Market phenomena - page 39
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
I'm just timidly asking: don't goal-directed systems work at all?
... and - what is governance: necessarily goal-setting?
Sorry it's not Friday
There are no classifications or long tails in the economy.
There is the production of goods and services which have inertia - trends
In markets there are lots of commodities - oversold and lots of money - overbought --- levels
All this passes through the psyche of the crowd (which didn't exist under socialism) and noise appears --- volatility
We trade trend, levels and volatility. You can also trade risk.
What do you mean?
Oh!!! So you suddenly woke up an econometrician. Good morning, econometrician. And you have started to write about something that you were asked about at the beginning of your econometric section and about which you have not written one clear word.
Well, tell me how does all this relate to your model, where you filter, seemingly without understanding what you want to get out of filtering, and moreover predict the filter by methods that cunning econometricians have stuck to themselves.
PS: "fundamentals" is a good thing, and indeed it has all the physics in it, but not what you wrote. In my secret laboratories I am developing a secret secret weapon to take over the world based on the foundation (kind of a joke) :o)))
Farnsworth, sorry if I'm wrong, I don't have time to read the whole thread. If I understand you well, are you talking about Markov Switching Models?
Unfortunately, the transition process is not Markovian. Taking my basic model, I use Bayesian networks to describe the transition between processes
.... no time to read the whole thread..... just can't guess what kind of subprocesses you use?
do i have to rewrite the whole thread again? no, and i don't have much time for you.
Another phenomenon is long-term memory.
For example, if we take the history of the EURUSD since 1999 on H1 and check the chi-square of the pair's returns, we see that in the range of "distances" between bars 10 and 6000, in about 90% of cases the current bar depends on the previous bars. 90%! At distances between bars of more than 6000 such dependences occur less and less frequently, but still occur!To be honest, I was stunned by this "discovery" as it directly shows that the Euro has a very long term memory. On EURUSD H1 6000 bars is about a year. This means that there are still bars among the hourly bars from a year ago that the current zero "remembers".
... This result is still purely theoretical and of no practical importance. Nevertheless, it clearly shows that for those who are looking for something, not all is lost.Hi all!
Alexey, it's been half a year since your post was written. Any progress in this direction. Very interesting.
Suppose 90% of bars are statistically interdependent in the area of a few thousand counts... Have you estimated the degree of dependence?
How do you measure it at all... is unclear. Linear dependence - clearly, it is the angle of the slope of the straight line drawn by least squares through the cloud of returns of the time series:
This is for the Eurobucks watch. You can see that the blue line lies on the horizon, i.e. the linear correlation coefficient between adjacent bars is zero.
As I understand the presence of non-linear correlation will lead to this picture:
In other words, you can visually see a non-linear dependence of the amplitude of the candle on the amplitude of the previous one. I wonder if it is possible that there is a non-linear dependence and it is not visually visible in the cloud, although, the chi-square of the pair return indicates its presence?
Hello, everyone!
...
Well done, Sergey!
Pleasure to meet you. I admire your ability to work. I am stagnant - I need new ideas.
I admired Pirat's performance at the Automated Trading Championship 2011:
The person traded Eurodollar with a fixed lot, i.e. the result is not obfuscated by MM. Several hundreds of transactions. It is something. In short, it is possible, if you try hard enough!
I plotted the distribution of bribe values using his data:
And I tried to reconstruct his trading algorithm on this bar chart.
I'll show you a picture...
Well done, Sergey!
Pleasure to meet you. I admire your work ethic.
Oh, come on, Serega. I just got out of the doldrums myself. Time is very short, unfortunately. :о(
I'm in stagnation - need new ideas.
Get busy with one of the "phenomena".
In particular, I wanted to post some phenomena related to different approaches. There is a direction in mathematics, "asymptotic random walk analysis", the main goal of which is to study processes with thick tails. In this direction the most interesting researches (from practical point of view) are connected with studying deviations of trajectories of realization of such processes.
Somehow I wanted to see a quote without "fat tails", i.e. to perform relatively simple filtering (classification) that would remove spikes, bursts, "weird", "unnatural" accelerations of trajectory deviations in extremely short sections, in general, everything that literally sits in these "tails".
You get the alpha process, this process is perfectly described by the tools of stochastic financial mathematics. And such processes, mainly growth processes, you will get on all the quotes. The complex analysis of all alpha processes for quotes is curious. Very curious. Just don't make faa mistakes, let the creative develop.
Try looking at kotir without fat tails. Just give it a try.
I was fascinated by Pirat's performance at the Automated Trading Championship 2011:
Yes, there are interesting algorithms, but what's bad is the lack of consistency.
Take up one of the 'phenomena'.
Try looking at kotir without fat tails. Just try it.
No problem.
What are we looking at. Eurobucks minutes or larger TF?
No problem.
What we're looking at. Eurobucks minutes or larger TF?
OK, briefly:
(1) For a start EURUSD, experience says that you should watch something small, maybe not more than an hour (or maybe 15 minutes or better), the "phenomenon" is not likely to manifest itself in a 24-hour period
(2) You need a "fttration" criterion. I have used several. The simplest one- everything which exceeds 1/2/3 of RMSP of increments - has a tail, from memory it works as 3*RMSP.
(3) I sorted increments into two threads, didn't fill up holes in every thread. Assumed that a stream is interrupted.
(4) separately collected statistics on "switch" or "hopping" between these threads