Discussion of article "Exploring Seasonal Patterns of Financial Time Series with Boxplot" - page 9
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Just admit that I've found a useful pattern, the rest is philosophising.
No, unfortunately. You have to be honest, first of all, to yourself.
You made a research, saw that you could squeeze profit out of it and wrote a TS, which bends at a certain area. Then you saw that everything sucks statistically on this area and concluded that this is the reason why it doesn't work there. Putting the cart before the horse.
Here I did a study on a blog by running just a script. I don't need any optimiser to create a TC with a profit on the study interval on the basis of this research. But I can't call it a pattern just because I can interpret the result of the study 100%. After all, I have performed implicit optimisation. And this is exactly what you did in the article.
No, unfortunately. You have to be honest, first of all, with yourself.
You conducted a study, saw that you could squeeze profit out of it, and wrote a TS that fails in a certain area. Then you saw that everything sucks statistically on this area and concluded that this is the reason why it doesn't work there. You're putting the cart before the horse.
Here I did a study on a blog by running just a script. I don't need any optimiser to create a TC with a profit on the study interval on the basis of this research. But I can't call it a pattern just because I can interpret the result of the study 100%. After all, I have performed implicit optimisation. And that's what you did in the article.
that there are seasonal fluctuations is an undeniable fact. It is as if they cannot be absent.
And even the fact that the market is usually determined at the hour "H" is also understandable, and can even be justified. But a specific instruction "here we buy, and here we wrap fish" simply based on the calendar and alarm clock is, to put it mildly, IMPOSSIBLE.
By the way, you can check: where to place an order with 300 pips TP/SL? There are demos, and no one is in a hurry - a couple of three months can be quite possible to see
I understand that this is more of an academic question - it is not verified by practice :-) we will definitely go grey here and not everyone will live to see the reliability of it (cfu, cfu, knock on wood).
how many deals should be made and for how long to test the hypothesis ?
that there are seasonal fluctuations is an undeniable fact. It's as if they can't be absent.
The point is that at any testing interval "cycles" will be found, regardless of the nature of the price. Even a random one. You have to look at what the test interval is.
How many trades do you have to make and in how much time to test a mortgage ?
None. One only has to look at the OOS to begin to have more faith in a hypothesis that is doomed to not be a theorem.
Would take a walk-forward to solidify my belief.
I can show you the drafts of the article, which is exactly what happened. The article was written on the fly, without prior knowledge of what could even turn out.
I gave an example of an entry from my blog. All sources are there, no TC or Optimiser. But this does not cancel the essence - statistical research, which means implicit optimisation. That's fine.
Talking about a pattern with a healthy dose of scepticism should probably be done when there is an OOS evaluation.
that there are seasonal fluctuations is an undeniable fact. It's as if they can't be absent.
And even the fact that the market is usually determined at the hour "H" is also understandable, and can even be justified. But a specific indication "here we buy, and here we wrap fish" simply based on the calendar and alarm clock is, to put it mildly, IMPOSSIBLE.
By the way, you can check: where to place an order with 300 pips TP/SL? There are demos, and no one is in a hurry - a couple of three months can be quite possible to see
And I realise that this is more of an academic question - it is not verified by practice :-) we will definitely go grey here and not everyone will live to see the reliability of it (cfu, cfu, knock on wood).
how many transactions should be made and for how long to test the hypothesis ?
The point is that "cycles" will be found at any test interval, regardless of the nature of the price. Even a random one. One has to look beyond the study interval.
Not how much. One only has to look at the OOS to begin to have more faith in a hypothesis that is doomed not to be a theorem.
Would take a walk-forward to solidify my belief.
it's crazy to look for arbitrary cycles.
But there are processes that determine some cycles. The fact that volatility rises in the morning and falls in the evening doesn't seem to bother anyone. This is a cycle generated by natural processes. From the point of view of trading - these are external conditions.
And there are many such cycles, not necessarily strictly calendar ones.
PS/ abstractly - Fedoseev (if I'm not mistaken in his nickname) showed a completely crazy thing here this week, he didn't even realise what it was...you should think about it :-).
PPS/ "nodal points of seasonal fluctuations together create a strictly linear structure". maybe old Gunn was damn right.
it's crazy to look for arbitrary loops.
It's crazy to reject crazy ideas.