a trading strategy based on Elliott Wave Theory - page 265
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
I try to be strong. :о)
If we divide the mentioned objects into statistical, phenomenological and microscopic, the first ones, in my opinion, are the least designed for evolution, since they are entirely based on the past. The latter are probably the most. My last posts are just that... dreams of such a model. :)
I also spent a lot of time collecting similar statistics, but then I came to the conclusion that using it to build a strategy is essentially a fitting of history, even if it's more correct than using parameters in the tester. However, this does not mean that it is useless at all.
Philosophy should not be allowed close to practice. Why, for example, one cannot look at a microscopic object as a statistical one? And what are phenomenological objects I do not know, but I have a suspicion that they can be any objects, including microscopic ones. Though I do not argue, for I do not understand.
I don't agree about gathering statistics. If you set the problem correctly, it is just statistics that eventually gives the possibility of obtaining empirical laws of possible evolutions of the system.
There is a sort of classification of theories. Microscopic is when equations are written down based on one's understanding of first principles (and then the data are fitted to this theory :). Phenomenological is when some general enough equation is chosen, parameter values are determined at which it satisfactorily describes data and it's stated: Who cares what's really going on there. Statistical is when no equations at all are specifically chosen for the data, but it is just assumed that everything will be the same again as always. These are not canonical definitions, but my own free interpretations :)
Seems clear, but slippery somehow. :о)
What if we try to simulate as fully as possible the price formation process for the EUR/USD pair, for example? Let's take a real tick history, for example over a year, and calculate in a known way all coefficients of an AR-model of order p for a series of first differences:
X[i]=SUM(a[k]*X[i-k])+s, where k runs from 1 to p, and construct our tick series by fitting the appropriate stochastic term s, coinciding with the real series of integral characteristics(IC) like residual distribution functions (at all time lags!) and their correlograms. This was done. The corresponding ICs coincided satisfactorily and we can say that there is no way to distinguish where the series of ticks is real and where it is artificial (well, almost impossible). The result of the modelling can be seen on the picture:
An interesting point has arisen. The point is that the residuals model series does not belong to natural number class and has to be rounded beforehand (points are always integers after all) and only those terms which >1 are used. It is equivalent to influence on the initial model series a filter which passes only integer values and rejects all others (many other things), including interesting figures with amplitude less than 1 point... If, like me, you had a good beer beforehand, then it's not difficult to guess that the informers give away in the form of an indicator the exhausted and therefore useless material, and the most interesting and tasty thing is within the spread and unavailable to us - the end users. But someone trades with this stuff. What do you think?
What if we try to simulate the pricing process as fully as possible, e.g. for the EUR/USD pair?
....
Hello Sergey. That's an interesting thought, except I don't understand "why?". And are you really sure you now know how the price is formed?
What are you doing? - Renco H-strategy is slowly gaining statistics on the account, and at the same time the code is being fine-tuned... All that can be counted - counted, so it remains only to drink beer, and ideas on the forum to plant in the hope of a response.
That's right, now I'll finish work and go drink beer too. The statistics have not collected all, and the experiment must be set competent. And that can be done after a beer. By the way, you wrote that you "gave up" with constructions and your current strategy gives more chances to earn. Although I've probably got it all wrong again. :о)))
I don't know about the reality, but it is hoped that the tick is the result of averaging a certain number of real trades. In that sense, there should of course be a fractional peaking part. But if it can do more than catch these very pips, I will be surprised. Although, of course, I have not seen these figures and anything may happen.
Is it possible to give more details about the type of stochastic member s? At least the kind of beer :)
So that's how it turns out: "Everything is brilliantly simple!".
Actually rounding causes white noise in quotes measuring plus or minus seven points.
Did the calculations results here
http://forum.fxclub.org/showthread.php?p=717988&posted=1#post717988
About the same can be found in posts where similar calculations have been done.
Actually it is strange that you ask such a question, because you can use your model to calculate how much white noise in the model series affects the prediction or am I misunderstanding something?