From theory to practice. Part 2 - page 134
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
There is an obvious problem - extrema of what? To buy, we have to rely on the lows of the Ask, and to sell, on the highs of the Bid.
So the obvious solution is to build tops using bids and troughs using Askas
I understand you perfectly, I was sick of it myself at first. Python shouldn't even be compared with Java and Sharp, Python is a scripting language, that's its main niche, while in 10 minutes to try out an idea, you can only use Python, in Java/Sharp you will spend 2-3 times longer, in pluses 5 times, you will have time to forget the idea, get stuck in the technical details of implementation.
Yes you're right, but I'm horrified by absence of control over many important things in Python, without which I can't trade at all, especially lack of strict typing...
Well, it's OK, as they say to each his own)
Yes, technical details are the bane of "plus-like" languages...
That's right. Many traders spend 10 years or 20 years optimising their TS parameters. In fact, they try to adjust to the current non-stationary intensity of tick flow, without making any effort to minimize its impact without losing the basic properties of market BP.
Optimizing the TS parameters via backtesting - doesn't work and never has worked, well maybe just a little in the pre-computer era. Do not believe the anonymous suckers like me, believe Prado, he kind of managed the money publicly and even had a profit. Forget optimisation in the backtester. And the non-stationarity of the price series isn't going anywhere, no matter how you filter it, the other thing is that non-stationarity itself isn't such an obstacle if you have price drivers that are minimally ahead of it, as they are directly or indirectly causing it to change. And the tick flow in kitchen forex is synthetic, there is no point in analysing it.
Data preprocessing is the key thing. Including in MO, if I'm not mistaken.
You are absolutely right, but we need quality data, which contains something valuable that has not yet been taken over by everyone who is not lazy. And most importantly being able to understand when there is nothing in the data.
Yes you're right, but I'm horrified by absence of control over many important things in Python, without which you cannot trade at all, especially lack of strict typing...
You don't need to trade in python. Only sketches of data transformation processes, statistics, MO, econometrics, individual strategy fragments. Python is similar to GUI in some sense, I just ran through the menus, found what I need, ran the test, without distracting at all, in fact, it's not even python itself that matters, but frameworks for it in our subject area. Python seems childish and limited at first, but later you get used to it (after about half a year) and realize that it's really more convenient when it comes to working inside the subject area with user-friendly frameworks. Without Python it is simply impossible to quickly check all the flow of ideas arising in the minds of quants, if you do it in pluses or sharpe\yawa, it is like flying in a helicopter to the shop for bread across the street.
I see that the mold has also wished to speak out on the theory and practice. It pains them, you know, that no one pays any attention to them. But they have nothing to say on the matter. So they have chosen the tactic of yapping from the back alley.
You went out and asked to be removed...
Then you insistently asked the moderators why you weren't deleted...
Why come back? Not enough attention?
And as someone used to say, "Who calls himself what he calls himself."
Quantum, if you do it in plusses or sharpe\yawa, it's like flying a helicopter across the road to the shop for bread.
So the obvious solution is to build tops by bids and troughs by ascas
With this approach, at small zigzag parameter its "zigzag" - strict alternation of tops and bottoms - may sometimes disappear.
There is no need to trade in Python. Only sketches of data transformation processes, statistics, IO, econometrics, individual strategy fragments. Python is similar to GUI in a sense, you quickly run through the menus, find what you need, run and check it, without distracting at all, in fact, it is not even Python itself that matters, but the frameworks for it in our domain. Python seems childish and limited at first, but later you get used to it (in about half a year) and realize that it's really more convenient when it comes to working inside the subject area with user-friendly frameworks. Without python it's simply impossible to quickly check all the flow of ideas arising in the heads of quants, if you do it on pluses or sharpe\yava, it's like flying in a helicopter across the street to the shop for bread.
In terms of matstat, python is still pretty rubbish compared to R. There's nothing there) But the future is clearly ahead of it.
So the obvious solution is to build tops by bids and troughs by Ascans.
Even if we look for extrema by High/Low candlesticks, they won't alternate. Of course we may look for the highest High and the lowest Low from Highs, but it's not that simple. The algorithm will be quite complicated.
I also thought that the obvious one, but then I gave up, it is more correct / much cheaper to search for highs and lows by one parameter.
With this approach, if the zigzag parameter is small, its "zigzag" - the strict alternation of peaks and troughs - may sometimes disappear.)
Why? I think it depends on the algorithm of the zigzag, if the "zigzag" is necessary (in my opinion, yes), it will remain.
The point is that such a construction has a physical sense (estimation of maximal profit on a period, which can be taken on the market) other constructions of zigzags have less evident sense.
It seems that just such a zigzag is used by fxsaber