|
PureBytes Links
Trading Reference Links
|
Gary wrote:
>There's also Tillson's T3 average, which gave JMA a run for its money
>until Mark got annoyed and found an improvement to keep JMA ahead. :-)
T3 is a fancy cascade of exponential moving averages with feedback
taps. It's an IIR filter, sharing no similarity with statistical
filters. Its main benefit is that its very efficient to calculate
and performs reasonably well. It has less lag for the same level of
smoothing than more primitive filters like the simple or exponential
moving averages.
However, I don't think T3 ever gave the original JMA a run for its
money. I saw the performance of the original JMA. T3 had more lag
and overshoot. T3 also can't adapt to sudden changes (like when
there's a limit move) without requiring significant settling time.
>Bob Fulks wrote a TS version of T3:
>http://www.purebytes.com/archives/omega/1997/msg00119.html
>http://www.purebytes.com/archives/omega/2004/msg03606.html
The suggested damping coefficient of 0.7 is too high; it has an
energy gain. See http://unicorn.us.com/trading/src/T3Avg7.gif for a
frequency response spectrum. The damping coefficient should be more
around 0.5.
-Alex
|
