Machine learning in trading: theory, models, practice and algo-trading - page 2120
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
Um... And then why don't the MOshniks have positive stats? After all, the ACF of the market series of increments is not equal to 0 and therefore should be predictable. Obviously, because the 2nd condition for predictability according to Kolmolgorov is not met - there is no constancy of the expectation for any sample of data. What's wrong?
Is it the ACF of prices? If for the price it is not much different from ACF of SB, and ACF of returnees is like white noise, on normal data, unchecked by ACF masters, sometimes the higher frequency (minutes and above) the negative component appears if the price is formed from deals, but not from the average (bitwofer), but deals influence bids and ask ask and there is a reverse pseudo regularity, but there is no profit from it.
And " then why don't the MOs have positive stats? " because the data they use is sterile, there are almost no patterns in it that can be traded, you need more data, better data, new types of data, etc.
But a negative result is also a result, MO allows you to "dot all the I's". As a matter of fact MO is just an extension of statistics, "heuristic statistics" it gives more powerful ways to extract patterns from data and prove more or less reliably if there are any regularities in this data at all, unlike TA and optimization of mashups.
ACF prices or what? If prices, then it is not much different from the ACF of SB, and the ACF of returns as in white noise, on normal data, not chemically chemists with DC, sometimes the higher the frequency (minutes and above) appears a negative component if the price is formed from transactions, rather than the average (bit/ofer), just transactions hit then on the bid or ask and there is a reversal pseudoregularity, but it does not earn.
And " then why don't the MOs have positive stats? " because the data they use is sterile, there are almost no patterns in it that can be traded, you need more data, better data, new types of data, etc.
But a negative result is also a result, MO allows you to "dot all the I's". In fact MO is just an extension of statistics, "heuristic statistics" it gives more powerful ways to extract models from data and prove more or less reliably if there are any regularities in this data, unlike TA and optimization of mashups.
So what is your approach to the market?
So what is your approach to the market?
There were different approaches, at the moment I'm trying to collect a list of some "target market states", which could be better predicted than returnees, which were then traded by usual trend or reverse strategies, "states" like trnd-flat, "with and without patterns", calm market - agitated, etc. Probably it is obvious to all that the market is not always the same, when it is necessary to trade when it is not.
And there's no point in asking me about my "approach to the market" as I haven't even earned a million dollars by trading, let alone a lakh, I'm trying this and that, it's interesting as a game, but I have nothing to boast about.
There were different approaches, at the moment I am trying to collect a list of some "target market states", which could be predicted better than returns, which were then traded by conventional trend or reverse strategies, "states" like trnd-flat, "with patterns" and without, calm market - agitated, etc. Probably it is obvious to all that the market is not always the same, when it is necessary to trade when it is not.
And there's no point in asking me about my "approach to the market", because I haven't even made a million bucks trading, let alone a yarmulke, I'm trying this and that, it's interesting as a game, but I have nothing to boast about.
got it
As I go along, the frequencies and possibly the phases are floating... The amplitudes are holding...
Here is the forecast for 500 points of the fitted model on the history of 10k of 4 harmonics
We can see that the forecast is relevant for all 500 points, but the frequencies are drifting, and they are drifting according to an incomprehensible algorithm
And this is an illustrative example, sometimes it's even worse
I have no idea what to do with them, I do not know what to do with them, they just keep on doing it. I failed with waves, spectrograms of market increments are indistinguishable from spectrograms of white noise, alas, at least in my experiments, however I know guys who claim the opposite and beat their chests that they feed on the market.
My thoughts on this matter are as follows: in principle waves should have a place. The market has reversal patterns, it can be seen in the results of simple forecasting of increments, it is obvious that the MO is trying hard to trade in the reverse (pullback), but it does not work so well. And since there are reversal forces there should be waves IMHO.
Waves are cool, one way or another any trader should try furying or alternative alchemy like Elliot waves etc. I failed with waves, spectrograms of market increments are indistinguishable from spectrograms of white noise, alas, at least in my experiments, but I know guys who claim the opposite and in addition beat their chests that they feed on the market, so...
My thoughts on this matter are as follows: in principle waves should have a place. The market has reversal patterns, it can be seen in the results of simple forecasting of increments, it is obvious that the MO is trying hard to trade in the reverse (pullback), but it does not work so well. And since there are reversal forces there should be waves IMHO.
Pardon me, of course... But how can one confuse market increments with "white" noise?
Here, just took the GBPCAD increments for the current month:
A total of 10 trading days have passed. Can't those 10 days be seen in this figure?
In short. There is no white noise in the market, never has been and never will be. And the problem with cash withdrawals is not that the market is a completely random process with no Wiener-type memory.
The problem of working with increments is that most TSs do not take into account at all the time (specifically - day of week, hour, minute) of arrival of this or that increment.
And if we take as a paradigm the thesis: "it is necessary to supply to NS inputs everything that comes to hand, and it will sort itself out," then certainly time, along with return, is the main input parameter.
Pardon me, of course... But, how can one confuse market increments and "white" noise?
Here, I just took GBPCAD increments for the current month:
A total of 10 trading days have passed. Can't these 10 days be seen in this figure?
you have in your figure a multiplicative process with cyclic activity (most likely, the cyclic activity depends on the time of day)
If you are able to convert cyclicality into estimated periodicity or seasonality... Profit then, but otherwise - simple curves, imho
Pardon me, of course... But, how can one confuse market increments and "white" noise?
Here, I just took GBPCAD increments for the current month:
A total of 10 trading days have passed. Can't those 10 days be seen in this figure?
In short. There is no white noise in the market, never has been and never will be. And the problem of cash withdrawal is not that the market is a completely random process with no memory like a Wiener.
The problem of working with increments is that most TSs do not take into account at all the time (specifically - day of week, hour, minute) of arrival of this or that increment.
And if we take as a paradigm the following thesis: "it is necessary to supply to inputs of NS everything at hand, and it will deal by itself", then certainly time , along with return, is the main input parameter.
I did it, but it did nothing.