compensation & leverage - how does it work?

On https://www.quantopian.com/fund, it says:

If your algorithm receives an allocation, we will pay you a share of the gross returns that your algorithm earns on our capital. Our target is to pay you 10% of the profit on your algorithm's allocation.

What is the definition of an algorithm's allocation? Is the profit adjusted to an effective gross leverage of 1 (i.e. only the actual capital put up by Quantopian, as implied by the "our capital" wording)? Or does the manager get whatever profit is generated, with him getting to set the gross leverage within the algorithm (up to 3X, presumably)? Or does Quantopian get to set the base gross leverage? What if Quantopian adds leverage, on top of that of the algo (on https://www.quantopian.com/posts/quantopians-first-discretionary-capital-allocations I posted an article suggesting Quantopian would go as high as 6X leverage)?

I guess the basic question is, for a given algo, would the author simply get X% of the profits of all capital running through his algo, regardless of the origin of the capital, cash or borrowed by Quantopian? Or is there a distinction between cash input by Quantopian and borrowed money? The implication here would be whether or not Quantopian plans to share fund profits, since I'm figuring that an individual algo might only support leverage up to 2X-3X, whereas once 30 or so algos are cobbled together, Quantopian might be able to get 6X. So would the additional profit from leveraging up be shared with managers? Or would Quantopian claim that they'd done all the work to get the additional leverage, so sharing the profit wouldn't be justified?

I'm pretty sure I understand your questions, and the way you phrased them in the 2nd paragraph is easier to answer.

The short answer: The algorithm author is paid on the basis of all (any) profits generated by their algorithm less costs attributable to their algorithm.

The long answer:

First, I want to clarify how we consider leverage when choosing algorithms for an allocation. When an algorithm has a starting capital of $1 million, and it buys $3 million in stock, that's a leverage of 3.0 (the calculation gets more complex when the algorithm is both long and short, but you get what what I mean). That 3.0 leverage algorithm is perfectly legal for the contest, but currently we wouldn't give it an allocation in that form. If that algorithm with leverage 3.0 does well in our selection process, currently we will ask the author to modify the algorithm's capital usage to be 1.0, and it is then tested further.

Now, that means that we're working with a portfolio of algorithms that individually don't typically increase the amount of leverage used. We can increase the leverage "on the back end" and give each algorithm starting notional capital allocation.

Now, imagine this hypothetical and simplistic example: We're have $20 million in cash to invest. We lever it to $100 million (5X). We have 25 algorithms. On the first of the year, we allocate each one of them a notional capital of $4million. At the end of the year, imagine that each one of them was up 5% ($200,000). From that $200,000, we would deduct costs including appropriate financing costs charged to obtain the leverage. The algorithm author would be paid their share of the resulting profit. This is a simplistic example and the details are spelled out in the license agreement we execute with the author.