Machine learning in trading: theory, models, practice and algo-trading - page 1582
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
there is something in it!
I had tested in the 1000 and 1-st configuration of a grid of orders, the main problem is that very poorly forward passes on real ticks, and the dynamics of the equity graph on the forward is shifted in time, in a couple of hours I will show you the screenshots of the tests - what I write about
Well, remember the last article you hated) where the shift of returnees by specific hours does not allow to build a short strategy at all, in principle
I've sent for checking another article, for other watches, on which it is impossible to build a long strategy, and the shorts have been bulldozing just like that for 5+ years. Moreover, the chart on this interval is both rising and falling, it makes no difference. But the purchases there are doomed everywhere.
And in the previous 10 years the average returnee is positive by the same timeframe, and there longs are in the red, although the chart is growing and falling, periodically.
Try adding my research on leaf groupings to random forests.
I assume that the forests are not balanced by votes, and after grouping the voting leaves, you can improve the results. I suggest you give it a try, since you have figured out in detail the library you can use to build forests in MT5.
You can figure out the library in about 5 days. Just sit down and read the code, write comments on it, describing what each line does.
After that you can edit the forest as you want.
I filtered the leaves as well. But not manually, but simply by percentage of successful hits on a training plot. For example I took all leaves with 90% success. But they also on the forward gave 50/50.
So I gave up. And I don't have time...
I'm now doing something else, which really feeds me, and not just promises, like forex with forests.
Forests work where there are regularities. Physical processes, for example. Once I read an article about using MO to control and manage natural gas purification and liquefaction processes. There MO works better than any automatics...
If the same scaffolding does not work at Forex, it means that there are no regularities. At least I did not find them in price returnees and in real SME volumes too. No other information is available to ordinary traders.
The library can be figured out in 5 days. Just sit down and read the code, write comments on it, describing what each line does.
After that you can edit the forest as you please.
I also filtered the leaves. But not manually, but simply by the percentage of successful hits in the training area. For example, I took all leaves with 90% success. But even they were 50/50 on the forward section.
So I gave up. And I don't have time...
I'm now doing something else, which really feeds me, and not just promises, like forex with scaffolding.
Forests work where there are regularities. Physical processes, for example. Once I read an article about using MO to control and manage natural gas purification and liquefaction processes. There MO works better than any automatics...
If the same scaffolding does not work at Forex, it means that there are no regularities. At least I did not find them in price returnees and in real SME volumes too. No other information is available to ordinary traders.
I also have a hard time with free time now - I had to go to work for a paycheck, and in the last 3.5 years a lot has changed in my professional field.
I didn't suggest culling anything by hand - there's a code in the thread.
I agree that MO methods work on stationary processes, but I don't agree that they are not on the market, they are, but mixed in with the chaos.
I agree that MO methods work on stationary processes, but I do not agree that they do not exist in the market, they do exist, but they are mixed with chaos.
remember the last article you hated)
the article is ok, the example is a bit far-fetched, but in short - it is bullshit
What I wrote here is the test of a grid of orders with some kind of order accompaniment - I do not even know what is there and how, the GA selects - not the main point, here is the optimum by real ticks, time of TS work 18 - 1 hour, the optimum for 6 months of 2019
In principle, not a bad dynamics, especially if the GA himself picked it up, it would seem possible to continue to work with it.... and here is the test for 12 months of 2019 of this TS - note on real ticks!
and here is the forward:
I don't think it's possible to continue working with this forex broker in 12 months of 2019 - note the real ticks!
the article is ok, the example is a bit far-fetched, but in short - it is bullshit
I am writing about that, here is the test grid of orders with some kind of order accompaniment - I do not even know what is there and how, the GA selects - not the point, here is the optimum by real ticks, time of operation of TS 18 - 1 hour, optimum for 6 months of 2019
In principle, not a bad dynamics, especially if the GA himself picked it up, it would seem possible to continue to work with it.... and here is the test for 12 months of 2019 of this TS - note on real ticks!
and here is the forward:
and it was expected that the drawdown would be no more than in the test? and the drawdown starts to float over time, the longer the forward, the more frequent the drawdowns
This example is not far-fetched. If you were confused by MA, I removed it in the next article, the essence has not changed.
This is normal for the grid, the cumulative loss becomes larger on the forward side, since the pattern is no longer the same, but it continues to withdraw at the expense of the grid.
Let me give you another example:
Fix stop, fix profit. No grid, no averaging, no optimization. TS has only 2 outputs. The regularity - the shift of the average increment over the entire period. It does not matter how a specific trade will close, on average they are in the +. TS is as simple as a stool.
The reality and unreality of ticks practically does not affect the results in my case, + and the same, because all the signals are closing prices.
Well and note that the test for 10 years, 15 min. TIME FRAME.
The example is not a far-fetched one. If you were embarrassed by the Mashka, I removed it in the next article, the essence has not changed.
Bullshit, whatever you want to call it.
Well and note that the test for 10 years, 15 min. TF.
I have not a problem at all, here I have found a test with 2 orders at 2 minutes that has been running since 2010 till now, I write that order accompaniment is more important than a certain regularity found through reasoning ;)
I think there are patterns in price behavior after political decisions are made, after important news is released. Here's a 10 minute window to know it all )))
Just, I think, that the price is corrected to "regularity" by impulses, which can be during news and other events, and between them chaotic accumulation/distribution.
It's not a problem at all, here I have found a 2-minute test with 2 orders that has been running since 2010.
I do not understand what the follow-up has to do with the regularities being traded. Apparently, it is not given.
I don't understand what accompaniment has to do with trading patterns. Apparently, it's not a given.
That is... If we look at trading from the position of the game theory, then it is very important to select the parameters of the TS which will correspond to the optimal game
And if we look at trading from the position of some found regularity on the history, then it's trading according to the forecast - how the forecasting is carried out... You may knock on a shaman's tambourine, or you may selectively thin out the ticks to fit the same shaman's tambourine )))