Statistics as a way of looking into the future! - page 11
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
No, it's not NS. The model is built on the method of local regression with optimal structure fitting at each point of the investment space. "Overfitting" is the hallmark of this method. In fact, we can say with great confidence that if the inputs to the model (in this case a series of price increments of some dimensionality), had any predictive value at all, a run on a test sample would have produced much better results.
So the conclusion from this picture is that either the market is very efficient and there is nothing to catch, or the model is trained on data that does not correlate in any way with future movement.
But we as true dreamers have to believe in the latter and keep looking :)
I.e. there is one conclusion from this picture: either the market is very efficient and there is nothing to catch, or the model is trained on data that does not correlate in any way with future movements.
The third variant is an inadequate model. But we can check it only if we obtain better results on the same data, but with a different algorithm. In principle, if the length of incremental vector is sufficient (at least 10 000 samples), I can try to build a derivative model and dump the result for comparison. But I must say at once that if it's just increments of cotierre on bars no NS will help.
Option three is an inadequate model.
A perfect finalisation.
I have a question for the participants of the discussion - what is the basis for your desire to approximate the market by a curve? Is it really just because we know some approximation method and let's ride this curve horse around the market and collect a profit? Or is it because we believe that the market fits into a polynomial with hidden parameters, and we only need to investigate these secret parameters? Tell me, how do you determine that the method you use is suitable in the original sense? Or is it a secret?
Excellent finalisation.
I have a question for the panelists - what is the basis for your desire to approximate the market with a curve? Is it because you know some approximation method and let's ride this curve around the market and make a profit? Or is it because we believe that the market fits into a polynomial with hidden parameters, and we only need to investigate these secret parameters? Tell me, how do you determine that the method you use is suitable in the original sense? Or is it a secret?
It's not a problem to find a polynomial, the most important thing is to know the factors influencing the market. If you use only time in a polynomial, you can't predict.
I have a question for the panelists - what is the basis for your desire to approximate the market with a curve? Is it because you know some approximation method and let's ride this curve around the market and make a profit? Or is it because we believe that the market fits into a polynomial with hidden parameters, and we only need to investigate these secret parameters? Tell me, how do you determine that the method you use is suitable in the original sense? Or is it a secret?
Well, firstly, the approximation is not a curve but a hypersurface. Secondly, what do you offer us, dear one? You can see that our methods will lead us to starvation. Waiting for your salvation :)
Well, firstly, the approximation is not a curve, but a hypersurface. Secondly, what do you offer us, my dear? You can see that our methods will lead us to starvation. Waiting for your salvation :)
In my opinion, an approximation is a better way of determining a trend than a moving average. At least for this reason you can search for patterns
It is not a problem to find a polynomial, the most important thing is to know the factors influencing the market. If you use only time in the polynomial, you will not be able to predict
Suppose we don't know the factors affecting the market, then what?
Time - why use it? For example, does the polynom "know" the holidays in the countries of the traded instruments? Easter, second day, all of Europe is asleep, volatility is close to minimum, i.e. there is nothing all day, so our polynomial should take this into account (lunar calendar).