Sharpest guy on the contest floor (July 2015)

I would like to congratulate myself on writing the algorithm whose live-trading Sharpe ratio was ranked #1 in the July contest:

Annualized returns of 90.68% (ranked #24), beta 0.4105 (#281), Calmar ratio 32.36 (#9). Thus I'm the Sharpest guy on the July contest trading floor. Either I'm a genius, or the market was bad when the algo was started (Greece), or the assets did exceptionally well (earnings), or I'm a genius, or all of the above.

The algo missed the low-beta badge, even though I hoped for just less than 0.30; I suppose it was due to the sharp post-earnings surge. It missed the "hedged" badge, because this requirement was announced too late for me to incorporate it in my code. There are 10 algorithms with higher scores and all three badges and 9 without; so if badges were disregarded and ranking was based on scores only, it would be ranked #20 out of 683. Not bad.

9 responses

@André - Congrats on your high quality algo!

1) Isn't the beta badge based only on the 2-year backtest? If so, you should be able to submit your algo before the deadline to see your beta score, and adjust your algo until it meets the requirement. It looks like yours was 0.385 during the backtest period, so you knew you wouldn't get a badge even before it started.

2) You could have made the top 20 simply by using lower leverage! Try running your same algorithm in this month's contest with a few different leverages, for example: (0.10, 0.20, 0.30). Your beta would drop below 0.30, and your volatility based metrics (calmar, drawdown, volatility, stability) would also drop. The winners of this contest have low returns and even lower volatility.

3) Regarding hedging, simply make a small purchase of a long and short stock and hold it for the entire competition. I don't like the current hedge requirement, but this will get your badge this month. I agree with the concept of hedging against down markets, but I don't like the way it is being enforced. This is partially being enforced by the the low beta requirement (but a portfolio of 100% SPY could still get the beta badge with a leverage of 0.30 or less!) I hope Quantopian comes up with a better way to enforce hedging.

Good luck on this month!

Congrats on the #1 rank André! A couple more tweaks to the beta and hedging, and I bet the algo will rise further in the ranks. Maybe something to try out with your 2 remaining entries? Keep in mind that the most talked about prize is the \$100K, but the lurking even bigger prize is selection into the fund. If you algo has all 3 badges and performs well, it's a solid contender. So as they say, keep your eye on the prize!

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.

Thank you, Alisa. My August algorithm is similar, but also is "hedged" according to Quantopian's definition. It replaced one that attempted to have a beta of 0, because "beta [was] king", and because I was curious to see what would happen to the Sharpe ratio calculations if it achieved it. ("Sorry - something went wrong on our end. Sorry for the inconvenience."?)

As far as controlling beta, I might take Tristan's suggestion even further, and lower my leverage to 0.01. I'll still be "hedged", make money, have a very low beta, and returns don't seem to matter much.

@Alisa - Thank you also for reminding me of the hedge fund, if and when it starts. My Icarus' wings have softened a bit (Vladimir might make a remark about various belt colors here), and the algorithm now has a Sharpe ratio of 7.801 (rank #19), paper-trading score of 79.72, and annualized returns of 79.91%. Two other algos I wrote, running but not in the contest, have better Sharpe ratios: one, trading since July 21, has a 9.94 Sharpe and 4.01% actual (not annualized) returns; the other, since June 25, 9.28 and 7.05%.

Hi Andre;
Congrats on High Sharpe, but than high Beta spoils the party. . . .

My simple test algo is giving 1/2 of your annual return but less than half the beta in paper trading.
its ranking very close to your's.

It even got highest score for one day on leaderboard
https://www.quantopian.com/posts/got-lucky-and-unlucky-at-the-same-time

Regarding inputs by Tristan on leverage. (my personal opinion)
Leverage reduced to 0.2 means why write the algo in first place. If a fund is going to keep
80% cash who would want to invest in it ? I would not recommend you this path.
You better come up with an algo with 0.7 ~ 1.1 leverage and beta at +/-0.2~0.3 this definatly takes
tall on sharpe but staying above water for long will help.

On hedge issue I am waiting for the sky to clear.

Best wishes to you to come up with much improved idea and algo.

@Yagnesh: I agree that using low leverage will not create the type of algos that Q wants in their fund. However, based on their current scoring system, using low leverage will help an algo achieve a higher score. Therefore, Q needs to continue to evolve their scoring system until it provides the right guidance on our algo designs.

I don't know what the best scoring system would look like, but a good start would be less focus on volatility and higher weighting for risk-adjusted returns (Sortino, Shape, etc). Low volatility should still be important, but it should not be the primary goal, like it is now.

@André: As the sharpest performer, I have to ask you - do you have any pointers? (Ha ha ha ha...c'mon finance crowd loves that jokes no?)

On a more serious note: are folks doing much serious collaboration? Joint contest entries? I'm curious as to people's behavior.

Right now I get the feeling folks are guarded about giving away their 'contest alpha', which means they are posting screen caps of perf algo or maybe if they're generous posting an initial algo. Once refined they're not going to just share it - folks would clone it, maybe make a few tweaks and beat you out.

@Robert - I have many pointers, some of them NULL. I also have many references, some of them null, some None. That's a C/C++, Java, and Python joke. Options coming soon, hopefully.

I had the same thoughts on collaboration. I asked one user to collaborate on an algorithm he needed help with, but got no reply.

@Robert - I expect we love jokes just as much as bears love knocking bee honey down from trees, and bulls love the color red (but shouldn't that be green)?

This forum has a lot of public collaboration, usually in the form of helping someone perform fundamental tasks. Selecting stocks, sorting lists, collecting indicator values, placing orders, tracking orders, logging output, etc. It is true that people hide their best ideas (what Grant calls "secret sauce") from the public.

Regarding private collaborations, I just tried my first one this past contest. Basically, it gives you the ability to share an algorithm with someone, including backtests, and a simple chat interface. We quickly discovered that two people editing the same code is a recipe for disaster. We decided to create three shared algorithms: My working copy, his working copy, and a master copy. How are other people managing private collaborations?

The part I was uncertain about is how to submit a shared algo to the contest. Based on the contest rules, if two people submit the exact same algorithm, the profits will be split between them. Do we need to have both people submit the algo to the contest? Or does having a collaboration already tell Q that we want to share the winnings?