Alternative and common approaches in the construction of TC - page 4

 
Azerus:

Then why did you start this thread?


An example:

You have two FIs and one TS. You test your TS on FI1, optimize it on different parts of the history, look at the results on Out of Sample, etc. Find the optimum (from your point of view) indices of your TS for Instruments of FI1. You also do the same for Fi2. The TC results on FI2 turn out to be much better than on FI1. Where will you run your TS?

You can diversify your risk by running it on both FIs. But you will still do it with different TC weights for FI1 and FI2. Because the results of FI2 are much better.

The bottom line is that based on your results on FI1 and FI2, the TC is more adapted to FI2. This says nothing about how the TC will behave next, of course.

The above describes the way most people do.

The more "advanced" ones start still generating synthetics (and there are a lot of them) and run their TC on them. This is very resource-intensive, but also an option (own tester).

The alternative option implies quite a different approach. You think about what financial instrument must be, to make your TS work the way you want it to. After you have thought it over, you start thinking, how to create such synthetic instrument out of those FI which are available. Having created such a synthetic, you will perform the same optimization operations on us. The results of optimization, Out of Sample and all the other tricks aimed to get rid of the fitting will be better than those of the original instrumentation. What FI will you run your TC on?

 
It's like they created a TS for the pound, then matched it to the euro, the results were better and decided to run it on the euro pound in general, having previously matched it to the euro pound's history. Have you got it right?
 
sanyooooook:
It's like they created a TS for the pound, then matched it to the euro, the results were better and decided to run it on the euro pound in general, having previously matched it to the euro pound's history. Have you got it right?

Wrong.
 
hrenfx:

Wrong.
Yep, ughhhhhhhhhhhhhhhhh, replied )))))
 
Svinozavr: In fact, you suggest the following scheme: EA --- function of creating a synthetic trading tool from the market mn --- market. In terms of form, it is a multi-currency instrument.

The essence of the idea (as I understand it) is to isolate the logical synthesis block of a trading tool. In my opinion, it is a sensible idea. Respect.

Yeah, I like it too. Now, it would be nice to do something practical. It's not obvious that the result would be better than the traditional approach, but it's still a curious idea.
 
hrenfx:


.......

The alternative implies a very different approach. You think about what the financial tool should be to make your TS work the way you want it to. After you have thought it over, you start thinking about how to create such a synthetic from those FI that are available. Having created such a synthetic, you will perform the same optimization operations on us. The results of optimization, Out of Sample and all the other tricks designed to at least somehow get rid of the fitting will be better than on the original FI.


Again, the alternative you propose should imply an obvious stability of the synthetic's future behaviour relative to the shown history compared to the existing tools.

If we say that the pound may do something funny in the future via yen (GBPJPY), and therefore it is profitable, but risky to trade on it, then let us trade wheat via spot gold (synthetic), because the behavior of synthetic will be very stable.... The question is: Why the hell should synthetic be able to guarantee us such stability? Clearly, no one will guarantee anything to anyone..... And if synthetic does not guarantee us stability of its behavior in the future, why do we need it?

 
Mathemat:
Yeah, I like it too. Now that would be a good idea to do something practical. It's not obvious that the result will be better than the traditional approach, but it's still a curious idea.

I posted an example of a dynamically changing synthetic in CodeBase (not advertising - for those who start accusing). There are a lot of variants.

It is a funny situation with martin - all sorts of tricks are invented to defeat EURUSD or GBPJPY. Why these pairs and not others? And after all, why not synthetics, of which there are far more "real" FI? And finally, why not at least answer yourself the question of what should be the optimal FI for trading this popular martin?

P.S. An example off-topic, but it also relates to the practical application of simple synthetics.

 
Azerus:
Read the first post on this page again and answer yourself the question at the end.
 
hrenfx:

Common:

FI1 (financial instrument) is taken and patterns are sought in it. On the basis of these patterns TC1 is created.
Then FI2 is taken and TC2 is created
the
same way.

Alternative:

A TC is invented. It is investigated what properties an FI must have in order for the TC to work stably.
After the properties are determined, opportunities to create a synthetic (artificial FI) from the set of FI available for trading are sought.
A common method just invented by you. There is 1) TS - trend, 2) trend is looked for on the charts and traded. It may be the same with a flat TS.
 
hrenfx:

Any TS will (in theory) fail! Whether it is Martin or something else is irrelevant.

As long as you know the distributions of the non-trending movements of the financial instrument over 20 years. And this is easy to find out by analyzing the history. Then taking into account this distribution you can create a Martin that will not fail on these 20 years of history.


Do you have any results?