Machine learning in trading: theory, models, practice and algo-trading - page 1474
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
In the other thread, a person who traded at CME says that the trading robots there trade with each other and make up fictitious volumes. And the real volumes of real buyers / sellers are many times lower. Now I understand why the volumes are only noise. Real volumes might be useful, but there's nowhere to get them.
Forex tick volumes in MT correlate with real volumes
look at the tick volumes, they repeat the structure from day to day, i.e. in the middle of the trading day they increase, then they drop
when the number of participants (volumes) increases, it not only needs to quickly execute all bids, but also to guess where the crowd is going - if MM guesses, then the trend is there, if not - a hairpin to restore the balance between buyers and sellers.
@Vizard_ wrote correctly - test your model in the Russian markets, you'll see if it immediately sees the volumes or not, but if I'm not mistaken, the expiration date of the contract also means a lot.
With M1 you can work all your life and everything passes. On a uniform timescale, there is pure SB with the corresponding result.
It is necessary to work in a non-linear coordinate system. But, it is not necessarily ticks :)
Result:
Learn, daddy, while I'm alive.
Where to learn? It's unreal to read your whole thread. Is the essence of the method described somewhere, in one place? A blog would be convenient.
1. Find anything that is traded in 1 place.
2. Make sure the real volumes in your "model" are really affecting.
3. Dokuchi - if you have mm, try to figure out his algorithm, but you won't understand it without the glass))))
Logical approach, I will try to test the MICEX
Forex tick volumes in MT correlate with real volumes
look at tick volumes, they repeat the same structure from day to day, i.e. in the middle of the trading day they increase, then they drop.
When the number of participants (volumes) increases, it not only needs to quickly fulfill all the requests, but also to guess where the crowd is going - if MM guesses, then the trend is there, if not - a hairpin, to restore the balance between buyers and sellers.
@Vizard_ wrote it correctly - test your model on the Russian markets, you'll see if it immediately sees the volumes or not, but if I'm not mistaken, there the expiration date also has a lot of meaning
If MM guesses, then the trend is in the direction of the buyers and sellers.
I wonder about the ratio between the bar height and the volume traded on it. There is some information about it on the cluster. But I'm afraid the tick and real volumes will be very different in comparison
Logical approach, I will try to test the MICEX
Well, the increase in total volume and just by time can be predicted without the tick.
I'm interested in the correlation between the height of the bar and the volume traded on it. There's a clasterdelta about it. But I'm afraid that the tick and real volumes will be very different in comparison
MOEX is in MT5, you can test it there faster.
About volumes and clusterdelta, etc. - There will be differences, but the correlation is more than 90% between the spot and the futures, you are interested in the trend, not in specific numbers.
Not bad!
Where to learn? It's unreal to read your whole thread. The essence of the method is described somewhere, in one place? A blog would be convenient.
... non-linearity of time, but ask not read, but the data. Date, time, what was made of, conversions. Then
Try to simulate something similar, you can do it on other tf(approximate), etc. You'll get a brain boost + maybe a good
You'll get a brain boost + maybe a good sign to jack off with. The obtained do not divulge, at a positive cut discount only to Sanka...
With M1 you can work all your life and everything passes. On a uniform timescale, there is pure SB with the corresponding result.
It is necessary to work in a non-linear coordinate system. But, it is not necessarily ticks :)
Result:
Learn, Daddy, while I'm alive.
Alexander, that's not fair, there is a large drawdown and no stops.
That's how we traded on forex, at 1000% a month and 90% drawdown.Not bad!
Where to learn? It's unreal to read your whole thread. The essence of the method is described somewhere, in one place? In the blog would be convenient.
I'll think about the blog :))
But, the essence of the method, in general terms, is as follows:
1. From work with ticks, as well as with M1, M5, ... I gave up working with ticks, as well as with M1, M5, .... And a long time ago. A year ago Doc and I studied the predictive ability of BP in its thinning. Even then Doc said that he was delighted and amazed by the results of the simulation. Then, just like that, he disappeared...
2. I was fascinated by these thinned tick series. I studied them for a long time. It turns out that these series form in the dynamics a certain temporal structure. That is, you can not take just any sample of data. This sample must form, in terms of time, a certain dynamic series: 8 hours in the daytime, 12 hours in the evening, and 24 hours at night. Well, this is simplistic. This series depends on the time intervals between thinned ticks (I call them events). The densification/decrease of tick flow during the day is taken into account.
Then I applied the Einstein-Smoluchowski formula to such a series relative to some average, and voila.
I don't think Doc went the same way as I did, but he and I started this mess together.
Because the distribution of price gains is symmetric, and this symmetric distribution is preserved in the sliding window.
Something like this is what I need to achieve on M15.
The yellow line is a moving average with a window of 100. Almost perfectly flat.
And here is the EURUSD M1 for comparison. 10,000 last bars, no thinning. The average is constantly going way off to the side.
This is one of the last entries I had in my PM from Doc... Something made me cry, remembering the old days....
Because the distribution of price gains is symmetric, and this symmetric distribution is preserved in the sliding window.
Something like this is what I need to achieve on M15.
The yellow line is a moving average with a window of 100. It is almost perfectly flat.
And here is the EURUSD M1 for comparison. 10,000 last bars, no thinning. The average is constantly going way off to the side.
This is one of the last entries I had in my PM from Doc... Something made me cry, remembering the old days....