A study on the applicability of martingale using simulations of the coin game - page 7
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
Once again, for all readers of this forum and this particular thread:
Only trading strategies based on the analysis of a set of historical (integral) and current parameters can bring any positive result. All strategies based on the analysis of current parameters ONLY (current price, current variance, current correlation coefficient, etc., etc.), such as Bollinger bands, Martingale, all kinds of oscillators based on Fourier transforms - are doomed to failure.
So you didn't go on familiarization courses on retail Forex? Maybe, you should read at least client agreements of some companies and think, why so many companies consider arbitrage as a forbidden technique (for clients, for themselves they consider choosing the best course available on the market quite legitimate)? After all, ONLY current rates are analysed there... Why do companies defend themselves against methods that are "doomed to failure in advance", and not at all against analyzing integral parameters, do you think? Because of their stupidity?
I notice the word "martingale" is used several times in the thread. Freedom of speech, of course. And word creation. But yet the word is already busy, and in the theory of random processes, for which the trading method known as "martingale" is being explored here. Why create confusion? Wiki:
"Martingale in random process theory is such a random process that the best (in the sense of mean square) prediction of the future behaviour of the process is its present state."
P.S. There is also another meaning: "A martingale for a horse is not a means of rearing it, but a helper that allows it to keep its head in the right ..." - enough, it seems.
Greetings, Vladimir!
No time now - lots of work, even abandoned my branch for a LONG time. But, I read your post there - it's curious, for sure. Especially about reading data at certain intervals. I'll say right off the bat - I don't think it's at randomly chosen intervals, but exponentially distributed. In this case we come to a pure Markovian process. I have a little work on this topic. They seem to show a geometric law distribution of increments with p=0.5, which again says that without analysis of historical data our chances are strictly 50/50.
But, I need some time and experiments in this direction - I could be wrong with p=0.5. Maybe = it is only on my data from my DC?
Here, you have sown my doubts - no argument. :)))))
Greetings, Vladimir!
No time now - lots of work, even abandoned my branch for a LONG time. But, I read your post there - it's curious, for sure. Especially about reading data at certain intervals. I'll say right off the bat - I don't think it's at randomly chosen intervals, but exponentially distributed. In this case we come to a pure Markovian process. I have a little work on this topic. They seem to show a geometric law distribution of increments with p=0.5, which again says that without analysis of historical data our chances are strictly 50/50.
But, I need some time and experiments in this direction - I could be wrong with p=0.5. Maybe = it is only on my data from my DC?
Here, you have sown my doubts - no argument. :)))))
"not through randomly chosen intervals, but exactly through exponentially distributed" - how so? You still haven't clarified what you mean by those words. Perhaps you can tell us now?
I notice the word "martingale" is used several times in the thread. Freedom of speech, of course. And word creation. But yet the word is already busy, and in the theory of random processes, for which the trading method known as "martingale" is explored here. Why create confusion? Wiki:
"A martingale in random process theory is such a random process that the best (in the sense of mean square) prediction of the future behaviour of the process is its present state."
P.S. There is also another meaning: "A martingale for a horse is not a means of rearing it, but a helper that allows it to keep its head in the right ..." - enough, it seems.
haha)))) for some reason I immediately associate it with the word "marginal"
Link: https://ru.wikipedia.org/wiki/Экспоненциальное_распределение
I specifically wrote a program that generated pulses at time intervals that obeyed this distribution with lambda=1. And strictly on the arrival of such pulses I read the tick data. Absolutely all the pictures of increment distributions became perfectly smooth with correct proportions. In general - a very beautiful Markovian process. Just take the usual linear equation of price movement and that's it. But if we make a histogram of increments taken modulo in this case, we'll get a nice geometrical distribution with p=0.5 which proves that in this case we deal with "coin flip" described in this thread. So I abandoned this case...
Link: https://ru.wikipedia.org/wiki/Экспоненциальное_распределение
I specifically wrote a program that generated pulses at time intervals that obeyed this distribution with lambda=1. And strictly on the arrival of such pulses I read the tick data. Absolutely all the pictures of increment distributions became perfectly smooth with correct proportions. In general - a very beautiful Markovian process. Just take the usual linear equation of price movement and that's it. But if we make a histogram of increments taken modulo in this case, we'll get a nice geometrical distribution with p=0.5 which proves that in this case we deal with "coin flip" described in this thread. And I've abandoned that case...
Interesting... So you had exponentially distributed increments of reading moments. And what was being read out? The Ask increments, or the rate itself?
And a question about generating a pseudo-random sequence of steps distributed with a density of exp (-x). In https://habrahabr.ru/post/263993/ I read:
"It has already been said that for an exponential distribution you can take the logarithm of a uniformly distributed value and that you can make the generation faster. Since any exponential value is obtained from a standard value by dividing by the density, the generation can be done by the proverbial Ziggurat. In case you hit a tail, you can re-run the algorithm and add x1 to the obtained value:"
- Have you done so? No particular requirements to the pseudorandom number generator, is it all right? I want to see how the sample rate diagram will change with your method of reading data. I understand about the tail, it is a consequence of the independence of the distribution exp (-x) from the shift of the starting point.
Interesting... So you had exponentially distributed incremental moments of reading. And what was being read out? The Ask increments, or the course itself?
And a question about generating a pseudo-random sequence of steps distributed with a density of exp (-x). In https://habrahabr.ru/post/263993/ I read:
"It has already been said that for an exponential distribution you can take the logarithm of a uniformly distributed value and that you can make the generation faster. Since any exponential value is obtained from a standard value by dividing by the density, the generation can be done by the proverbial Ziggurat. In case you hit a tail, you can re-run the algorithm and add x1 to the obtained value:"
- Have you done so? No particular requirements to the pseudorandom number generator, is it all right? I want to see how the sample rate diagram will change with your method of reading data. I understand about the tail, it is a consequence of the independence of the distribution exp (-x) of the shift of the starting point.
1. The prices themselves were read and the increments were then calculated.
2. Yes, exactly so. I only took an integer part of the generated number and added 1. Thus, I obtained time samples of 1, 2, 3, ... ...seconds, distributed according to the exponential law.
It was a beauty... However, p=0.5 scared me off and I'm working now only towards investigating the combination of current parameters and average historical parameters. There are some results. I will finalise them and publish them in due course.
1. The prices themselves were read and the increments were then calculated.
2. Yes, exactly like that. I only took an integer part of the generated number and added 1. Thus, I got time samples of 1, 2, 3, ... ...seconds, distributed according to the exponential law.
It was a beauty... However, p=0.5 scared me off and I'm working now only towards investigating the combination of current parameters and average historical parameters. There are some results. I'll finalise them and publish them in due course.
Yes, interesting... One last question.
What is the reason for choosing lambda=1?