Machine learning in trading: theory, models, practice and algo-trading - page 3247
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
Not Marty then, congratulations!
Well, it would be strange if no patterns were found at all, then no TCs would work.
I'm just saying that's what you can look for.
I'm just saying that' s how you can look, too.
Didn't see a methodology, just a graph.
Didn't see a methodology, just a graph.
correlation matrix between the rows of given attributes, then the most correlated rows are selected, a graph of n closing prices is built for each row to see how it was in the future, statistics on all rows is taken, how it was in the future on average, all patterns are filtered according to these statistics, the best ones are selected.
the pattern reference is saved, in the tester we look for correlation of current values with the reference, open trades according to the selected logic.so far I have not been able to beat the results of MO
I don't understand what correlation has to do with it? I think it's enough to just take a string from scratch and look at its behaviour over the mining interval.
Well done! It turns out it's no longer MO, but a complete overkill without any hint of blackbox. I like it.
The length of the string? Judging by the animation - the length is four of 10 possible values.
I don't understand what correlation has to do with it at all? It seems to be enough to just take a string from scratch and see its behaviour in the future.
that example from the tester had a length of 9 (increments of different periods).
different periods in one 1D array, stacked into one curve, allow you to shove different signs into one sequence, which allows you to search for a pattern through correlation, not MO.
I'm trying with different ones, I still don't have enough lightning speed to count everything very fast, I'll try to speed it up more
In the optimiser you can take one, then another... and race it like that. But I do in python and calculate correlation for all possible pairs at once, then choose from that.
The most important thing here is speeddifferent periods in one 1D array, added into one curve, allow shoving different features into one sequence, which allows you to search for a pattern through correlation
I don't get it.
I'm trying different ones, still not fast enough to count everything very quickly, I'll try to speed it up some more.
I didn't get it.
Let's say there are several indicators, you can write their values into 1d array sequentially. And compare with other similar ones through correlation.
Let's say there are several indicators, you can write their values into 1d array sequentially. And compare them with other similar ones through correlation.
Then it is necessary to bring the indicators to some uniform parrots. Even if the indicator is the increment at different intervals, otherwise the correlation will be strange.
I myself would move a window in a 1d-array and look at the number of samples through any sign of "similarity". More precisely, not even the number, but the total "profit" of such places (entering by pattern, exiting in n hours).
Only the number of found samples on the animation is confusing: several hundreds. In four years, there are only 4*365*5/7*24~25000 samples. To find 500 samples among 25K is either a very rough pattern (or a sign of similarity) or some kind of off the scale repeatability (with a hypothesis of regularity).
Then it is necessary to bring the indicators to some uniform parrots. Even if the increment at different intervals serves as an indicator, otherwise the correlation will be strange.
I myself would move a window in a 1d-array and look at the number of samples through any sign of "similarity". More precisely, not even the number, but the total "profit" of such places (entry by pattern, exit - in n hours).
Only the number of found samples on the animation is confusing: several hundreds. In four years there are only 365*5/7*24~6000 samples. Among 6000 to find 500 samples is either a very rough pattern (or a sign of similarity) or some off the scale pattern.
there in the loop for each sample the correlation with the others is filled in using the matrix already calculated, so it is a lot of