arbitrage. The showdown. - page 81

 
Maxim Kuznetsov #:

n1. in daytrips

you don't need to predict it, you just need to know it.

PS/ by the way, on the screen you can clearly see the "two-headedness" of the distribution. Or how quoters differ from the chaotic SB

What is this indicator?

 
Sergey Mitrofanov #:

What's that indicator?

It's me twiddling my thumbs over the "convergence of currencies to a single scale." I'm studying relationships and dynamics.

if you read the thread, there is about how it was made through Yoxel
and the initial version of the indicator together with the source.

On the screenshot it is not the initial one at all, but I don't like to publish it :-) not because it is a secret, but just the software itself needs to be redesigned.

There are a tonne of crutches and patches, everything was corrected almost in real time from fresh ideas.

 
Maxim Kuznetsov #:

I'm the one who's been tinkering around with the "convergence of currencies to a single scale." I'm studying the relationships and dynamics

if you read the thread, there is about how
was made via Yoxel and the initial version of the indicator together with the source.

On the screen is not the initial one, but I don't like to publish it :-) not because it's a secret, but just the software itself needs to be redesigned.

There are a tonne of crutches and patches, everything was corrected almost in real time from fresh ideas.

So what is the complexity of the indicator and why is there a ton of crutches?

It's just normalised prices of different currencies, 10 lines of code in a normal language.
 
mytarmailS #:
So what's the complexity of the indicator and why are there a tonne of crutches?

It's just normalised prices of different currencies, 10 lines of code in a normal language.

There should have been a lot of off-topics here, which would have been deleted anyway, and even with a ban.

But there isn't :-)) and there are a ton of crutches.

 

so that the branch does not go to waste :

since all currencies with denomination x cobelieve as ~ln(x) we can try to construct a basket (someone's favourite) triangle to start with.

maybe try to find a third solution other than the degenerate "nobody anything at all" and "all to one" solutions

PS/ (added a bit after publication) it is worth adding what exactly we want from the triangle. Without boundary conditions it seems to be unsolvable. Or outline the area of boundary conditions (fantasies can be from and to and edges are mutually bounded).

 
Maxim Kuznetsov #:

so the branch doesn't go to waste:

If by triangle you mean 3 valpairs, then, IMHO, it is not worth it. The more pairs - the more spread you will give and the more swap you will pay.

I think you need to think about the entry condition.

 
trampampam #:

If by triangle you mean 3 valpairs, then, IMHO, it is not worth it. The more pairs - the more spread you will give and the more swap you will pay.

I think you need to think about the entry condition.

first of all, analytics, that is, to figure out what it is.

and then methods of inputs, outputs.

about triangles, in the examples we decide to trade USDJPY vs USDCAD(JPY vs CAD).

We open 1 lot on CADJPY.

We are surprised to see that the profit/loss on the trade strongly depends on one leverage. (No less surprised people have been calculating correlations for several years).

Why? incorrectly calculated volume - USDJPY and USDCAD fluctuate with the same %, but the price of interest is different.

Somehow we recalculate volumes, we get different shoulder lots for USDJPY and USDCAD, but we really trade JPY against CAD.

But the shoulder lots are different, the dollar is out - we got a triangle (basket ) JPY-CAD-(USD).

 
Maxim Kuznetsov #:

analyse it first, that is, to find out what it is.

and then the methods of inputs, outputs.

Okay, analyse:

Let's say we want to take USDJPY against USDCAD "at random". First, we should check how currencies move against each other. What period are we looking at? A week, a month, a year? Let's say we take a month and a year:

It is immediately obvious that the currencies are poorly correlated with each other. Question: Are they suitable for pair trading?

We can, of course, then try to trade them for divergence rather than convergence.....
Files:
CADJPYH4.png  50 kb
 
trampampam #:

Okay, analysing:

Let's say we want to take USDJPY against USDCAD "at random". First, we should check how currencies move against each other. What period are we looking at? A week, a month, a year? Let's say we take a month and a year:

It is immediately obvious that the currencies are poorly correlated with each other. Question: Are they suitable for pair trading?

Surely we can then try to trade them for divergence rather than convergence....

Above on the topic was "all, any bundles diverge". Increasing the time horizon will come to a disappointing 3-5% per year

 

for those looking for immediate entry points, as a hint: