From theory to practice - page 52

 
Alexander_K2:
... Where there is a geometric distribution there is no memory and there cannot be. ...

MOST

 

Look.

This is what Vladimir and I were talking about.

Again. The raw data - ticks were read at exponential intervals and the average value between two consecutive values was taken. The following picture is obtained:

Column D - real probabilities of tick increments.

Column E is the probabilities calculated from the t2-distribution.

Column F - probabilities calculated from geometric distribution (q^n)*p.

Don't you see that values in columns D and F are almost the same and the discrepancy is explained only by measurement error?

Or do you think that this mismatch is "memory"?

Files:
EURJPY.zip  6513 kb
 

Why this topic, why these 500 comments?

your research will show that most forex deals are 10 pips higher. what are you going to do with it?

you know that the next increment will be 10 pips. but you don't know in which direction.

First develop a trading system that allows you to trade with the knowledge about the gaps, and then investigate the gradients.

 
Максим Дмитриев:

Why do we need this topic, why do we need 500 comments?

Well, your research will show that the most frequent hand hand hand increases by 10 points, and what are you going to do with it?

You know that the next increment will be 10 pips, but you don't know in which direction.

First develop a trading system that allows you to trade with the knowledge about the gaps, and then research the gaps.

Yes, it has already been developed and I ran it today on 4 pairs at once so far.

The question is whether the market process is Markovian or non-Markovian. If it is a non-Markovian one, then historical archive tick data should be taken into account, if it is a Markovian one, then it should not. Well, we're bickering here...

 
Alexander_K2:

Yes, it has already been developed and I ran it today on 4 pairs at once so far.

The question is whether the market process is Markovian or non-Markovian. If it is a non-Markovian one, then historical archive tick data should be taken into account, if it is a Markovian one, then it should not. Well, we're bickering here...


how to trade a non-Markovian process?

 
Alexander_K2:

Look.

This is what Vladimir and I were talking about.

Again. The raw data - ticks were read at exponential intervals and the average value between two consecutive values was taken. The following picture is obtained:

Column D - real probabilities of tick increments.

Column E is the probabilities calculated from the t2-distribution.

Column F - probabilities calculated from geometric distribution (q^n)*p.

Don't you see that values in columns D and F are almost the same and the discrepancy is explained only by measurement error?

Or do you think that this mismatch is a "memory"?



Do some numerical experiments.

1) Generate an array with some distribution - these will be the "increments".

2) Create a process from these increments

3) Determine the regularities of the process

Do the same for other distributions -- beta-, bnom-, cauchy-, xi2-, exp-, gamma-, geom-, lnorm-, logis-, norm-, Poisson-, Student-, uniform-, or whatever else you want...

Do these processes have memory? Is it possible to make a profit on these processes?

 
Максим Дмитриев:

how to trade a non-Markovian process?


If my EA shows good results - then I will tell you in detail. OK? Because I've already said too much, and not enough to do. I agree with the critics on this point. But these are questions of principle! No?

 
Alexander_K2:


The question here is whether there is a Markovian or non-Markovian process in the market.


what do you mean "in the market"? there are 5000 stocks, futures.... You can also run them in a test to see "what process is there".

 
Alexander_K2:

If my advisor shows good results - then I will tell you in detail. OK? Because I've already said too much here, and not enough to do. I agree with the critics in this. But these are questions of principle! No?

I say "you figure out how to trade a non-Markovian process first, and then research financial instruments on what processes there are".
 
Alexander_K2:

Yes, it has already been developed and I ran it today on 4 pairs at once so far.

The question is whether the market process is Markovian or non-Markovian. If it is a non-Markovian one, then historical archive tick data should be taken into account, if it is a Markovian one, then it should not. Well, we are bickering here...

What conclusion have you drawn - Markovian or non-Markovian?