Back to Community
What about a new contest?

I was wondering if there was another contest that could be done that doesn't require shorting (as I'm working to prove that you can long only and make great returns). I have a long-only strategy that matches all the requirements besides shorting but it does hedge into gold and bond ETF's. I was wondering if you would consider a long-only contest (same criteria as the current ones because it's possible as I've done it) and see what you think of the algorithms that come in. :) My long only algorithm has a beta of 0.23 and volatility of 0.15. Let me know! For long-only, I would just change shorting to gold and bond ETF's since that's all you can really do. :)

3 responses

Hi Dustin -

A couple thoughts -

  1. You could try adding an itsy-bitsy short, since I don't think the contest requires a specific implementation of longs and shorts. One approach, perhaps, would be to use a so-called "hedging instrument" to bring beta down closer to zero (e.g. SPY).
  2. In theory, if you've isolated long alpha, you may have also isolated short alpha and just not realize it. So, maybe your algo could be re-jiggered to be the classic long-short that Quantopian needs.
  3. The contest guidelines, as I interpret them, are meant to match up with the requirements for getting a Q fund allocation (https://www.quantopian.com/allocation) and what is described on https://blog.quantopian.com/a-professional-quant-equity-workflow/. My sense is that the Q fund team is looking for something kinda-sorta specific, that can be leveraged up to 6X to trade tens of millions of dollars per algo, that will be attractive to institutional investors. As I understand, the all-equity alpha needs to be isolated, since that's what the market is willing to pay a premium for. Sloshing tens of millions of dollars in and out of ETFs probably isn't what they want, but I could be wrong.
  4. Personally, I've been trying to focus on what the Q fund team sees as fund-able, versus trying to shoe-horn something into the contest, or game it, since I don't expect that they will set up alternate contests (e.g. ones that might accept low-capital, long-only algos, of the type that retail investors would trade).

Just my two cents...some of my interpretations may be off-base, so feedback from the Q team is certainly welcome.

I only put money into gold and bond ETF's for hedging purposes since if markets go down, people move into gold and bonds as a usual reaction. But as for ETF's, I don't trade ETF's if they don't pass my screen in the universe (not sure if the Q1500 trades ETF's or not). But reading the requirements, it says it must hedge by short and longs and I would like to rejigger it but I don't necessarily want to mess with it as once I can get the money to be able to run it myself, I would be using. I have isolated long-only alphas with the help of alphalens and some gold/bond etf buying for hedging. My beta is 0.23 over all. Completely positive except in the 2008 crash (only a -10.3% return as the max loss during the 2008 crash) and has been nothing but positive returns after that, including in different time periods. But from what I read, it seems as though they want long-short algorithms in contests only because it's for getting into Q fund allocation. But about the long-only, low-capital algos that retail investors would trade, my algorithm actually does better the more initial starting capital you have so that more you have, the better you chances of the drawdown being reduced in my algorithm. Although my algorithm takes only a minimum of $10k to run, it does significantly better the more that is put in it, returns might be a little less (as expected because it takes more money to get to the same amount of return percentages) but drawdown ends up being reduced. Although a retail investor could trade on my algorithm, so would hedge funds with millions of dollars :)

Hi Dustin -

Not sure what to say, but if you've really got something, then it shouldn't take much to hedge it with shorts to get it down to beta ~ 0. What does it look like if you run a backtest starting in 2002, up to the present? If you think you have something that Quantopian could pour $10M into, then I'd contact them, and maybe they can advise.