Adaptive digital filters - page 35
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
A filter looking into the future is not yet possible. So what of it? Why get all hot and bothered? :)
A filter looking into the future is not yet possible. So what of it? Why get all hot and bothered? :)
Like - it takes a little effort :-)
Hi Alexey. The paradox is that to build a "no lag" and no future looking filter you need a priori knowledge about a process to which you are going to screw in this filter (it is how Kalman filter works, for example). We, on the contrary, try to get such knowledge by observing filter's operation... - A vicious, vicious circle with no clearance to the solution of the problem. Moreover, having received, as a result of analysis (or shamanism), this very a priori knowledge, we do not even need MA for decision-making. So if some clever person speaks about building a non lagging and non redrawing filter for market BPs, probably he is deceiving - it cannot be done in principle, because the martingale is (almost) unpredictable.
It is interesting to solve the problem of minimizing the lag for a smoothing filter in the absence of a priori knowledge of the process itself. This problem has been solved absolutely rigorously by Bulashov and the smoothing algorithm is called MEMA. In derivation of the recurrence relation a minimum deviation of the smooth curve from the original BP and its maximum smoothness is required. Well, there is no less lagged and smoother moving average in nature, and it lags be damned! Everything else is bogus. Adaptive filtering will not give any benefit, because the main problem of searching for patterns hidden in BP is not solved (adaptation is not done by this parameter).
Neutron:
Now, there is no less lagged and smoother moving average in nature
Incidentally, no one was willing to comment on a lag-free, non-delayed filter, or the possibility of creating one. Indicative.
Try using Bayesian filtering with complex conditions. Admittedly, it is a hell of a thing, but if you manage to build such a filter for a quote process, you are at least good at it and you can get closer to a good trading strategy.
You can build a Butterworth filter without redrawing, but the result of filtering will actually be not much different from the initial process and will hardly help.
My opinion is simple - it makes no sense to filter the quote process.
to Mathemat
Well, suppose you already have one, and even Juric is smoking on the sidelines. How would you use it?
Oh, that's the easiest one. Once you have such a signal, it means that the distribution of residuals is close to normal and their correlation is close to zero. The trading strategy will be trivial - by extremes and taking into account the power of noise. But it is impossible to detect such a signal that would meet the criteria, even on the history. All we will get is almost the initial quote. The lag in determining the extemma itself, plus the variation of the real quote at the moment of making a deal, plus ... in general, there are difficulties.
"Why do you need one, mikfor? Well, let's say you already have one, and even Jurik is smoking on the sidelines. How would you use it?"
I'm not a fan of mais, because I've observed him on other forums as well.
"Suppose we have a price chart. And a long-period moving average, the SMA. Consider some time interval, say a day. It is obvious that the standard deviation of the mean square deviation of the original price chart is larger than that of the corresponding SMA. In other words, the SMA is "smoother". In other words, if at any time there is a difference between the price chart and the SMA, it is MORE ORDERED that most of it will be eliminated in the future by bringing the PRICE to the SMA rather than the SMA to the price. Simply because the SMA does not know how to run at the same speed as price. A long-period SMA can change quite slowly, by definition, by its very nature.
Therefore, it would seem that if at time t0, if price is, say, 100 pips greater than the corresponding point of some SMA, and we sell with TP=SL=100 pips, we should be in the middle of a large number of similar trades. This is a brilliant and brilliantly simple trading system. It essentially tells us that price tends to the average. But it certainly doesn't work. Because the SMA is LATER. And the value of the SMA in the bar is giving us OLD information.
That's if we had a REAL (and non redrawing) filter, then we could implement this simple and obvious strategy on a grand scale."
i share. for i myself have come to similar thoughts. and i thought about it long time. by the way, i recommend to look there again. it seems that people understood that created an index IS NOT a non lagging non re-drawing filter, but in terms of trade it possesses all its properties.
"all that will come out is pretty much the original quote"
read the page http://forextechnologies.ru/for/viewtopic.php?f=55&t=83 there are pictures demonstrating no lag, as well as comments from smart people. i've been really interested in this topic for a long time, although i'm not welcome on that forum.
" All that will come out is pretty much the original quote.