From theory to practice - page 273
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 think I've got it.
Step1 - Make your own beat generator, with random pauses between beats like that distribution. At each cycle - take current bid/ask price, make a time series of them.
Step 2 - I don't understand it yet, I need to think about it. I won't be able to trade such time series, it's already HFT.
Step 2 is most likely a wrist with a random process attached to it. When the process exceeds its average limits, it trades back to the wrist... I guess that's how it is in simple terms...
Am I correct in assuming that the K-factor of the Erlang distribution is a good indicator of the integrity of dealing?
More likely about the quality of its infrastructure.)
Am I correct in assuming that the K-factor of the Erlang distribution is used to judge the fairness of the dealing?
A bit of specific humour:
I think the more k, the better for us)))) because if k=1, then it is a pure exponent, a Poisson process, in general, it is very bad. This is when you start comparing between very bad and so, average, maybe something for a profit))))
I think the more k, the better for us)))) because if k=1, then it is a pure exponent, a Poisson process, in general, everything is very bad. That's when you start comparing between very bad and so, average, maybe something for a profit))))
For neural networks and prediction tasks - yes, the more k, the better.
I stand by my opinion - parameter k should be selected forcedly, depending on model used by trader. I have k=1.
There is another reason why I need the K=1 parameter, i.e. Poisson flow emulation.
And this reason is the need for a mysterious trend/flux parameter.
I am absolutely convinced that this parameter is the system's non-entropy coefficient calculated as the ratio of the current probability distribution parameter to the Gaussian distribution parameter.
So, the hypothesis is that with K=1, i.e. when simulating a real tick flow by a Poisson process, the non-hentropy coefficient of the system will be calculated correctly, in other cases it will not.
For neural networks and forecasting tasks - yes, the more k, the better.
I stand by my opinion - parameter k should be chosen forcibly, depending on model used by trader. I have k=1.
Alexandre, you don't say what you need k=1 for, namely to calculate the difference between the current state and the state without any consequence, on this difference you calculate the trading ratio that shows the trend between sessions.
Alexander, you don't specify what you need k=1 for, namely to calculate the difference between the current state and the state without any aftereffects, this difference is used to calculate the trading ratio, which shows the dynamics between sessions.
That too, of course.
So this difference between the real tick flow and its emulation is needed for 2 things:
1. to calculate trading intensity (resolved)
2. to calculate non-entropy (in progress...)
There is another reason why I need the K=1 parameter, i.e. Poisson flow emulation.
And this reason is the need for a mysterious trend/flux parameter.
I am absolutely convinced that this parameter is the coefficient of nonentropy of the system calculated as the ratio of the parameter of the current probability distribution to the parameter of the Gaussian distribution.
So, the hypothesis is that with K=1, i.e. when simulating a real tick flow by a Poisson process, the non-hentropy coefficient of the system will be calculated correctly, otherwise it will not.
Is it possible to theoretically tune the existing real process, to approximate the normal distribution, say, as in the case with k--> to infinity in the Erlang distribution, but not to take it as a whole, but as a part where a part of this distribution tend to normal, projecting on the whole process, as if replacing it, to calculate the trade intensity ratio correctly, because currently it is not quite correct exactly because of the Erlang "tail" where the exponent with Cauchy still sits. After all, all your success, as far as I understand it, can be attributed to the application of this coefficient. Correct me if I am wrong.
Is it possible to theoretically adjust the existing real process close to normal, say, as in the case of k--> to infinity in the Erlang distribution, but not to take it as a whole, but as a part, where part of this distribution tends to normal, projecting on the whole process, as if replacing it, to correctly calculate the rate of trading intensity, because at present it is not quite correct exactly from the Erlang "tail", where the exponent with Cauchy still sits. After all, all your success, as far as I understand it, can be attributed to the application of this coefficient. Correct me if I'm wrong.
To be honest, I haven't considered cases of K > 1, but I probably should have. There must surely be something there.
In my excuse, I'll say that my goal was to solve the problem as quickly as possible - to start replenishing my purse immediately.
Once I realized that only working in Erlang's flow (already at K=1 !!!!!) gave good results, I somehow started to think more about money than about physics and mathematics. Alas, human weaknesses, fuelled by my father-in-law and wife, are just as intrinsic to me...
So, K>1 cases, I hope someone will consider and get not just good, but unrestrained profits.
To be honest, I haven't considered cases of K > 1, and I probably should have. There must surely be something there.
In my excuse, I was aiming to solve the problem as quickly as possible - to start replenishing my purse immediately.
Once I realized that only working in Erlang's flow (already at K=1 !!!!!) gave good results, I somehow started to think more about money than about physics and mathematics. Alas, human weaknesses, fuelled by my father-in-law and wife, are just as intrinsic to me...
So, cases with k>1 will hopefully be considered by someone and get not just good, but unrestrained profits.
At k--> to infinity we will get analogue of normal distribution, I suggest to do otherwise, not to look for such k, but here and now to transform residuals, which we have in a tail, we have not received them simply as a result of a delay on the way.
It is possible that you and I are talking about the same thing, only you from the non-entropy side and I from the Erlang side.