On Kaufman And Tillson - here is an extract from a summary I made of an article I once read on T3 (or Tx)
"So, if you multiple run with twicing and turn down the volume factor (Amplitude Response) by an appropriate factor you can get what might be called a
Generalized DEMA (GD)
GD(n,v)=EMA(n)*(I+v)-EMA(EMA(n))*v
Where v ranges between zero and 1. When v=0, GD is just an EMA, and when v=1, GD is DEMA. In between ,GD is a less aggressive version of DEMA. By using a value for v of <1 we cure the multiple DEMA overshoot problem at the cost of accepting some additional phase delay.. Now we can run GD through itself multiple times to define a new, smoother, moving average (T3) that does not overshoot the data.
6 T3(n)=GD(GD(GD(n)))
Results are v similar to DEMA but smoother
In filter terminology T3 is a six-pole non linear Kalman filter. Kalman Filters use the error - in this case (the time series - EMA(n)) â to correct themselves. In TA these filters are known as adaptive MAâs; they track the time series more aggressively when it is making large moves."
The bit that interests me most is the
"Where v ranges between zero and 1. When v=0, GD is just an EMA, and when v=1, GD is DEMA. In between ,GD is a less aggressive version of DEMA. By using a value for v of <1 we cure the multiple DEMA overshoot problem at the cost of accepting some additional phase delay"
On the other hand if you traded on crossovers plotted in a separate window it seems to me that v of > 1 may well be an advantage in identifying the crossover point more clearly at no cost in lag?
