From theory to practice - page 679

[sibirqk]

Natalja Romancheva:
Dependence of profitability on the parameters of the ticking MAs used at the channel boundary to determine the probability of a rebound from the boundary.

Is the testing period long? Half a year, a year, five?

[Mihail Marchukajtes]

What can I say, Trickster is right as always. To be honest, I saw only one word COINTEGRATION I liked this tool in NS very much. I think this topic is interesting now, especially if you attach to it NS.... In general, the bomb would be...... By the way, please come to our fireplace where I will tell you one idea, about which I have already told more than once..... so you are welcome....

[Alexander_K]

Man, you guys, look at this!

https://demonstrations.wolfram.com/TheReturnDistributionOfTheVarianceGammaProcess/

Well, the market is 100% Variance Gamma Process!!!

There's some other additive added to the variance calculation and that's it!

And who knows the TS on this model!!!

[Violetta Novak]

Alexander_K:

Man, you guys, look at this!

https://demonstrations.wolfram.com/TheReturnDistributionOfTheVarianceGammaProcess/

Well, the market is 100% Variance Gamma Process!!!

There's some other additive added to the variance calculation and that's it!

And who knows the TS on this model!!!

I sent to Bulashev, he has a generalised exponential distribution through a Gamma function

[Violetta Novak]

Addition. For Brownian motion, the square of the sum of the increments equals the distance of the point from the initial coordinate, okay, so that's what you get when generating a random walk. But the ideal process doesn't have this distance, it = 0. What is the problem?

[Oleg Papkov]

Oleg Papkov:

End of the US session, start of the Asian session. Change in the forex market. Picking up the dough at the exchanges. Closing the banking day. Accrual of swaps on open trades. Number of deals falls sharply.

Volatility depends on the number of deals, or rather, on the average speed of increasing the number of deals, or rather, processed volumes in lots, and as a consequence, the increase of the average value of price increment. As if the "average temperature" on the instrument increases. The value and the number of spread jumps increase. At the beginning of the London session the number of deals for the period and the volatility increases sharply. At the moments of slackening of the market (change of sessions, etc.) the amplitudes of all frequency components of the "white noise" of the market just decrease to small insignificant values.

[Natalja Romancheva]

sibirqk:
Is the testing period long? Half a year, a year, five?

Period 2018.06.18-2018.10.18 and 1000ms delay in execution. A year is also possible.

Five is unlikely, there is almost no ticking history for such a period.

[sibirqk]

Natalja Romancheva:

Period 2018.06.18-2018.10.18 and 1000ms delay in execution. A year is also possible.

Over five is unlikely, there is almost no ticking history over such a period.

What is your confidence that this is not over-optimisation based on? If the testing interval is a year, with the same parameters - will the picture change much?

[Natalja Romancheva]

sibirqk:

What is your belief that it is not over-optimisation based on?

Any parameter selection is an optimisation or over-optimisation.

In this particular case, there is some rationale behind the selection, so there is hope...

Of course it's not as cool as the author of the thread, the rationale is empirical - there is no strict theory.

[Violetta Novak]

Alexander_K:

Man, you guys, look at this!

https://demonstrations.wolfram.com/TheReturnDistributionOfTheVarianceGammaProcess/

Well, the market is 100% Variance Gamma Process!!!

There's some other additive added to the variance calculation and that's it!

And who knows the TC on this model!!!

The link is very good, everything is clear, super