Machine learning in trading: theory, models, practice and algo-trading - page 1029
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
In real BP you can observe a repetitive (24 hour period) at certain times of the day increase in volatility associated with the opening of sessions. In the random one there is no such thing.
If you do not know what to do and how to do it you will get a kick out of it, then you may start to trade on weekends only and forget about it on weekdays.)
Forum on trading, automated trading systems and trading strategies testing
New trading strategies and techniques in trading: theory and practice (trading and not only)
Aleksey Nikolayev, 2018.07.22 11:00
You can try to apply goodness-of-fit criteria (Kolmogorov-Smirnov, for example) between samples of increments of generated and real series. For example, real prices are said to give thicker tails and a sharper center than the Gaussian distribution.
I think that the random walk may be applied as an alternative history for the optimization of primitive trend systems that do not have predicting properties, but are simply trend following.
For example, to generate a 3000 year alternate history and optimize a trend robot on these data, it seems to me that with the obtained parameters the robot will behave better in real trading than if it was optimized for the last few years of real history, but I'm not interested in this anymore, I haven't experimented with it
oooh!
Are you still in search?
However, each new program will be graver than the previous one....
Here I take my leave for a heated discussion.
PS
Man is smarter anyway... Smarter than indicator, neural network, etc.
And how did you account for density? How did you even calculate it?
I called a dense accumulation of values a cloud, and tried to figure out what was better - to close at the beginning, middle or end of the cloud, in any case, such a phenomenon was talking about high risks of rebound. I invented the density bicycle here.
I think that the random walk may be applied as an alternative history for optimization of primitive trend systems that do not have predicting properties, but are simply trend following.
For example, I can generate 3000 years of alternative history and optimize a trend robot on this data, it seems to me that with the obtained parameters the robot will behave better in real trading than if it was optimized for the last few years of real history, but I am not interested in this anymore, I haven't experimented with it
Theoretically the balance will be martingale (because it can be represented as a stochastic integral of Ito). That is, it won't completely resemble a random walk, but the property of having zero expectation remains. Therefore, there is unlikely to be anything other than a rearrangement.
Perhaps it makes sense to add a small trend to the wandering and compare different systems by sensitivity to it. In fact, it would be a variant of the Monte Carlo simulation that was once scolded here)
I called the dense accumulation of values a cloud, and tried to figure out what was better - to close at the beginning, middle or end of the cloud, in any case, such a phenomenon spoke of the high risks of rebound. The density bicycle was invented here.
The method of nuclear density approximation does not suit you?
The nuclear density approximation method doesn't work for you?
I looked at the article with one eye. This method does not identify individual cluster groups, but mine does.
I looked at the article with one eye. This method does not detect individual cluster groups, but mine does.
If we are talking about clustering, then the density is usually approximated by a mixture of unimodal densities (most often Gaussian). For example like this.
there was a topic by the user, I think demi, where they tried to distinguish the real market from the pseudo-random, Dmitriy Skub, it seemed to work, used the Hearst index
https://www.mql5.com/ru/forum/143224
Buddy))) Sorry, but if you draw the curves on the randomly generated row and look at any targets on it then you are wooden at least up to your waist :)
Before replying to the post, it would be a good idea to read it.
And how did you account for density? How did you even calculate it? I need this tooThe fact that it is random I got it. I wanted to show that even on a random series you can apply tech analysis and see the intended targets, or do you think that none of my goals will not be achieved by continuing the series????
And therefore the technical analysis can be applied to ANY row. The main thing is that this series changes over time!!!!
As for the formation of levels. The most useful levels are the volume profile and delta at a specific price. And only these two components form them, not your infamous indices or paternas, as in my case with candlesticks.
The candlestick indicator is quite usable, but if it is filtered by the market profile. That is, when the candlestick pattern is confirmed by a large profile volume in the area of its formation.
If the volume profile is a past level, the future levels are formed in the market in the form of pending orders that are later considered as "Open Interest". That's all in the wisdom of the market. Look at the kotir as a market, not as a temporary unsteady series for statistics. And things will start to work out...
If I didn't have a good TS, I'd spin the levels better. As it is... I can't bring myself to do anything, if the question is already solved and the robot is trading on its own!