From theory to practice - page 1050
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
On the subject, Kolmogorov was succeeded by Einstein and someone else.
Einstein and his wife were sort of developing theories))))
Einstein and his wife were sort of working out theories))))
No, they were Curies.
No, it was the Curies.
I saw a film about him. I think his wife was lame and they met at the institute. And she had, like, 50 percent of the discoveries.
Tomorrow, 14 March 2019 will mark the 140th anniversary of the birth of Albert Einstein, the man who apparently made the most significant changes to our scientific understanding of the world. In terms of his contribution to science, Einstein is probably the greatest scientist of all time. In the 1905 "year of miracles" alone, he created the special theory of relativity, laid the foundations of quantum mechanics in the hypothesis of quantisation of electromagnetic radiation, and published work explaining the mechanism of Brownian motion, which confirmed the molecular structure of matter. Einstein was a Nobel laureate in physics, author of about 300 scientific papers.
Einstein's first wife Mileva Maric is Serbian by nationality, and her role in her husband's scientific successes is a folklore fantasy based on working together on his degrees. Certainly the young Einstein discussed his ideas and plans with his wife and perhaps she advised him on something, but no more. She never had any scientific publications.
Yep....
The topic has turned to rubbish - no research, no coveted formula... Too bad!
So far I've done a little experiment.
I took series generated by Doc - with Gaussian distribution for incretals, on the basis of which it's convenient to simulate Wiener process (aka Brownian motion).
So - both for initial increments and for thinned ones (in every 2nd, 3rd, etc. values) we ALWAYS have ONE and the same Gaussian distribution - with the same expectation and other moments of a random variable. This confirms the invariance of Brownian motion as a classical stochastic process with respect to time and its self-similarity.
But if we take a market tick series of quotes and start to thin it out, then for each case we obtain DIFFERENT probability distributions, i.e. the market is NOT self-similar and thinning (for example OHLC M1) leads to a distortion of the process and loss of important information.
Tick quotes are inapplicable because of their different number in different brokerage companies.
Deadlock. Is it a deadlock? I think - to obtain a quote flow that has to be seen and handled, it is important to learn how to thin out the initial tick quotes correctly, showing a false quotes from brokerage companies and at the same time not losing information with unnecessary data loading, as it was in the OHLC M1 case.
How to do it? Well, of course, by converting the tick stream to Erlang of a certain order! I followed this way some time ago, then gave up, and now have decided to go back again.
In competent thinning is the salt and power of the Grail.
Yep....
The topic has turned to rubbish - no research, no coveted formula... Too bad!
So far I've done a little experiment.
I took series generated by Doc - with Gaussian distribution for the incretals, on the basis of which it's convenient to simulate a Wiener process (aka Brownian motion).
So - both for initial increments and for thinned ones (in every 2nd, 3rd, etc. values) we ALWAYS have ONE and the same Gaussian distribution - with the same expectation and other moments of a random variable. This confirms the invariance of Brownian motion as a classical stochastic process with respect to time and its self-similarity.
But if we take a market tick series of quotes and start to thin it out, then for each case we obtain DIFFERENT probability distributions, i.e. the market is NOT self-similar and thinning (for example OHLC M1) leads to a distortion of the process and loss of important information.
Tick quotes are inapplicable because of their different number in different brokerage companies.
Deadlock. Is it a deadlock? I think - to obtain a quote flow that has to be seen and handled, it is important to learn how to thin out the initial tick quotes correctly, showing a false quotes from brokerage companies and at the same time not losing information with unnecessary data loading, as it was in the OHLC M1 case.
How to do it? Well, of course, by converting the tick stream to Erlang of a certain order! I followed this way some time ago, then gave up, and now have decided to go back again.
Proper thinning is the salt and power of the Grail.
Maybe before subtracting and thinning you should learn how to add up the flows?
Imagine that you receive 2 quote streams from two brokerage companies and generate your own. And see what properties of the original streams remain as a result and what addition options depend on.
So, while I'm not trading, I'm just speculating theoretically.
And theoretically, we must, by all means, reduce the market process to the Ornstein-Uhlenbeck process, which guarantees a return to the mean.
Its distinguishing features are stationarity, stable and infinitely divisible distribution of increments and exponentially decreasing ACF.
There is an opinion that similarity of such a process will be observed in the sliding time window = 24 hours, in the Erlang's quote flow of a certain order.
Next week I will try to get back on track and show you how to properly thin market time series.
Stay awake, children! The grail will be found, full stop.
Yep....
The topic has turned to crap - no research, no coveted formula... Too bad!
So far I've done a little experiment.
I took series generated by Doc - with Gaussian distribution for incretals, on the basis of which it's convenient to simulate Wiener process (aka Brownian motion).
So - both for initial increments and for thinned ones (in every 2nd, 3rd, etc. values) we ALWAYS have ONE and the same Gaussian distribution - with the same expectation and other moments of a random variable. This confirms the invariance of Brownian motion as a classical stochastic process with respect to time and its self-similarity.
But if we take a market tick series of quotes and start to thin it out, then for each case we obtain DIFFERENT probability distributions, i.e. the market is NOT self-similar and thinning (for example OHLC M1) leads to a distortion of the process and loss of important information.
Tick quotes are inapplicable because of their different number in different brokerage companies.
Deadlock. Is it a deadlock? I think - to obtain a quote flow that has to be seen and handled, it is important to learn how to thin out the initial tick quotes correctly, showing a false quotes from brokerage companies and at the same time not losing information with unnecessary loading, as it was in the OHLC M1 case.
How to do it? Well, of course, by converting the tick stream to Erlang of a certain order! I followed this way some time ago, then gave up, and now have decided to go back again.
Proper thinning is the salt and power of the Grail.
Alexander, you call yourself a physicist, but operate solely with statistics. Where's the physical approach? Where is the physics?
Alexander, you call yourself a physicist, but you operate solely on statistics. Where is the physical approach? Where's the physics?
The power of recursion is unstoppable...
with such a simple question you take the discussion back a thousand pages :-)