Despite having read some posted comments like: "Fundamentals are for loosers "[sic], (presumably means getting looser & looser all the time ;-)) personally i think that in the context of trying to write Equity Long-Short algos for Quantopian, i have found more benefit from experimenting with fundamentals data (Morningstar and the new FactSet databases) than i have from playing around with those old, trusted favorite (at least they were 20 years ago) "technical indicators" that we all know & love ... or don't. Actually i doubt that anyone has much hope of winning the daily Q Contest using moving average crossovers or similar .... although maybe i might be wrong. Anyway, in the context of Q algos, personally i do spend most of my time working with fundamentals data.
Nevertheless in a different context, outside of Q, i have been doing some experimentation looking at the Relative Entropy signatures of different technical indicators, mostly but not all well-known ones, and i have been rather surprised at how much they actually differ in their "information content", as measured by relative entropy. Of course this says nothing about the actual RELEVANCE of that info but, all other things being equal, the indicator with the higher Relative Entropy "should" contain more (hopefully useful) info.
Two questions now for friends & colleagues here at Q:
1) Have any of you who like using TA-type indicators used this sort of Relative Entropy approach here at Q with your ranking of indicators for algos? I think this is a perfectly reasonable, sensible, and probably useful question, although i'm a bit less sure about the next one.....
2) Picking up on several posts related to good selection of Fundamental data for input to Q algos, has anyone tried something similar for FUNDAMENTAL input data? Is this a useful / meaningful question? Even if it is not quite useful directly, then perhaps it may lead on to other useful ideas, in a similar way to how a consideration of Relative Entropy with regard to TA Indicators can certainly help us to design better indicators. Any thoughts / comments?