Discussion of article "Exploring Seasonal Patterns of Financial Time Series with Boxplot" - page 6

 
Maxim Dmitrievsky:

For me, as an arbitrageur, it is quite obvious that the market is extremely efficient and any regularity on it is simply destroyed, this is normal

when I wanted to write an ordinary TS on some pattern, nobody could show me any interesting stable pattern.

Different approaches show that these "regularities" are extremely weak and short-lived, but sometimes you can pull it out, an example is in the article. A step to the left - a step to the right and you lose.

you can go the other way round - the market is really extremely efficient, and it studiously avoids direct/mirrored repetitions.

That's where you can dig in. Due to the efficiency of the market and the laws of information, it won't take long, but,

perhaps the collected statistics and found regularities should be used in a different way - in hours other than HH and days other than DD, the matter is a pipe - we trade in the pipe :-) it is not enough to find a regularity - it is important to exploit it.

 
fxsaber:
Running Optimisation.


Complete overrun, since it took ~50 minutes to calculate. The result should be shitty, because BestInterval was not calculated (we need to think how to fix it) to work on opening prices (many trades occur at the same time). For this reason, in particular, the new biblical feature is not used, so the potential will not be unlocked yet. But as an experiment of what we have, let's see.

For comparison, the result of the Expert Advisor from the article.

Profit = 1462 pips, PF = 1.85, MaxDD = 1099 pips.

What is that red spot in the upper green bar - I don't know.

It's just an adjustment to a piece of history, isn't it? I mean, if you take half of it, the second half won't be profitable.

 
Maxim Dmitrievsky:

For me, as an arbitrageur, it is quite obvious that the market is extremely efficient and any regularity on it is simply destroyed, this is normal

depends on the period you want to consider.

on the period of a year there are regularities and forward tests take place.

on longer periods there will be periods when the TS works or not.

i.e. the market is efficient, but the efficiency of the market lies in "who will be the first to discover a pattern and stop using this TS in time".

in defence of "mindless optimisation", here is the search for a pattern by entry time, I searched from 01.01.2019 to 01.05.2019, then forward - this TS forward has the same tendency as on the optimisation

with this approach you can quickly go through the symbols and find a TS that works, but the question remains open how to find a mathematical evaluation that the TS stopped working.... evaluation on a drained deposit is not the best way ))))

 
Maxim Dmitrievsky:

It's just fitting a piece of history, right? I mean, if you take half of it, the other half won't be profitable.

That's the EA from the article, not my version.

 
fxsaber:

This is the advisor from the article, not mine.

why are the systems so smooth, is this a tweak to the current 30p euro market?

 
fxsaber:

This is the advisor from the article, not my version.

Oh, I see.

You can use ticks as an option, it will be even more correct.

 
Igor Makanu:

depends on the period you want to consider

on the period during the year there are regularities and forward tests pass

on longer periods there will be periods when the TS works or does not work

i.e. the market is efficient, but the efficiency of the market lies in "who will be the first to discover a pattern and stop using this TS in time".

in defence of "mindless optimisation", here is the search for a pattern by entry time, I searched from 01.01.2019 to 01.05.2019, then forward - this TS forward has the same tendency as on optimisation

with this approach you can quickly go through the symbols and find a TS that works, but the question remains open how to find a mathematical evaluation that the TS stopped working.... evaluation on a drained deposit is not the best way ))))

I have never seen such charts on real, except for arbitrage

 
Fast235:

why are the systems so smooth, is this a tweak to the current market at 30p on the euro?

This is to the author.

 
Maxim Dmitrievsky:

you can use ticks, it's even more correct.

So we take a rough estimate as in the article. I am interested in what the Optimiser will give compared to the approach in the article.

 
Attached is the result of the Optimisation.
Files:
Python.zip  14024 kb