Back to Community
Definition of a good Stat Arb

The question is in context of low frequency strategies that balance say quarterly and rely on fundamental data.
What is the acceptable range of risk/performance metrics ? I see such strategies do have an appeal of low beta but they seem to move in regimes , work for a few years then lose the steam and tend to be in drawdowns for a lot of months(or maybe I haven't found my golden ticket yet). Do these strategies have to be profitable each year of the backtest or be profitable just overall ignoring the drawdown years . Maybe someone from Q can comment on this in general and in the context of hedge fund .

4 responses

I can't answer most of your questions, but I have heard it mentioned that the fund is targeting algos with a trade frequency that is more than once a month and less than twice a day. Your example of quarterly re-balancing doesn't fit that criteria, but I'm sure they would make an exception for an exceptional algo. :)

Well, statistical arbitrage seems to be a very elastic term. Arbitrage is a more clear term. In principle and in academic use an arbitrage can realize risk-free or low risk profits. Imho in market slang statistical arbitrage refers to nearly everything that generates profit. A lot of this strategies should imho be called a (risk containing) trading strategy and not arbitrage.

Consuli

Hi Yatharth,

Right now we are looking for algorithms that trade more frequently, turning over the portfolio value between a few times a day and say once a month at the very low end. The reason for this initial preference is that more frequently trading algos accumulate independent 'bets' or 'decisions' at a higher rate, so we need comparatively less out of sample data to gain confidence in the algorithm's behavior and profitability.

For a strategy that makes quarterly or fewer decisions a year you could easily argue that you'd need to see years of out of sample performance before you felt confident in attributing returns to skill as opposed to chance.

Setting the frequency question aside, you also asked about how consistent the performance should be to be considered 'good'. When I evaluate algorithms for allocations I am indeed looking for very consistent positive returns month over month and year over year.

I'll actually be giving a webinar on the topic of using the pyfolio tearsheet in our research platform to grade your own algorithms on Thursday of next week. Please join us or catch the recording after the fact.

-Jess

Disclaimer

The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by Quantopian. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. No information contained herein should be regarded as a suggestion to engage in or refrain from any investment-related course of action as none of Quantopian nor any of its affiliates is undertaking to provide investment advice, act as an adviser to any plan or entity subject to the Employee Retirement Income Security Act of 1974, as amended, individual retirement account or individual retirement annuity, or give advice in a fiduciary capacity with respect to the materials presented herein. If you are an individual retirement or other investor, contact your financial advisor or other fiduciary unrelated to Quantopian about whether any given investment idea, strategy, product or service described herein may be appropriate for your circumstances. All investments involve risk, including loss of principal. Quantopian makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances.

Thanks Jess , I look forward to the webinar.