The WL link published below (WL's builtin Kalman) refer to a very simplified version of Kalman's idea. Its so much watered down, that it should no longer have the name "Kalman" attached to it.
A somewhat better approach is the "alpha-beta" filter. The WL implementation is here:
and all other "Optimal Tracking" variants and "alpha-beta" things in WL's Code Library and Forum...
Basically we have here moving averages with the time constant dependent on the amount of noise (or call it volatility). If noise is low, the "true value" is better known and the average moves faster to the "best estimate". If noise level is high, we have not much clue where a "good value" may be found, so the average moves slowly from where it is, just like a wanderer in the fog...
Yes, I know, this sounds like a great idea to apply to trading, unfortunately...
Our Trading System at C2: Topaz
Saturday January 12 2008 by