From theory to practice - page 25
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
It's not hard to make a Winner's NEMARCKOW moving average. Partly you do it yourself. Once you do MA the Wiener process no longer exists, even if it was one originally).
!!!!!!!!!! Yes.
But about that - probably tomorrow.
I just don't have the time right now.
While I'm away from the forum, can someone justify what Yuri said?
Can someone describe the physical meaning of choosing MA or EMA or WMA?
Why do you choose this particular type of moving average (of those traders who use them, of course)? The answer - "because there are more profits and fewer losses" is not accepted :)))))
By the way, we will discuss again tomorrow - whether or not the process of price returns is stationary at the chosen sampling time T for tick quotes, as opposed to the non-stationary process of Ask, Bid or Last price movements itself.
I feel anxious in advance - but we both know the answer to this question, don't we? :)))))))
And yet there has been no justification as to why ticks?
What's worse than opening points for minutes? There are 2 million data points.
Maybe it's better to debug the model on minutes, and then get into ticks.
The minutes are ready, while with ticks, you may encounter problems with getting the long history and replenishing the history.
They need compression algorithms, infrastructure in the form of VPS, connect to VPS for automatic downloading ticks for yourself, etc.
You should not hope that you can put in a terminal ticks collector and they will be collected, the history will be leaky, I tested it on my own experience.
And yet there has been no justification as to why ticks?
What's worse than the opening points of a minute?
Maybe it's better to debug the model on minutes, and then get into ticks.
The minutes are ready, while with ticks, you may encounter problems with getting the long history and replenishing the history.
They need compression algorithms, infrastructure in the form of VPS, connect to VPS for automatic downloading ticks for yourself, etc.
I've tried it myself.
Here you are, Nikolay - respect once again. I'm amazed at how accurately you understand the essence of the problem.
Exactly!
In order to take the integral component into account - you need historical data... Every brokerage house has different data... Some don't have them at all. Do we have to do it ourselves, like I do now? And when developing a trading robot, what should I say to my client - well, you may have to collect some data for a month?
I am thinking about it.
You don't understand the theory, but you also have technical problems. Alas...
The Bollinger Bands do not take into account the non-markness of the process. We must consider "memory". And memory is an average of process values based on historical data.
Wait a minute. Bollinger is based on the SMA, each current point is a function of N previous price values. Where is "no process memory" here?
Further, the deviation is a function of N previous SMA values, and thus contains information for even 2*N previous price values.
Could you make your point more clearly?
Here you go, Nikolai - kudos once again. It never ceases to amaze me how accurately you get to the heart of the issues.
Exactly!
In order to take the integral component into account - you need archival historical data... Every brokerage house has different data... Some don't have them at all. Do we have to do it ourselves, like I do now? And when developing a trading robot, what should I say to my client - well, you may have to collect some data for a month?
I am thinking about it.
You don't understand the theory, but you also have technical problems. Alas...
If we calculate by opening prices and a certain minute is absent, we fill it with closing price of the previous bar. I will write the algorithm on my knees.
The only thing that leaves gaps is weekends and holidays, but the model should take into account deferred demand. I would say something was happening in the world, but I didn't see any trades, and after days off everything goes like crazy. But I have no idea how to deal with it.
The solution lies in the same plane as activity changes at the opening of trading sessions.
To calculate the integral component - we need archive history data... They are different for every brokerage company...
I don't quite understand why you're fighting over ticks either. With your scale of transactions (and they last for hours and have a value of tens of points), the differences in single ticks do not play any role at all.
The brokerage companies' data is almost the same at the minute level, at the tick level the difference is units of points and units of seconds.
Wait a minute. Bollinger is based on the SMA, each current point is a function of N previous price values. Where is "no process memory" here?
Further, the deviation is a function of N previous SMA values, and thus contains information for even 2*N previous price values.
Could you please state your point more clearly?
I don't quite understand why you're fighting over ticks either. With the scale of your trades (and they last for hours and have a value of tens of points), the differences in single ticks do not play any role at all.
At the level of minutes the brokerage companies have almost the same data, at the level of ticks the difference is units of points and units of seconds.
Yes, yes, I'll have to think about it. Maybe really take CLOSE for the minute bar and that's it?
Open, they make it easier to write algorithms and fill in holes.
For the calculation of klose there is one moment while it is stable in realtime, the open is set once and for all.
After the open appears, a new calculation is made and a trading decision is made, while the klose in the meantime is constantly bouncing until the close.