Distribution of price increments - page 6

 
Alexander_K:

If this conditional variance is exceeded, a trade can be made.

The main question - which is common to all probability-based systems - is in which direction to trade.

If we assume the system adheres to the rules, then we should enter against the trend, believing that price will return to the boundaries.

If we assume that the longer some rule works, the higher is the probability of failure (by the way, it is nice to count MTBF), then we should enter in accordance with the trend, i.e. with the idea of continuing of border breakdown in even more evil way. Then comes the idea that one can calculate conditional probabilities of a breakout of more distant boundaries when the current boundary is breached. And so on - in terms of complicating the model and making (profitable) decisions.

 

Of course, this is the most important question.

I think in the case of a non-Markovian process we should trade against the trend, while in the case of a Markovian process we should trade with the trend.

Next week I will investigate the probability distribution of tick arrival time - let's see what it is for different pairs.

If it is nonexponential - then the processes are not Markovian and vice versa.

I will post the results on the forum.

 
Alexander_K:

I think that in the case of a non-Markovian process you have to trade against the trend, while for a Markovian one you have to trade with the trend.

Can you explain why the fact whether the process is a Markov process or not should affect the performance of trend or counter-trend trading? So far it reeks of scientific nonsense.

 

Alexander_K

Distribution of price increments

No problem).

On TF1m in points.

X-increase, Y- probability.

 
anonymous:

Can you explain why the fact whether the process is Markovian or not should influence the efficiency of trend or counter-trend trading? So far it reeks of scientific nonsense.

I agree, it is too early to draw such conclusions without a serious evidence base. But there is a long way to go to collect it, and I would like someone along the way to say "no this is not right, I have already checked" or "yes it seems to be right". So far, alas, there are no such people...

For now - this is my hypothesis and I am working on proving or disproving it.

PS. Promised my daughter to look into this topic - it's not like I can refuse her :)))))

 
Yuriy Asaulenko:

No problem).

On TF1m in points.


What on the axes is a variable, a unit of measure?

It's not clear yet.

 
Renat Akhtyamov:
What on the axes is a variable, a unit of measure?

It's not clear yet.

As usual in distributions. X-increase, Y- probability. TF1m 3 months. ~54000 mins.
 
Yuriy Asaulenko:
As usual in distributions. X-increase, Y- probability.

I see. An increment in the size of the spread is the most likely.

Everyone who bought or sold will definitely notice a price movement on the spread.

Possible calculation error.

Most likely the +/- spread is closer to 0.5 rather than 0.05 in probability.

 
Renat Akhtyamov:

I see. An increment in the size of the spread is the most likely.

Everyone who bought or sold will surely notice the price movement on the spread.

Could be a calculation error.

Most likely the +/- spread is closer to 0.5 and not 0.05 in probability.

You are wrong. The algorithm is standard, and not mine, but from a professional program. Like MathCad, but different. It's done in 3 keystrokes).

Sum of all increments=1.

 
Yuriy Asaulenko:

You are wrong. The algorithm is standard, and it's not mine, it's from a professional program. Like MathCad, but different.

Sum of all increments=1.

I just drew a conclusion based on logic.