@sle, could you share some insight on what is "threshold chaining"? Also, I wouldn't take what kut2k2 said seriously.
I think the key is to construct a stationary time series via transformations or linear combinations. I believe once you have stationarity, then you can make inferences on population parameters, which in turn lets you formulate probabilities. So it's not about being scared or wimpy, but simply rational.
Simple noise analysis and simple autoregressive modelling. Sounds like stuff they teach in CS-oriented physics programs. I do optimal noise reduction and multitemporal trend extraction. Aka stuff they don't teach in CS-oriented physics programs.
If what quants do is just profitable, why aren't they all rich?? I was referring to the fact that directional trading is just plain more profitable than market-neutral strategies when done correctly. So if the quant way is the right way, why are they scared of directional trading?
The problem is that price series are so emphatically nonstationary, trying to forcefit them into an ARIMA model is akin to forcing a square peg into a round hole and declaring it a good fit. Tsay's book deludes readers into thinking they can get some useful information from ARIMA and other ill-fitting models of financial series. Maybe econometricians can, but it's a waste of time for traders.
Exactly, price series are emphatically nonstationary. That's why one needs to transform or take linear combinations or trade volatility or some other novel tradable derivative combination that is stationary.