Machine learning in trading: theory, models, practice and algo-trading - page 270
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
Why is everyone so fixated on models? Why doesn't anyone develop the topic of signs? Why doesn't anyone talk about non-stationarity? Why doesn't anyone try to solve these problems? Why doesn't anyone think about what drives prices?
If you use a stochastic, it doesn't matter what model you use, be it a usual KNN or the most sophisticated deep net,the accuracy will be 51-53%, no matter how deep it is. What's the use of these models if the input is garbage? but 95% of attention goes to the models, for me personally the models are the last stage of the system, and it's only 2% of the work
And it is quite another case, when the MO is applied to a large volume of data, where the algorithmic analysis (using direct formulas) is very difficult or even impossible. But here, in general, only combinations of stochastics coupled with machcs are spoiled, so the question "why?" is not particularly relevant here.
Why is everyone so fixated on models? Why doesn't anyone develop the topic of signs? Why doesn't anyone talk about non-stationarity? Why isn't anyone trying to solve these problems? Why isn't anyone thinking about what drives prices?
You've got the wrong idea about the thread you're on.
Look at my posts, and not only my posts, which say that the main problem is in data mining. I even mentioned the figure of labor intensity distribution - over 70% in data mining.
Moreover I claimed and still claim that the choice of model had little effect on the final result.
Moreover I and other forumers cited specific algorithms that would allow to sift the original set of predictors from the noise. At the same time it is stated that without noise predictors the model is NOT RETRACKED.
This is all available on this thread.
PS.
Non-stationarity was not considered because classification models are considered, not regression models, and the effect of non-stationarity on the performance of classification models is not entirely clear.
Why is everyone so fixated on models? Why doesn't anyone develop the topic of signs? Why doesn't anyone talk about non-stationarity? Why isn't anyone trying to solve these problems? Why isn't anyone thinking about what drives prices?
If you input stochastic it does not matter what model you use .....
Non-stationarity does not mean non-predictability, it says that simple statistics such as expectation and variance drift, not even the regularity of this drift is analyzed, if they drift then they are not stationary. In the context of MO, non-stationarity is not a problem, non-stationarity is a problem for systems built on the assumptions, piecewise constancy of expectation and variation. MO may use window expectation(maschines) and variation(bollinger) as features, but it is a very small part of features and errors of these features can be partially eliminated. The main problem is in fast market reactions to new information, which is not determined by available features, the only hope is on the insiders and connected with them different "premonitions", when there are certain patterns of participants' behavior before the news announcement. That is, because of insider actions, the market is more predictable.
Why do you need stochastics? In fact the difference between the MO stochastic and the standard momentum is not great, you don't need to use other than momentum as the simple window expectation of returnees.Since all that stuff with "an ideal indicator" especially in the context of smoothing, it looks like alchemists trying to make a philosopher's stone or an eternal engine, they are ignorant fanatics. I mean the levels that we see with our eyes on the chart, where people place stops. For example "levels" could be one of the most important features, I mean those levels that we can see on the chart with our eyes where people place stops.
Non-stationarity does not mean non-predictability, it says that simple statistics such as expectation and variance drift, not even the regularity of this drift is analyzed, if they drift, then non-stationarity. In the context of MO, non-stationarity is not a problem, non-stationarity is a problem for systems built on the assumptions, piecewise constancy of expectation and variation. MO may use window expectation(maschines) and variation(bollinger) as features, but it is a very small part of features and errors of these features can be partially eliminated. The main problem is in fast market reactions to new information, which is not determined by available features, the only hope is on the insiders and related diffusive "heralds", when certain patterns of participants' behavior appear before the news is announced. That is, because of insider action, the market is more predictable.
...
What?
Non-stationarity does not mean non-predictability, it says that simple statistics such as expectation and variance drift, not even the regularity of this drift is analyzed, if they drift, then non-stationarity. In the context of MO, non-stationarity is not a problem, non-stationarity is a problem for systems built on the assumptions, piecewise constancy of expectation and variation. MO may use window expectation(maschines) and variation(bollinger) as features, but it is a very small part of features and errors of these features can be partially eliminated. The main problem is in fast market reactions to new information, which is not determined by available features, the only hope is on the insiders and related diffusive "heralds", when certain patterns of participants' behavior appear before the news is announced. That is, because of insider action, the market is more predictable.
1. Non-stationarity = variance equals infinity. "Drift" is groundbreaking!
2. Highlighted in red - stocked up on popcorn and beer. Looking forward to the show predicting the price range by MO methods!
1. Non-stationarity = dispersion equals infinity. "Drift" is groundbreaking!
2. Highlighted in red - stocked up on popcorn and beer. Looking forward to the show predicting the price range by MO methods!
We're not really interested in your support of beer and popcorn makers here.
We here are interested in thoughts on identifying problems in the marketplace and solving them. not in general, but in deciding on positions.
To me, there are two such problems:
1. Unsteadiness in predicting the SIGNIFICANCE (VOLUME) of the quotient
2. Overlearning when predicting the direction of kotir movement.
In doing so, the MO can name not only the problem, but also discuss the problem-solving tool, moreover, justify the accuracy of the result obtained.
What?
What's wrong with it?
1. Non-stationarity = dispersion equals infinity. "Drift" is groundbreaking!
2. Highlighted in red - stocked up on popcorn and beer. Looking forward to a show with price range predictions by MO methods!