Funny enough, you can use Alphalens turnover analysis to discover that. As the turnover is computed as the proportion of names in a factor quantile that were not in that quantile in the previous period, we can simply create a single quantile that contains the full QTradableStocksUS universe and then run turnover analysis on 1 day period (disregard factor autocorrelation plots).

Thanks Luca! Also, the original post announcing the QTradableUniverse in experimental mode has a similar analysis. You could re-run the notebook originally provided: https://www.quantopian.com/posts/working-on-our-best-universe-yet-qtradablestocksus

So what’s the answer? How much daily turnover is there in the QTU, roughly?

It seems ~0.1% or ~2 securities per day.

Thanks Luca & Josh -

From Luca's analysis above and the attached notebook, it appears that a handful of stocks per day are swapped out (out of ~ 2000). Seems pretty high. I wonder how this compares to a common market index (e.g. Russell 2000)? And with all of the other required risk controls (constraints), is there a net benefit to the turnover? Or might it be a drag on returns?

I guess the follow-on question is which stocks are getting swapped out? Is it random, or do they have predictable characteristics?

I'm also wondering if certain stocks are jumping on/off the QTU, and if so, on what time scale? I'd think that this sort of filtering could lead to oscillatory portfolio churn, without any real benefit.

Generally (and I'll keep asking), when (if ever) will the QTU code be published (along with the Optimization API and risk model)? Possible answers:

a) Never
b) Eventually
c) Don't know, but I'll get back to you
d) (no response)

There are a variety of "use cases" to provide users with access to the code (aside from it being in keeping with the presumed Quantopian operating principle of transparency wherever possible).