Hi, what is your practical experience with Kalman Filter? Is it worth to learn for the domain of finance, markets, trading system development?
I don't think so. I've played around with many types of filters over the years...Kalman, Butterworth, low-pass, and various IIR and FIR filters. My opinion is that none of them have an advantage. You're better off sticking with indicators that other people follow such as the 50, 100, and 200 SMA. Indicators are not magic. Although they are easy to curve-fit, the only predictive power that they have comes from other people using them and thinking the same thing. Do you really think that there's anything magic about Fibonacci numbers? They only work because enough people follow them and expect them to work. It's a self-fulfilling prophecy.
TA = "self-fulfilling prophecy" Well said, I too had the same thoughts about it. It works (mostly) just because the TA crowd follows it... Ok, thx, this helps me to decide between Kalman Filter and Time Series Analysis as the next thing I wanted to study. I'll stick with the second.
It depends on what you use it for. Within the context of simply filtering a timeseries it is not very useful. For parameter or regime learning it is a different story. I'd say gradient boosting, for instances, with a specific target function is kind of like a Kalman Filter.