Nice to see your name in the news eh?
Nice to see your name in the news eh?
I liked the headline, "College Kids Are Now High Frequency Trading From Dorm Rooms." Nice to see they did their research there. Nevertheless it was pretty cool to be the lead.
Quantopian’s founder and chief executive John Fawcett said its membership has surged to 60,000 from 35,000 less than a year ago...
Yet, when I look at Contest 13 (https://www.quantopian.com/leaderboard/13), there are only 288 submissions. Assuming 3 per contestant, that's only 96 contestants. The participation rate is 0.2% out of 60,000. This doesn't really make sense, given that Q will simply hand over $5,000 to the winner, each month. I wonder what's going on?
I recently reanalyzed the Quantopian Open contest's specification and rules. In addition to all of the criticisms that it deserves (or not), one notion struck me as obvious and most likely foolish. This has been mentioned in other posts, often perhaps, but...
Additionally, the repeating nature of winners being allowed back into the contest after a 3 month suspension, allows them to continuously tweak their algos, in minute increments, and then be re-entered. It is therefore possible that a cadre of 4 algo writers could theoretically own the contest. The likelihood of this occurring is slight, I'll admit. But regardless, to me, the restriction on algorithm diversity and the repetitive reintroduction of past, but tweaked, algos would be a great way to build a dead-end fund.
The other thought that occurred to me is how does Q monetize 60,000 users, aside from the Q fund idea. Let's say each puts in $25,000 of capital (not likely), and is willing to pay 1% per year (or 1% is gleaned through paying for order flow or whatever). You only end up with $15M per year in revenue, from which expenses have to be subtracted. With 30+ employees and investors to pay back, it seems like a tough business to be in. Maybe this is why they shifted to the fund in the first place? It was never really explained, vis-a-vis what seemed to be the original business plan.
Or, maybe Q's exit strategy has been (perhaps all along) to build an algo generation engine and sell it to the likes of Citadel or Blackrock. The top talent at the big hedge funds are major overhead. If you could prove that you could offer something comparable, with low, controlled costs (bright, poor, college grads and students) perhaps you could lure a major player to buy you outright for $100M.
And, as I am wont to do, I pontificated, this time upon Q's quandary. I considered an easy way for Quantopian to expand their restrictive algo selection method; offer multiple, specification altered contests.
Right now the contest's rules have been distilled down to a narrow, conservative, might I say boring rule set that any state pension fund might be able to replicate with little effort. So, twist the rules for a separate contest. And twist them again for yet another.
What if you couched a new set of rules thusly: beta 0.5+, Sharpe 1.0+, volatility < X, security count 20+, 10 year back test. The 6 month paper trade period would continue to be applied, but now you've got a whole other set of strategies that can be tested. And then twist the rules for yet another strategy variation. Different rule sets = different contests = different risk profiles = different strategies = different income options. The Q would get expanded participation of their so-called 60k members, and eventually, when futures come on line, different strategy types will be easily incorporated into the Contest.
And, one last thing, and this has been posted ad nauseam, formalize the scoring, per rule set, such that a strategy can be graded outside the ranking system.