Machine learning in trading: theory, models, practice and algo-trading - page 800
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
This is an erroneous opinion. The notion of trend-flat is very relative. What is a flat for one person may well be a trend for another.) And vice versa.)
Of course, Bolinger-like strategies have their limitations. Just like any other.
There's a lot of counter-trend systems on boyls and stuff, all the way to the basket.
even if you're lucky enough to make a couple of months in a pacific session, it's a blessing.
It's been around for a hundred ice years.
for such systems, the ratio of profit trades to loss trades should be much more than 0.5, small profits and big losses
FAP Turbo was a popular bot in the past, it used to take sacks of trades, now it seems that the third version gives something, I don't know
There are a lot of counter-trend systems on the fries and so on, the path is the same - in the basket.
even if you're lucky enough to chop a few months in a pacific session, it's a blessing.
It's been around for a hundred ice years.
for such systems, the ratio of profit trades to loss trades should be much more than 0.5, small profits and big losses
FAP Turbo was a popular bot in the past
Bollinger by itself is a great pit. You should never use the standard formulas to calculate the variance. But, you can and should use nonparametric methods to calculate the measure of process deviation from the mean.
Thanks, I'll read it.
Here, by the way, is the appendix to the vid from Kuznetsov above. 2018 something interesting, but I don't understand it yet. There are examples of predicting bitcoins, forex, etc. And a comparison of his method with arima.
https://arxiv.org/pdf/1803.05814.pdf
I calculated the predictive ability of 23 predictors for 12 currency pairs based on the difference in distributions for the teacher, with a correct prediction the profit would be over 50 pips.
The results are as follows:
1. The predictive ability of the same predictors for different currency pairs is different.
2. Predictive ability of different predictors for one currency pair can differ by two orders of magnitude
3. Predictive ability will change as the window moves. As the window moves over 500 bars the statistics of variability of the prediction ability stabilizes
4. The slope of the predictive power obtained by moving the window varies from values of less than one percent to over 100 percent. And the "bad" predictors (with a large sko) are always bad, and the "good" predictors are always good.
5. We have studied 12 currency pairs. Three of them are hopeless: I couldn't find any good predictors for my target variable among 23 used ones.
6. For one and the same currency pair the predictive power of longs and shorts is radically different.
The Bollinger itself is a complete pit. Under no circumstances should you use standard formulas to calculate variance. But you can and should use nonparametric methods to calculate the measure of process deviation from the mean.
I can't help it - you're talking nonsense. Bollinger is just an indicator with a bunch of settings. On its basis, you can build all sorts of strategies, including return to the mean. Your concept is essentially the same Bollinger - the design is different. And the Bollinger is a pit. What did you do? - We built it differently - we replaced the MA, reconfigured the borders. That's all. Call it by your name, like some people do here.) It's funny.
I can't help it - you're talking nonsense. Bollinger is just an indicator with a lot of settings. On its basis, you can build all sorts of strategies, including return to the average. Your concept is essentially the same Bollinger - the design is different. And the Bollinger is a pit. What did you do? - We built it differently - we replaced the MA, reconfigured the borders. That's all. Call it by your name, like some people do here.) Ridiculous.
Pure Bollinger is a pit. And bollinger-like systems are okay. I see no contradiction.
I have calculated the predictive ability of 23 predictors for 12 currency pairs based on the difference of distributions for the teacher, if correctly predicted the profit will be more than 50 pips.
I have an idea to start by taking n-distributions of market ones, no matter what their characteristics are, and fitting them with n-models that will compete with each other.
But it's a blunt instrument, the main idea will be born in the process, as usual :)
I realize I don't know anything about tervers, and there's a lot of interesting stuff buried there. + 2 more weeks to study, found something to do.
And then something more complicated and interesting, I will read what you have given and then something new food for thought
Pure Bollinger is the pit. And bollinger-like systems are fine. I don't see any contradiction.
As described in books on TA - yes, of course. But everything is in the pit.)
Okay, we've come to a compromise solution.)
I have an idea, for a start, to take n distributions of market ones, no matter what their characteristics are, and make n-models that will compete with each other, which is the most important right now
But this is a dumbass sketch, the main idea will be born in the process, as usual :)
I realize I don't know anything about tervers, and there's a lot of interesting stuff buried there. + 2 more weeks to study, found something to do.
And then something more complicated and more interesting mb, I'll read what you've given and then something else that would give food for thought a new
I wanted to show in my post that the success of classification models is completely determined by stationarity of predictive ability of predictors. If that predictive ability changes manifold, then collapse is guaranteed.
Another great conclusion from my post: BEFORE the tester, before the demo and the real, one should investigate the predictive ability of model predictors. Any thoughts about it are the way to stable performance of TS in the future.
With my post I wanted to show that the success of classification models is entirely determined by the stationarity of the predictive power of predictors. If this predictive ability varies manifold, then you're guaranteed to fail.
Another great conclusion from my post: BEFORE the tester, before the demo and the real, one should investigate the predictive ability of model predictors. Any considerations on this subject are the way to stable performance of TS in the future.
The Bollinger itself is a complete pit. Under no circumstances should you use standard formulas to calculate variance. But you can and should use nonparametric methods to calculate the measure of process deviation from the mean.
You just do not know how to make real money on it, apparently.
There was a report in the forecasts branch from the real, it was amazing...