For all the quant traders out there, how do you combine signals? What are the pitfalls, hard learnt lessons? @therobotjames @macrocephalopod @HangukQuant @macro_srsv @robertmartin88
-
4/4 Also I think it's worth checking out the literature on aggregrating probability estimates. In superforecasting tetlock mentions that you shouldn't just take the average – you should subsequently extremise this average else you lose information. e.g Satopaa and Ungar 2015)
Basically agree with all the above. Point 4/4 needs care. If you are forecasting a real-valued qty like a return, then extremisation happens naturally in any reasonable regression framework. You don’t need to extremist further after getting your regression coefs.
