Sorry for the double dip, and perhaps another 'rookie' question, but why is there a cap on the negative correlation side? Wouldn't a negative/inversely correlation to peers be preferred even, over a zero correlation, as long as both the algo and peers are both otherwise performing well? Or maybe it's not possible for two consistently negatively correlated portfolios to both consistently return positive results? (I'm suspecting this might be the case but I can't stop my head from spinning.)

Hi Joakim -

Yeah, that's a bit of a mind-bender, in light of the requirement:

Consistent Profitability
We are looking for uncorrelated algorithms that show stable profits. If your algorithm makes money while managing risk exposures and avoiding long drawdown periods, it might be a great addition to our portfolio. We are looking for algorithms that consistently have a Sharpe ratio over 1.0.

Conceptually, the idea is that if Q has a single algo in their portfolio that is consistently profitable, as measured by a SR > 1.0, it will have an average daily return of r_avg. A plot of the daily return stream versus time would be a fuzzy flat line that crosses the y-axis at r_avg. Another algo could be added to the portfolio, with exactly the same SR, and the same r_avg, but the fuzz needs to be different.

Mathematically, we are talking about fuzz terms that look like this:

r_1(t) = r_avg + delta_r_1(t)  
r_2(t) = r_avg + delta_r_2(t)  

where delta_r_1 and delta_r_2 are the fuzz. A plot of one delta against the other should yield a blob, centered on the origin. However, if one gets a line instead of a blob, then there would be no point in adding r_2 to the portfolio, all other things being equal.

If I were Quantopian, I would not think about this in such a strict way, since there may be advantages to admitting a broad diversity of quants into the allocation pool. I suspect that the Low Correlation to Peers requirement may be more relaxed, given that they've allocated to 20 or so quants. My intuition is that it might be really challenging to find that many uncorrelated alpha streams just from U.S. equities, filtered down to the QTradableStocksUS.

The other thing is that there probably isn't that much incremental cost in adding new algos to the portfolio. As I understand, there is just a contract between the quant and Quantopian; there's no overhead of hiring a full-time employee. So, if I were Q, I'd think again about the Low Correlation to Peers requirement, since it would seem better to have 5 correlated algos from 5 different quants from 5 different countries, etc. versus 1 algo from 1 quant, from 1 country, etc. There are unknown risks that potentially could be diversified away.

Also, from a marketing standpoint, if Q is too strict, then they'll only allocate to a handful of quants. Given that they did away with real-money trading, there needs to be some draw to entice the crowd (unless, of course, their main focus is now on the SaaS model).