Essentially, I believe this is what Quantopian is doing. They are trying to find alpha generating strategies that are uncorrelated to other strategies in the fund (and uncorrelated to other "known" factors where exposure can be achieved in a cheap manner). Then they are combining them in their fund with the goal of producing a higher overall sharpe ratio.

@ Ben -

Yes that was the idea when Q introduced the fund idea. The same concept applies to a set of factors in an algo. The noise in the factor returns should be uncorrelatd as much as possible so that on average the zigs cancel the zags.

I agree, but they have allocated to relatively small number of strategies, with one single strategy taking a huge proportion of the allocation. This would suggest the idea is to find a small number of exceptional standalone strategies.

My point is that a large number of relatively mediocre strategies (sharpe < 1) can give exceptional returns, and may be easier to put together than looking for the "holy grail".

From the 'Get Funded' page, section 'Our Process':

We are building a portfolio of uncorrelated investments. Financial theory, backed by evidence in practice, shows the benefits of aggregating a large number of uncorrelated returns streams. The key is that the algorithms should be uncorrelated.

This is what Quantopian is doing, and they're not hiding it.

I'm sure this is the intention of Q, "...building a portfolio of uncorrelated investments". They must have an evaluation metric that calculates the correlation between peers of allocated algos. This would ensure that funded algos are uncorrelated to each other.

They also must have another evaluation process that determines how many times each allocated algo will be levered at final execution on the hedge fund level. This was confirmed by Dr. Jess Stauth and this I believe is proprietary information.

That said, the recent $50M allocation to one algo vis-a-vis, I believe , 23 other recepient algos with allocations between $3M-$10M , has put significant weight on this one particular algo, which to me means they have high confidence on this one.

Q has no requirements (“constraints”) for algo authors in this area, so we are flying blind. My sense is that they aren’t too worried about it. The diversification will come from the introduction of other markets and flavors of assets.