That's interesting - page 17
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
The same model parameters will remain on the shuffled data as before shuffling.
(EURUSD + Const1) * (GBPUSD + Const2) / (AUDUSD + Const3) / (NZDUSD + Const4) * USDJPY / USDCAD.
Market models consider additive and multiplicative mixing.
Please give me a link to where you read that. The idea itself is good, but as usual you're understating the context or just making up a bunch of your own.
Any operation on time series implies its own invariants. And this "shuffling" has them too.
Send me the link where you have read it. The idea itself is good, but, as usual, you are not providing the context.
Any operation on time series implies its invariants. And this "mixing" has them too.
Didn't keep track of where I read it. But it seems logical. Initially I suggested that a study of a set of financial instruments should be carried out.
The implemented method of analysis, which I suggested more than a month ago, bypasses mixing - it does not affect the results.
There are some regularities in the market. Obviously, the fact that we have mixed some data with each other does not make the patterns disappear.
Mixing has to be reversible, which is why we are mainly talking about additive and multiplicative types only.
And if you analyse only one financial instrument. You can mix up the other one to such an extent that it will show nonsense. That is why I say that it is necessary to analyse the totality.
ah, you mean the same model parameters. OK, if the market is a set of sinusoids with known parameters, mix randomly - will we get the same parameters on the new data?
It is my deep conviction that this is a dead-end road. You need to analyse the processes behind the "pieces" of the price. Then the model will be "physically meaningful", reflecting at least the essence of things. Not just descriptive. The way you have it now, it exactly describes (tries to do so) the "picture" of the price. There is no "physics" behind it. And it should be. Why, because the price doesn't fully describe the state of the market. Looking only at the price, you have a small piece of information that looks like a nasty martingale. And then you'll be banging your forehead against the Oak, because you know that no matter how you divide a martingale into "pieces" it will be a lump. That said, mind you, the market processes themselves are not martingale. That's how it is.
That's exactly what I'm doing, i.e. the main challenge is identifying the process that drives the "chunks". So far on the time line. So far, so good. If it adds up in the long run - I'll plug in the influencing factors, but again, as a time series. Perhaps I will add the Bayesian logic in the estimation of individual events and their influence on the series. But it is unlikely that in the near future I will build complex fundamental models, i.e. real "physics". I would very much like not to do this, because it's all complicated.
. No need for any fine structures.
I am tired of...
All you need to know in the market is when to buy cheap and when to sell expensive. That's it.
"that's the reindeer." thanks. I didn't know that. It's as simple as that.
PS: but why do you always put a full stop at the beginning of the paragraph? I hope it's not a trade secret?
I worked with MQL long time ago. Then Yuri promised to help me.
When could I have promised such a thing? (surprised mug).
Sergei, I can hardly help you now. For a month now, I haven't been able to get myself to finish one simple task. Not a line of code in a month. That's how bad it is to do Hirst. :-((
I think Alexey will do it better than me. You should only take into account that every broker has exact time when it starts quoting and when it stops. Taking that into account it is very easy to distinguish the start of trading on Monday (or Sunday) from the end on Friday.
The historical data can be treated the way as suggested by hrenfx . Only holidays should be filled not with previous value (like holes), but with zeros. And the program should be written so that it does not process zeros.
I think you will get the same parameters.
no. I fundamentally won't get the same sine waves after mixing (imagine that I perfectly defined the "market" before mixing). If it doesn't work on such simple objects, it is unlikely to work on anything more complex.
Maybe you're confusing things a bit. Stirring is usually used to check the patterns found, and that with some caveats.
That's exactly what I'm doing, i.e. the main task is to identify the process that drives the 'chunks'. So far, on the time line. That's how it is for now. If it adds up in the long term, I will connect influencing factors, but again, as a time series. Perhaps I will add the Bayesian logic in the estimation of individual events and their influence on the series. But it is unlikely that in the near future I will build complex fundamental models, i.e. real "physics". I would very much like not to do this, because it's all complicated.
. And how do you want to solve the problem of identifying "chunks", on a time series, if the time series is martingale?
boring...
. Couldn't get past it and not kick it.
"there he is - the reindeer" (C) Thank you. Didn't know that. It's as simple as that.
. An examination of the revelations of a group of not-so-lucky traders, fairly solid in numbers, indicates that it is. The rare tales of successful traders with a sophisticated mathematical apparatus remain unconfirmed. By the way, it's not just my observation, someone else on the forum mentioned it.
. But, of course, no one can prevent you from being the first highly scientific successful trader. Well, maybe the second isn't a bad thing either. The main thing that it should happen before you are 80. :)
PS: Why do you always put a full stop at the beginning of the paragraph? I hope this is not a trade secret?
. I don't know, I wonder myself every time.
Not a single line of code in a month. That's how bad it is to do Hirst. :-((
I've wasted a lot more time on it. And I haven't even begun the work on our discussion yet. :о(
When did I ever promise such a thing? (gasps)
I must have misinterpreted this:
Come up with some other torture that's not so sophisticated. :-))
to Mathemat
I think Alexey knows better than me.
Alexey, ... give me a script to check all sorts of nonsense
(I hope Mishek won't claim the copyright for this ... geez, who is it?).
PS: though I don't think you have the time. So, fuck that shit.
No. I fundamentally won't get the same sinusoids after shuffling (imagine that I perfectly defined the "market" before shuffling). If it doesn't work on such simple objects, it's unlikely to work on anything more complex.
Suppose we have some data - a finite BP of length N. The point of the model is to describe the behaviour of the VR with parameters much smaller than N. Otherwise, any VR can be represented as N sinusoids, and that, of course, would be a bussymessel. Regularities in VR are always described by fewer parameters than the original N.
In one thread I talked about the notion of "information" as the minimum number of parameters describing BP. But I won't go back into the terminological discussion.
Regarding your quote above. A market model as a set of sine waves is not a model. So anything can be said to be a set of sine waves and you won't be wrong. A tree is also a set of sine waves. Only to describe a tree in sine waves, you'd need to use as many sine wave parameters as you would if you wanted to describe the tree in polynomials.
So a "set of something" is a model, but a lousy one.