|
PureBytes Links
Trading Reference Links
|
It seems to me that the most straightforward way of dealing with the
displaced average is to use the last value of the average for the
current bar. This is not perfect, but is a workable approximation. The
normal way to "extend" a line would be to calculate the slope. This has
lots of pitfalls with price graphs and length-limited indicators. THere
may have been a very recent change in slope (neg to pos or vice_versa)
which is shorter than your length parameter. Then your first point may
be before the change, and the slope would make little sense,
representing what happened much less recently. Or the change may be
very sudden, and, if within the length parameter, may give a very sharp
slope which will put the price target in the stratosphere (or the
basement).
Or average the last value of the Displaced MA with the more recent bars,
essentially creating an Exponential MA. But then you don't have a true
displaced MA :-)
donc
>Subject: Re: Unidentified subject!
> Date: Fri, 8 Oct 1999 05:52:08 -0400 (EDT)
> From: editorial@xxxxxxxxxxxxx
> To: "Anshien, Joe" <Joe.Anshien@xxxxxxxxx>
> CC: Omega List <omega-list@xxxxxxxxxx>
>
>Aha! You've discovered the problem with "displaced"
>averages.
>
>In general, a displaced average is moved back in time.
>The most common use of this is the "centered" MA, in
>which the MA is moved back half its length.
>
>By definition then, such an MA has no value on the
>current bar and cannot (by itself) be used to trigger
>a signal on the current bar.
>
>Is there any way around this? Those who recommend
>displaced MA's often recommend "projecting the MA
>forward" to the current bar by "extending it along
>its current path". Unfortunately, there is no way
>to program such an ill-defined extension. In fact,
>you'll find many references to "extending lines by
>eye" or "using a pencil to extend the line..." (No
>kidding... the methods are totally subjective and
>impossible to objectively backtest.)
>
>What you'll find, I think, is that, almost regardless
>of the mathematical complexity of the method you
>might choose to extend a displaced MA forward to the
>present, the extension will be wrong when the price
>turns. And, the "bigger" the turn, the more wrong the
>forecast will be. (This holds true for even a variety
>of non-linear forecasting methods.)
>
>
> ---- you wrote:
>> Please tell me how to create an exit signal when a close below a displaced
>> moving average.
>>
>> Thank You
>> Joe.anshien@xxxxxxxxx
>>
|