Ward 6 - page 37

[Рустам]

standard volumes = number of ticks ~ Frequency (at least you can think of it that way)

[Леонид]

FAQ: The feedback is the best, or rather the FFO. It is possible to tune it, but whether it will give results (it will lag - definitely). And according to my observation it is necessary to filter several floating carriers at once. All in all there are a lot of questions. And the main one - is it worth it?

It will lag behind definitely and irrevocably)))) That is why the anreal ))))

[Hide]

FAQ:

Price has nothing to do with it. If you base it on price, we're not getting anywhere with this filter. I was basing it on price acceleration.

That's a tricky one.

[[Deleted]]

I see some horror here :-))) I used to make a filter in my time too, focused on the speed of the process.

[Леонид]

FAQ: standard volume = number of ticks ~ Frequency (at least one can think of it that way)

It is hardly possible to recognize ticks as frequency, even if you take e-signal (for example) rather than MT, because each tick does not represent an equal volume of a trade for a given instrument, but simply a trade that has a completely different volume in each new incoming tick.

[[Deleted]]

here's even a good filter I can lay out

[[Deleted]]

And also tell you about its inner workings.

[Леонид]

Dr.Drain: And also tell you about its inner workings.

Tell me, tell me....))

[[Deleted]]

LeoV:

Tell me, tell me....))

Imagine you have a price chart. Let's say EURUSD. We can numerically differentiate it. Obtain the values of the derivatives, or rather the first differences, in units of minimum price change, e.g. 0.0001 or 0.00001 or whatever we want.

Next, I do a clever trick with my ears. What if I draw not just the derivative itself (first differences), but a powerfully non-linearly distorted derivative? For example, I arrange a stepped function, and I take it to a power, or even to the square. That is, where the derivative was 2 units 0.00001 will become 4, and where ten - a hundred.

[[Deleted]]

I can then use this non-linearly distorted array of first differences to draw an "extended price chart". Here is the rule: we take every price value and repeat it as many times as the value of the corresponding bar of our "derivative". That is, if it is 1000, for example, we will repeat the corresponding value of the price 1000 times.

Then we smooth the "extended array" using the usual SMA algorithm with a constant number of bars of averaging. Then we conduct "narrowing" of the smoothed array selecting bars from which we draw the actual filter. As a result, it turns out that smoothing for the sections having different volatility has orders of magnitude different lags.