From theory to practice - page 381

New comment
 

A trade that took place last night:

AUDJPY. Profit +116 pips.

However, as you can see, the entry into the trade occurred ahead of time. The very "heavy tail" of the distribution, in which the Grail sits, was out of focus.

The reason - the distribution quantile was chosen from the Chebyshev inequality and was =3.5355, which corresponds to the 93% confidence level for multimodal distributions.

Not much...

Now the quantile =3.849, which corresponds to the 97% confidence level for the unimodal distributions from the Petunin-Vysokovsky inequality.

Sooner or later we will get to the right quantile. The grail, to put it simply.

 

Next trade:

Quantile = 3.849, which corresponds to a 97% confidence level for unimodal distributions from the Petunin-Vysokovsky inequality.



Currency pair EURJPY. The loss is -39 points.

And again, the trade was entered earlier, much earlier than the time... This is fucked up, gentlemen!!!!! We raised the quantile from 3.5355 to3.849! What's wrong with that?

We'll get to the bottom of this.
 
Alexander_K2:

This is fucked up, gentlemen!!!!! Let's get to the bottom of this.

What is there to sort out? If the market is predictable, then everything is clear and requires no comment. If the market is random, then it is free to go anywhere and anytime, and you must always remember that, not only when calculating distributions). The first commandment is that no one has promised anyone anything).

What's there to think about, you have to shake!

 

Let's look at what the quantile = 3.849, (97% confidence level for unimodal distributions from Petunin-Vysokovsky inequality) corresponds to.

We look at the quantile of the 99.99% confidence level for the Student's distribution at 14400 measurements (4 hours=14400 sec.).

It equals = 3.89168.

As you can see, working within the Student's distribution (read - normal distribution) it is almost impossible to achieve a profit. Everyone knows this, but I got more evidence of it on my deposit.

Therefore, let's move on to distributions with "memory", which describe non-Markovian processes.

Of course, first of all, here are these:


 
Alexander_K2:

That's fucked up, gentlemen!!!!! We raised the quantile from 3.5355 to3.849! What's wrong with that?

For the thousandth time, although no one will hear me again: quantiles for input are good, but it is much more important what happens afterwards.
If you really want to work with distributions, "what happens after" is described by a conditional distribution.
And by itself, taken alone, no distribution says anything about the presence/absence of memory, read the textbooks at last)
 
Alexander_K2:

Quantile = 3.849, which corresponds to a 97% confidence level for unimodal distributions from the Petunin-Wysokowski inequality.

What's wrong?Let's look into it.

Petunin-Vysokovsky should not be bothered off-topic for nothing. ))

One can see that trades are executed only on strong trends in the holy and righteous hope for an obligatory pullback.

Even Petunin would agree that this gloomy hope is too shaky and even anti-scientific if not illiterate at its root...

The most important thing is not to ignore the Russian proverbs, which say that if you make a fool to pray to God, then the cancer on the mountain will whistle. ))

 
bas:
For the thousandth time, although again no one will hear: quantiles for entrance is good, but it is much more important what happens after.
If you want to work with distributions, "what happens after" is described by a conditional distribution.
And by itself, taken separately, no distribution tells about presence/absence of memory, read textbooks at last).

What happens before is described by distributions, but what happens after is described by events). As an example of A_K2 you can read about the notorious Schrodinger's cat).

 

This class of distributions can include the Weibull, Xi-squared, lognormal, etc.

But let's start with the Maxwell-Boltzmann distribution, which describes the velocity distribution of molecules in a gas.

What is its quantile function?

Do I know?!?

All I know is that it has Pearson's asymmetry coefficient = 0.0854. That's what we'll use in the algorithm.

And the quantile...

Well, let's try to take the Chebyshev confidence level = 94%. Quantile = 4.0825.

Fit it in. Wait for it.

See you later.

 
Alexander_K2:

As you can see, working within the Student's distribution (read - normal distribution) it is almost impossible to achieve profit....

But, let's start with the Maxwell-Boltzmann distribution...

Well, let's try to take the Chebyshev confidence level = 94%. Quantile = 4.0825....

Fit. Waiting.

See you later.

She's got half a dozen points on her;

♪ she twirls her glasses around and around ♪

She presses them against her skin and puts them on her tail,

and then she smells them, and then she licks them;

The glasses have no effect.

 
I am literally digging my teeth into the ground and crawling towards the coveted Grail. No one and nothing can stop me.
1...374375376377378379380381382383384385386387388...1981
New comment
Reason: