1. Welcome to the new Styleforum!

    We hope you’re as excited as we are to hang out in the new place. There are more new features that we’ll announce in the near future, but for now we hope you’ll enjoy the new site.

    We are currently fine-tuning the forum for your browsing pleasure, so bear with any lingering dust as we work to make Styleforum even more awesome than it was.

    Oh, and don’t forget to head over to the Styleforum Journal, because we’re giving away two pairs of Carmina shoes to celebrate our move!

    Please address any questions about using the new forum to support@styleforum.net

    Cheers,

    The Styleforum Team

    Dismiss Notice

Starting an investment business

Discussion in 'Business, Careers & Education' started by NameBack, Oct 19, 2011.

  1. legorogel

    legorogel Well-Known Member

    Messages:
    425
    Joined:
    Feb 26, 2010
  2. Sir Humphrey Appleby

    Sir Humphrey Appleby Well-Known Member

    Messages:
    1,921
    Joined:
    May 28, 2011
  3. wootx

    wootx Well-Known Member

    Messages:
    519
    Joined:
    Mar 2, 2009
    Arrested on fraud?
     
  4. gettoasty

    gettoasty Well-Known Member

    Messages:
    12,277
    Joined:
    Feb 8, 2010
    Location:
    Home
    You may try speaking with someone over at WBI regarding your regression methods. They specialize, in their back office, something similar
     
  5. NameBack

    NameBack Well-Known Member

    Messages:
    2,882
    Joined:
    Oct 4, 2010
    Been a while since I've ventured out of DT -- forgot about this thread!

    Anyway, we're off our peak a bit -- down to about 14%, so nothing catastrophic. Everything continuing as normal, still pursuing leads for anchor investors.
     
  6. SkinnyGoomba

    SkinnyGoomba Well-Known Member

    Messages:
    12,589
    Joined:
    Jan 3, 2008
    Location:
    The American Gardens Building, West 81st
    Pretty solid, keep up the good work.
     
  7. NameBack

    NameBack Well-Known Member

    Messages:
    2,882
    Joined:
    Oct 4, 2010
    So we're still flat from January. This led us to do some experimentation with our backtesting, and we found that the correlation between performance at the end of two years is not as well correlated to the third year's performance as we thought -- this obviously has a significant implication for our method of selecting algorithms to use (we select which algorithm to use going forward by looking at the hypothetical performance over the last two years).

    Basically, we freaked out.

    However, I've managed to come up with a more rigorous selection method and I'm getting an r-squared of about 0.5 right now, and I'm thinking I can drive that up further with more experimentation.

    All in all, I think we can fix this bug but it will definitely come at the cost of lower expected returns, because it's going to force us to average the outcomes of a bunch of algorithms instead of betting the whole farm on a single algo, in order to reduce our risk appropriately in light of the new findings.
     
  8. JoelF

    JoelF Well-Known Member

    Messages:
    1,459
    Joined:
    Dec 31, 2007
    Location:
    MA & NYC
    Nameback just out of curiousity what exactly are your qualifications to manage money?
     
  9. NameBack

    NameBack Well-Known Member

    Messages:
    2,882
    Joined:
    Oct 4, 2010
    

    None! I have no experience managing money prior to this. My experience is strictly in statistics and analytics.

    Obviously that makes this whole endeavor more difficult from the fundraising side, but we think with a good year under our belts and a lot more backtesting and analysis, we can recruit funders.
     
  10. StephenHero

    StephenHero Well-Known Member

    Messages:
    14,187
    Joined:
    Mar 10, 2009
    

    Don't forget 'bad ideas'.



    No, but seriously, I'm relieved you've found success outside the Beltway.
     
  11. NameBack

    NameBack Well-Known Member

    Messages:
    2,882
    Joined:
    Oct 4, 2010
    

    You think I'll stay outside the Beltway forever? What do you think I'll do with all that money when this finally pans out? :devil:

    Also, it appears that we may have stumbled on the solution to our problem. While finding a consistent method of high correlation between periods of performance has eluded us, we've realized that the first 3-6 months of any algorithm's performance tend to be its best months. Therefore, even picking algorithms at random, if we diversify (both in terms of algos and equities) and restrict our models to durations of no more than six months out from the build/training period, we think we should be able to clear strong returns.
     
  12. gettoasty

    gettoasty Well-Known Member

    Messages:
    12,277
    Joined:
    Feb 8, 2010
    Location:
    Home
    So are you hoping to eventually sell your algorithms or put it to use and raise your AUM?
     
  13. StephenHero

    StephenHero Well-Known Member

    Messages:
    14,187
    Joined:
    Mar 10, 2009
    

    I have a feeling that your brain will be a less fertile ground for half-baked socialist schemes now that you're the one footing the bill.
     
  14. NameBack

    NameBack Well-Known Member

    Messages:
    2,882
    Joined:
    Oct 4, 2010
    

    Put it to use and raise our AUM.

    We've already raised enough money for all the legal and administrative costs required to set up a hedge fund; we're just waiting to build a track record and make sure everything works before we pull the trigger on founding the fund.
     
  15. NameBack

    NameBack Well-Known Member

    Messages:
    2,882
    Joined:
    Oct 4, 2010
    New backtesting with new more rigorous methodology is lookin' fucking sick.

    We abandoned trying to find any kind of correlation between returns during different time periods, because there wasn't any.

    So instead we select ten models at random, average their positions together, and use them for a two week period before rebuilding all the models and doing it again.

    Thus, the results of our backtesting represent exactly what we would have earned on a given equity in a given year, because this methodology eliminates any hindsight whatsoever.

    So far our average annual return is looking like 75%, but it's still very early. It's much more labor intensive to test using this process, since we have to build and backtest 26 sets of models for every year (and there's 864 models per set, which means ~22,000 models per year, which ALL have to have to be backtested).

    Exciting stuff! Will keep you guys posted.
     
  16. otc

    otc Well-Known Member

    Messages:
    14,201
    Joined:
    Aug 15, 2008
    

    You're not still trying to do this all with a bunch of excel formulas are you?
     
  17. Nereis

    Nereis Well-Known Member

    Messages:
    1,374
    Joined:
    Feb 12, 2009
    

    I'd imagine he has something more powerful, like matlab.
     
  18. SkinnyGoomba

    SkinnyGoomba Well-Known Member

    Messages:
    12,589
    Joined:
    Jan 3, 2008
    Location:
    The American Gardens Building, West 81st
    Throwing around numbers like 75% annually seems a bit aggressive, IMO.
     
  19. NameBack

    NameBack Well-Known Member

    Messages:
    2,882
    Joined:
    Oct 4, 2010
    

    Some of it is done in Excel, but a lot is done in R.
     
  20. NameBack

    NameBack Well-Known Member

    Messages:
    2,882
    Joined:
    Oct 4, 2010
    

    It's no guarantee of future returns, but like I said, our backtesting is totally unbiased, so that is what we would have earned. I'll post a chart of our results so far in a bit.
     

Share This Page

Styleforum is proudly sponsored by