Dr.Fx corner - 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
Thank you very much, Prival. I am, however, digging in different directions. And my attempts to catch fractions of a pip are not alien to me, although it is not HFT, but it is not scalping as well, i.e. all the same ideas about filters are based on. It's important to hear exactly from you that there are fish :-)
I will try to help you some more, to give you some hints. What I think is important. It's not about pips, many people mistakenly think that if you analyse ticks, it's necessarily about pips. This is not the case with the theory of optimal control. In our case the control object is our account. All the formulas of optimal control obtained by Lyapunov, Pontragin and others are received in continuous time (integrals). But when you start to implement them in practice, you face difficulties, the most important one being that observation and control is discrete.
You take 5 minutes, i.e. during this time you can't control the object (I don't take TA and SL as a control signal, they are rather boundary conditions, limitations, like an emergency stop tap :-)). So it turns out that your object moves uncontrollably for 5 minutes (if we made such an autopilot, they would put us up against the wall :-), and here, like Forex, everything is possible, there are geezers who work on daily, hourly candlesticks ..... It's like driving a car on a mountain road and touching the wheel once an hour (disaster is inevitable) ...
That's why you should always monitor the market, get the data as often as possible, and not only for one currency pair (the market is not just the Euro / dollar rate, it is much wider). Obtaining this data, analyze and decide whether to continue in this direction or turn around.
I've already given a link to one of my best trades, I held a position for half a year, took 16 figures and all the time I was investing in the direction of movement.
So it's not about pips, it's about the quality of management and the quality of the data you get.
One more argument to move to smaller timeframes. There is Kotelnikov's theorem and it defines the sampling frequency, at 5 minutes all oscillations with period less than 10 minutes are simply unavailable for analysis...
One more thing, if you watch the whole market, build multicurrency analysis, some kind of synthetics, then from my point of view, the algorithm's entry should be given not as Close. The closest to the input price is a point in space equal to (ask+bid)/2 in my opinion.
But here you have to wait for new deutafide, because now the asc is the average temperature in the hospital...
Colleagues, closing all trades.
EURUSD has gone down as predicted,
GBPUSD does not want to go up, but it did not go down either,
EURJPY went down as predicted,
USDCAD went down as predicted, then went up, but not higher than it was. There is generally no loss either, even if one closes now rather than early.
P.S. Privat, thank you. Most of what you say in the post above is just common sense, already obvious to everyone. I agree that you need to analyse as complete a baseline as possible. I don't have the M5 not out of a desire to ignore the minutes and below, but simply out of a desire for speed of computation.As for the comparison between hourly and daily candlesticks Iagree, but I don't really think it's important.
...
P.S. Privat, thank you. Most of what you say in the post above is just common sense, already obvious to all. I agree that you need to analyse as complete a baseline as possible. I don't have the M5 not because I want to ignore the minutes and below, but simply out of a desire for speed of computation.As for the comparison between hourly and daily candlesticks Iagree, but I do not really think it is important.
That it's common sense comes out of your area of expertise. For you it's the obvious things + I try to write in simple language (without complicated math). Want I can provide you with a couple of respected members of this forum and try to convince them of that.
Here's a link for example, it took me many years to do it http://forum.mql4.com/ru/5190/page2#19194
I can also give you a couple of others who like diaries (there he found, analyzes weeks:-)), try to convince them that he is wrong. They will not understand what we are writing about here, because for us it is obvious, for them it is not, because it is beyond their knowledge and understanding.
Regarding (ask + bid)/2.
You are involved in metrology by profession, so the theory of estimating an unknown quantity is close and familiar to you. It could be voltage (current, Price, etc.). Nobody knows the truth, we can only get an estimate of this unknown quantity and taking into account the uncertainty of the device, we can only say that the true price lies in a certain confidence interval.
Let us transfer this knowledge to the market. What is the meter? What is its accuracy?
I assume that bid and ask are the limits of this interval (may the instrument lies and its error value is unknown to me), but there is nothing better, the true price nobody knows, we have only the estimate of this unknown value by other bidders.
