Theorem on the presence of memory in random sequences - page 15

 
Yury Reshetov:

...

Digging up the posts is a clear sign of 'scholarship'.

It's a particularly great sign of scholarship to rank taste by colour.
 
They sell an invisible hat on the internet. It hides donkey ears well.
 
charter:
They sell an invisible hat on the internet. It hides donkey ears well.
No wonder... although no, it's a very surprising world you live in: invisible hats, memory by the coin, trend by the cube... an orderly with a jacket around the corner...
 
charter:
You know what this "theorem" is about?
 
Дмитрий:
Do you know what this "theorem" is about?

Even if I don't understand something, that doesn't give me the right to say it's heresy.

It's just an excuse to work harder on the subject.

 

Created an EA based on the theorem. I'm testing it in the tester now.

Here is the chart:

Horizontal line - size of protective stops, offset - period in bars

Here are the results of forwards:


 
Yury Reshetov:

Created an EA based on the theorem. Trying it out in the tester for now.

It will be very unsurprising if a closer look at the EA will turn out that there is no hint of the theorem in it, just like in the game with dice from the first post.

Or that the EA runs on momentum.

 
Dmitry Fedoseev:

It will be very unsurprising if a closer look at the EA reveals that there is not even a hint of the theorem in it, just like in the dice game from the first post.

However, you have failed with your "scholarship". The Expert Advisor once again confirms that price time series have a memory, if the analysis of past values of the series to a depth of at least two. Therefore, such results are obtained even on forwards. Nevertheless, I expected that the expected payoff would be less volatile.

Dmitry Fedoseev:

Or that the Expert Advisor works on Momentum.

Almost. I.e., more simply - differences in opening prices of bars. Hence, the Expert Advisor has no indicators and no TA oscillations. There are only three input parameters, so complete optimization over the whole range is 6000 passes. The testing speed is fast because no TA indicators are used and the mode of testing is by bar opening prices.
 
Yury Reshetov:

However, you have failed with your "unsurprises". The Expert Advisor once again confirms that price time series have a memory, if you analyse the past values of the series to a depth of at least two. Therefore, such results are obtained even on forwards. Nevertheless, I expected the mathematical expectation to be more sober.

Almost. I.e. simpler - differences in opening prices of bars. And consequently, the Expert Advisor has no indicators and no TA oscillators. There are only three input parameters, and therefore complete optimization over the whole range is only 6000 passes. The testing speed is fast because no TA indicators are used and the mode of testing is by bar opening prices.

Now you are experimenting with data which is not random. But you have everything in one pile... that's not surprising either. In addition, it has long been known that even on random data, but a limited interval, there can be profit (on history).

The difference between two prices is momentum, whether it is in the form of indicator or directly in the Expert Advisor.

Forward Testing. We need to see how the test results are sorted. We should sort the results by backtest profitability and look at the forward ones here instead of the fact that there are profitable forwards. How is the table sorted?

 

There... In addition to Momentum, the terminal has a bunch of other indicators: stochastic, rsi, etc. etc.

By the way, there is a whole book about stochastic with theories and proofs that it is incredibly profitable.

Or a chapter (which is also not insignificant) in some book, I do not remember exactly, sorry.