I recently read a couple papers describing 3 or 4 factor models that explain a huge number of anomalies (purported sources of alpha), such as this one by Lu Zhang et al:
http://rfs.oxfordjournals.org/content/28/3/650.short?rss=1&ssource=mfr
Usually the models include a factor for the market, a factor (or two) for value/mispricing and maybe a factor for momentum. These models are basically claiming that using an anomaly that correlates closely with the value factor, for example, is getting extra returns because it is exposing you to more risk, not because it is a source of alpha. Value and momentum are things that don't work all the time (e.g. the period from 1996 to 1999 was horrible for value), so by definition, it is a risk to expose yourself to it even if, over the long run, it will bring you excess return (this is the risk compensation).
On Quantopian, the results for beta and alpha are computed using a simple CAPM model, where the only factor included is the market factor. A user just has to expose themselves to the value or momentum factor over the long run and it can look like "alpha" on Quantopian.
I'm wondering if the Quantopian team has thought about this and if it would make sense to somehow incorporate risk factor models like these into the analysis to see if algorithms are actually finding true alpha or just exposing themselves to extra risk.