Now let's analyze the situation, there are volumes at bid and ask, no deals yet (no one knows at what price the next deal will be, at bid or ask). Therefore it is better to take for analysis a point equal to the middle of this interval (it is possible to take the entire interval, the spread), now the next step is no deals, but ask and/or bid start to take volumes, i.e. there are market participants that change their estimation of this price (and I want to note, these participants are not afraid to stay at the ends of the spread!If the spread spread widened symmetrically, the price estimation did not change, it still lies at (ask + bid)/2 and if it is not symmetrical, the ask just widened, then this estimation has shifted ...
I realise it looks like common sense, but I am proceeding from my knowledge and understanding of pricing in the market, the physics of what happens in the cup. It helped me, and it can only confuse you, do it the way that is convenient and understandable for you. And as an option, when you build TF, leave this one, just give this value at the input and see, maybe the filter will work better.
S.U. And concerning Lyapunov and Pontryagin's maximum principle, they can't be realized in practice (with few exceptions). The most viable is Letov-Kalman criterion here http://drive.ispu.ru/elib/kolganov2/l7.html
Colleagues, closing all trades.
EURUSD has gone down as predicted,
GBPUSD does not want to go up, but it did not go down either,
EURJPY went down as predicted,
USDCAD went down as predicted, then went up, but not higher than it was. There is generally no loss either, even if one closes now rather than early.
P.S. Privat, thank you. Most of what you say in the post above is just common sense, already obvious to everyone. I agree that you need to analyse as complete a baseline as possible. I don't have the M5 not because I want to ignore the minutes and below, but simply out of a desire for speed of computation.As for the comparison between hourly and daily candlesticks Iagree, but I don't really think it's important.
You came out of the market and soon there was good movement on all the pairs. Why? Don't trust your filter?
I seriously wanted to finish the branch. Heard Privat's approval, that's enough. Besides, off to work. Intense day. That's me behind the monitor yesterday. What if people lose money from overconfidence in my predictions, without control of the market on my part?
Privat, you are very good indeed. I read many of your threads on Forum 4, I get useful information for myself. But you mentioned that I deal with metrology. You're not. I had a period of time between science and industry. But now I'm in industry.
The filter I'm going to show today is more of a lag-free filter. That is, what I call non-lagging (there are skeletons in the wardrobe, if you start strictly mathematically figuring out what I mean) is achieved with shorter time transients. In short, the public will find it more bizarre. However, questions on lousiness like "why do you have a big fat maximum on the right of a big fat maximum in the price" will cause less. At the same time, since I have deciphered what I understand by the OEM (volatility ratio), I will sign this characteristic of the filter on the interval shown (as before 2*288 bars M5).
Since we liked USDCAD so much, let it be: a series of 5 filter realizations with different smoothing at the top, only one curve below, Z5 = F, and the OEM value.
The filter is not so smooth, so it is more "by eye" to choose a direction to enter the market, smoothing the filter by your own brain. So, selling.
EURUSD is also selling. Generally speaking, when the filter does not have the appearance of a smooth line with a clear derivative sign, different thoughts begin to emerge as to how to trade on it. It is more difficult to "follow it". Besides, an idea emerges that, since the price volatility is larger, any mismatch is more likely to be eliminated by the price movement than by the filter. That is, you should trade on a price move towards the filter. In reality it is more complicated than that. Not only such characteristic as ratio of volatilities is important. But also the nature of these volatilities. Imagine there a curve similar to this one but with a meander of amplitude five points superimposed on it. The resulting volatility of the "filter" will be higher than that of the original curve, while its (the filter curve) will not actually change. The nature of both price and filter volatility is non-stationary. Thus, if the filter is not distinctly "smooth", the question of how to trade using it is not as easy as it may seem. I will mentally (or rather by ksmooth operator in 8 hours) smooth out the filter curve and trade in its direction.
I will buy GBPUSD though.