Exponential Moving Average - Lag

Quote from jimns:

Advanced

Don't know if you've read Ehlers paper on Signal Analysis Concepts, but you can find it here. In his paper, Ehlers explains that a simple moving average (SMA) reduces a number of periods to a single point that lags one half the width of the observation window. He then goes on to discuss the lag in an exponential moving average (EMA) and, from his formulas one can derive the following formula for EMA lag: Lag = ( L - 1 ) / 2 where L = length of the EMA.

On the other hand, in Trading for a Living Section 25, Elder says "The proper way to plot a simple moving average is to lag it behind prices by half its length. For example, a 10-day simple MA properly belongs in the middle of a 10-day period, and it should be plotted underneath the 5th or 6th day. An exponential moving average is more heavily weighted toward the latest data, and a 10-day EMA should be plotted underneath the 7th or 8th day. Most software packages allow you to lag a moving average."

So what am I trying to accomplish?

I'd like to understand whether anyone lags their EMA as Elder suggests, is there really any value to doing it that way, etc.? And I'd like to understand how the EMA lag is supposed to be calculated since Ehlers and Elder seem to disagree about that.
More...
Ehlers got it right. Another way of calculating lag is (1-A)/A, where A is the smoothing factor of the EMA. That frees your EMA from the restriction that it must correspond to an integral lookback period (A = 2/(N + 1), where N is the number of bars in the lookback).

A lagging indicator need not be the limitation that most traders think it is, but that calls for thinking outside of the box, which most traders are not capable of. Good luck.
 
Here's a very advanced concept: you could perform some experiments yourself, measure your results, and formulate your own conclusions.

You could make yourself a 3-column spreadsheet in which column A is BarNumber (or "date" if you prefer), column B is Price, and column C is exponential moving average of price. Then you could plot the data and directly measure the lag right on the plot itself. Lag is just the horizontal distance on the plot, between Price (plot of column B) and exponential moving average of Price (plot of column C). What could be simpler?

A bunch of old farts, long long ago in a place far far away, once decided that the simplest way to do this is to use a "ramp" for Price, column B. (They did this long before personal computers or spreadsheet software were available, so they didn't call it "column B" of course.) The price on barnumber 1 is 1.00. The price on barnumber 2 is 2.00. The price on barnumber 3 is 3.00 The price on barnumber 4 is 4.00 and so forth. When plotted, this price data is a straight line from southwest to northeast: a ramp.

Now if you are right handy with math, if you remember that high school senior math class that came after geometry and trig, you'll be able to write equations that exactly describe column B and column C. You won't even need to use a spreadsheet, you can measure the lag with equations!! But if you are skeptical about your math abilities, or if you just want a double-check, an empirical confirmation, then you'll want to measure the lag on the spreadsheet plot too. Belt and suspenders.
 
kut2 no offence taken. Not sure what you are talking about as there are many different approaches to cycle analysis.

If you can explain what you mean I can tell how it differs. For eg do you mean a linear 21-day time cycle where a top or bottom forms around the 21-day cycle date?
 
Quote from yoohoo:

kut2 no offence taken. Not sure what you are talking about as there are many different approaches to cycle analysis.

If you can explain what you mean I can tell how it differs. For eg do you mean a linear 21-day time cycle where a top or bottom forms around the 21-day cycle date?
More...
I'm just going by what Millard posted:

"Three cycles are shown for the FTSE100 Index in early August 2007. These are the short term 15 day cycle, a slightly longer 41-day cycle and a long term 201-day cycle."

There are lots of "magic" numbers in TA, not even counting all the fibonacci stuff. J. Welles Wilder was especially fond of the number 14, allegedly because it is half of the lunar cycle, but then you have to completely ignore the fact that the lunar cycle is 28 calendar days, not 28 trading days.

I've learned to be very leery of "magic" numbers until the proponent bothers to explain exactly what is so "magical" about them.
 
kut2 - no presets - no magic numbers - way off the mark here - just a tool that presents endless opportunity for creativity and big rewards for hard work.

I understand where you are coming from, but nothing esoteric here.
 
Quote from yoohoo:

kut2 - no presets - no magic numbers - way off the mark here - just a tool that presents endless opportunity for creativity and big rewards for hard work.

I understand where you are coming from, but nothing esoteric here.
More...
Every TA indicator is alleged to be "a tool that presents endless opportunity for creativity and big rewards for hard work."

And if you don't think there are any magic numbers, then explain why 201 is all over that website. What's so special about 201? This isn't a trivial or impertinent question, vendors have an obligation to explain why what they selling is worth it.

For an example of a TA vendor who doesn't prey on the naivety of his customers, go here:

www.jurikres.com

Jurik doesn't disclose any algorithms but at least he explains why his indicators live up to the claims he makes for them.
 
not sure if you're asking me, telling me, or wanting me to tell you MUST get this software.

In any event, all I got to say is, if it's not for you before you try it, fine. I found it to be amazing, but then I used it.

Enjoy it... or enjoy not using it - whatever.
 
You must log in or register to reply here.
Back
Top