Discussion of article "Filtering Signals Based on Statistical Data of Price Correlation"

New comment
 

New article Filtering Signals Based on Statistical Data of Price Correlation is published:

Is there any correlation between the past price behavior and its future trends? Why does the price repeat today the character of its previous day movement? Can the statistics be used to forecast the price dynamics? There is an answer, and it is positive. If you have any doubt, then this article is for you. I'll tell how to create a working filter for a trading system in MQL5, revealing an interesting pattern in price changes.

Author: Тарачков Михаил

 

Interesting article. I myself once made similar calculations on mt4.

Question to the author: you did the tests outside the statistics period.

In your short table the results are far ahead of the long table, which may indicate that the statistics was collected on the trend period.

Therefore, it would be good to do a back-test, otherwise such impressive results look like a fitting.

 
I liked the article. Especially interesting is the TDW strategy itself and all possible places of realisation (there were certain ideas in this direction).
 
Urain:

Interesting article. I myself once made similar calculations on mt4.

Question to the author: you did the tests outside the statistics period.

In your short table the results are far ahead of the long table, which may indicate that the statistics was collected on the trend period.

Therefore, it would be good to do a back-test, otherwise such impressive results look like a fitting.


The period of statistical research is the last 10 years. (otherwise, what kind of statistics is it!).

The tests were conducted for the last year on EUR/USD pair. This time turned out to be extremely rich in trends. However, most of the time the rate was falling. This explains the superiority of shorts over longs.

You cannot go far without optimisation. By the way, I carried it out on the period from 01.01.2010 to 31.12.2010. Markets change, the system parameters should follow this trend strictly, otherwise there is a risk of losing the deposit without trading.

Here is a picture of the back-test for the last decade with the same parameters:

Back test

It was draining until 2004. Then it seemed to grow for six years. I would like to emphasise that the Expert Advisor is a trend-follower, and flat is as destructive for him as a storm is for a fishing boat.

 
And here I am curious to see this idea in the mullet (then as I think some of the fitting issues are gone).
 
Interesting:
And I'm interested to see this idea in a mullet (then as I think some of the fitting issues are gone).

You can. And if we compare the movements of currencies - for example, the pound grows against the dollar, so we buy EUR/USD. And vice versa.
 
Urain:

That's why it's a good idea to do a back-test, otherwise such impressive results look like a fit.


Unfortunately, this is exactly the case... optimising for time (hours, minutes, day of the week) of input, is actually no better than for any other parameters: at least on OOS it generally works just as badly)).

Filters, filters ... eh, I don't like this word. Here's the thing. If we have a certain TS, it does not matter, even if it is unprofitable (or profitable, "but not very much", otherwise why do we need a filter:)), and a certain filter, which gives improvement of the system's indicators, then it means first of all one thing - that we can throw out our old TS with peace of mind and use our wonderful filter as a trading system (because, to tell the truth, it is he who is responsible for the profitability of "filtered" signals:)). Many people who "almost finished their system" and who "only need to pick up the necessary filter of signals" do not even think about the fact that they go round in circles, simply substituting concepts and self-deception....

So here goes. A good filter is no worse than a good TS, mmm....

 

alsu:

...it means first of all one thing - that we can throw out our old TS and use our wonderful filter as a trading system (because, to tell the truth, it is responsible for the profitability of the "filtered" signals:)...).

Not quite the right statement, in my opinion.

alsu:

So. A good filter is not worse than a good TS, mmm....

I agree with that.
 
Interesting:
Not quite the right statement, in my opinion.
I reread it this morning and it does sound a bit strange)) I was a bit too much, I guess)))
 
alsu:
...

So there you go. A good filter is as good as a good TC, yep.....

Actually, there's not just one filter in the article, there's 5. Monday filter, Tuesday filter, etc.

It's just that these filters are applied head-on. But I think the author did not set the task to develop a strategy, but just wanted to show the importance of filters.

In fact, each filter can hide a different strategy. On Monday work on one TS on Tuesday on another.

 

Another positive feature of filtering is that at 4-fold reduction of trades the profit drops by only 25%

It is not a secret that every entry into the market is a risk. And the winning strategy in coin is not to play at all.

And the fact that you can make much fewer trades with almost the same profit is good, very good.

12
New comment