It's impossible to make money on Forox!!! - page 36
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
Этим вроде Mathemat занимается.
No, I don't anymore. Too much difficulty - if you have to do something like real data. But joo does not seem to need it. However, the task doesn't get any easier from that.
The main problem and intrigue here is something else (I was even going to write an article about this and something else, but didn't want to). I am not a professional mathematician, so I will try to express myself as best I can.
Each trading system uses only some statistical properties of quotes series. We can say that for each simple TS there is some "variety" (space) of possible quotes histories, that with a given real series would give exactly the same results as for the real series.
Here's the first example, simpler at first. Let's take the complete history of the oira (say, H1 since January 1999) and build one simple waveform on it with a period of, say, 11. The question: are there other histories ("synthetics") where this one will be exactly the same as on the real quotes from the beginning of the history as a function of the bar number?
Yes, there are, and there are a lot of such stories, an uncountable number (continuum). Obviously, there are 10 less linear equations defining the Mach value at each bar through prices than there are bars themselves (because the Mach values at the first 10 bars of the story are not defined). There are fewer linear equations in this system than there are unknown prices. Hence, the solutions form a linear space of dimension 10. So if we only look at the waveform without paying attention to the price, we lose a lot of information about the price series (we cannot unambiguously recover it from the waveform history).
Second example. On the same story we build two wagons (11 and 17) and play by the intersections. The question is the same: is there another story, different from the real one, on which the points of intersection of the dabs will be the same? The answer is similar: yes, very much so. There are obviously far fewer equations for the unknown prices that determine the already known crossover points than there are bars. The remaining constraints are already inequalities. I.e. there are even more degrees of freedom for prices.
I strongly doubt that the complete set of solutions (price vectors) in this case defines any given distribution of returns. On the other hand, it is obvious that real quotes show us something statistically definite, though not well known to anyone and also non-stationary.
Both examples above were connected with simple indicators.
The third example - the game on a standard zigzag. Here the answer is obvious: there are also many synthetics with the same zigzag vertices.
joo, how will you determine by the TS itself how much you can change the quotes?
P.S. Yes, my answer is very late, but sorry for the slight tardiness.
>> joo, how you going to determine by the TS itself how much you can change quotes?
I'll be spinning quotes all the way to the tomato. Until the adaptive TS squeaks: ".... and we're getting stronger!"
I don't know yet, but come on, would that be question number two?
I'll be spinning quotes all the way to the tomato. Until the adaptive TS squeaks: ".... and we're getting stronger!"
I don't know yet, but come on, will that be question number two?
Okay, number 2. Now a couple more things.
1) If you spin too much, you'll end up rejecting really robust, good systems (here's an idiot's dream: I have a pile of diamonds, and I throw them away, throw them away, throw them away...).
2. Another thing that hasn't been talked about here yet. If you want to do something similar to real series (and you'll have to!) - you'll look at ARIMA and other stuff (not my place, grasn is a specialist here). The problem is, you'll probably still get a martingale (again grasn's question: always?), even if it looks like a real series. And what's the point of testing a system on a martingale if there's a theorem of the Dub that forbids returns other than zero on a long horizon?
I'll be spinning quotes all the way to the tomato. Until the adaptive TS squeaks: ".... and we're getting stronger!"
Don't know yet, but come on, will it be question number two?
Heh, your question Alexey, knocked me off my feet, it was the middle of the night after all.
I didn't originally intend to rotate BP's stats. I was going to generate SVR with changing stat characteristics from the beginning of the row to the end, and then see where TC starts to bog down. Identify weak points of the system, and if possible to treat them. After all, this is the point of robustness - to work the system in a wide range of conditions.
The SVR is the first step towards a rubber woman. On the outside it looks like, but the soul is not.
===
For the record, the rubber woman was invented by Adolf Hitler
It is possible to generate a series based on the sampling distribution of the real series without estimating the distribution parameters. We visually distinguish on the real series (or take several series) trend sections, flat sections, channel sections, etc. according to the feeling. We draw a distribution histogram for each sample. When generating synthetics, we randomly select one of the histograms and use it to generate a series for some time (randomly), then change it, etc. It is even possible to control the phase change and their duration: for example, to give longer trend samples, or with a higher probability to give a flat sample after a trend sample.
This is an excellent suggestion, as an additional test of survivability.
1. the idea of synthetics was embraced by many (as here), and in the end many realised it wasn't the right idea.
2. a number of prices have a single statistical characteristic to guide us - sustained profits.
1. the idea of synthetics was embraced by many (as here), and in the end many realised it wasn't the right idea.
2. a number of prices have the only statistical characteristic to be guided by - sustainable profits.
Are you sure you understand what this is for?