Adaptive digital filters - page 13

 
eire:
Prival:

I have not been able to find any real application of the Kalman filter to quote analysis.

Zizhelev V.I. "Optimal strategies for making profit on the FOREX market and the securities market". I looked through it diagonally. As I understand it, the book considers an extreme control system using Kalman filter. And the results are given. It may be useful.


Thank you. Where can I get it? Is there any way to download it from somewhere?

 
Prival:
rsi:
Prival, you also consider the broker's offer not the true price, but his expert estimate. Well, dear ones, wait until the broker's price
Otherwise you will not be able to make a deal at the price calculated according to your forecast!
Yes I do. And when you set the TP level, you make a prediction that it will get there, that "true" price. And I want my prediction to be more accurate. I don't want to set TP of 100 pips or 3 times the SL.
So I didn't put a question mark - I see what you're counting. It's a somewhat unexpected approach for me. You have to find a "slow" broker and - no problem, arbitrage in a single DC. I do not remember who described it even easier: if you find a "slow" broker and a "normal" one - and that's all - the price of the latter is taken as the true one, and you just have time to place orders. However, he says he hasn't been working for a long time...

About the illustrations in the article: the input velocity range does not coincide (does not even overlap) with the output velocity range. Or am I missing something?
 
Prival:

Well Mathematician was scary :-)

I've got scarier pictures than that. The red dots are the radar input (measurement stream) Fig.1. But Fig 2 is the output of the Kalman filter (what he selected from that stream). One of his successful applications.

Fig.1

Fig.2

So the dog in filtering has eaten a bit.

Z.I. These are graphs from my articles, to be published this month in Fazotron (Hardware-software complex for determining the speed of small and hypersonic objects) and in Vestnik of Computer and Information Technologies. All open printing.

Prival, that's impressive!!! Is there a picture of how the rocket actually flew?
 
eire писал (а): Zhizhelev V. I. "Optimal Strategies for Profit Extraction in the FOREX and Securities Market".
Here it is: http://forex.360.net.ua/ukraine-upravlenie-kapitalom-i-riskom-forex-articles/v-zhizhilev-optimalnie-strategii-izvlechenija-pribili-na-rinke-forex-i-rinke-tsennih-bumag. html
 
rsi:
Prival:
rsi:
Prival, you also think the broker's offer is not the true price, but his expert judgement. Well, dear friends, wait until the broker's price
Otherwise, you will not be able to make a deal at the price calculated according to your forecast!
Yes I do. And when you set the TP level, you also make a prediction that it will get there, that "true" price. And I want my prediction to be more accurate. I don't want TP of 100 pips or 3 times SL.
So I didn't put a question mark - I can see that you're counting. For me it's a somewhat unexpected approach. You have to find a "slow" broker and - no problem, arbitrage in a single DC. I do not remember who described it even easier: if you find a "slow" broker and a "normal" one - and that's all - the price of the latter is taken as the true one, and you just have time to place orders. However, he says he hasn't been working for a long time...

Regarding the pictures in the article: the range of input velocities does not coincide (it doesn't even overlap) with the range of output velocities. Or am I misunderstanding something?

Why go up against a broker. Let him live. The essence is the other thing, you need to know (measure) what is now as accurately as possible to make a forecast as far as possible on the time axis, so that it would be more accurate. Although if I have identified a gap (it appeared on the foreign market there), and the broker fell behind and allows to enter at the price before the gap, why not. The broker should be careful, he should get the money from him :-)

The images are indeed different (they are simply pulled out from various experiments). Observation, a very good quality. I had a whole pile of experiments, with mush on the way in and a nice clear trajectory on the way out (no way to handle that mush). I got it, just a month ago made the device (and algorithm) in the State Register of military measuring instruments. A whole year tortured until it has confirmed all the stated parameters. The benchmark was not, just like here :-). Nobody knows this "price=true tarektory", but want to find out :-).

And they also want to predict. It is difficult but I think it is possible. You just have to try to do everything right, and then maybe it will turn out well, but if you just do rubbish, then nothing good will come out :-). I was taught that way. I apologize for the ... but words can't be taken out of my mouth :-)

 
Prival:

State Register of measuring instruments for military use. Tried for a year until it confirmed all the stated parameters.


You got very carried away - you are revealing Russia's military secrets. You could be tracked down and kidnapped....
 
Prival:
Sart:
Prival:

State Register of measuring instruments for military use. Tried for a year until it confirmed all the stated parameters.


You get very carried away - revealing Russia's military secrets. You could be tracked down and kidnapped....

There's nothing wrong with that, the registry is accessible to all. And so is an open seal.
Too bad it's not! Because some people already wanted to make some easy money! :)
 
Prival:

The pictures are really different (just plucked from different experiments). Observation, very good quality. There was a whole bunch of experiments, mush at the input, clean beautiful trajectory at the output (no way to deal with this mush). I've got it. Just a month ago made the device (and algorithm) in the State Register of military measuring instruments. A whole year tortured until it has confirmed all the stated parameters. The benchmark was not, just like here :-). Nobody knows this "price=true tarektory", but want to find out :-).

And also to predict. It's difficult but I think it's possible.


Prival, as far as I understand it was all built using a Kalman filter. And to use it you need a signal model. In this case Kalman filter outputs that component of data stream which corresponds to this model. Apparently, the model used in your hardware-software complex quite corresponds to the real motion of the object. This is what the success of the Kalman filter in this case is based on. But here at least the laws of dynamics and aerodynamics can be used to build the model.

And what about the price? Where do you want to derive a model for its variation? What do you want to base it on? And one more thing. Whatever the model is, the Kalman filter will give us a prediction. What about a criterion for its reliability? Or should we estimate it only on the basis of historical estimates?

 
rsi:
komposter:
rsi:
Well, dear ones, wait until the broker price equals the true price - otherwise you will not be able to make a trade at the price calculated by your prediction!
Why? We will open towards the "true" price. The "broker's" price will still come to it (+/- a couple of pips).
So you've already built an infallible prediction of the "true" price. Congratulations and good luck!

komposter, sorry for being abrupt - I must have been in a hurry. This problem statement ("let's go by the true price prediction") coincides with most trading tactics (or strategies - doesn't matter). The only difference is in the construction of the forecast. If I say that the forecast should be built from the broker's price, then your team, which does not accept that price as true, argues: no we will build it differently. But if both sides admit that with +\- a couple of pips accuracy these forecasts should coincide, then what are we arguing about? We could argue about methodology. Mine is clearly simpler - there is no need to doubt the veracity of broker quotes, while yours should be questioned and verified with information from additional (trustworthy, paid, just multiple, etc.) sources. But "why pay more if the result is the same?" (c) (I don't remember which detergent commercial).