For the full presentation on crowd sourced alpha by Dr Jessica Stauth, Click Here, For the Slides, Click Here. Dr Stauth opened with a quote by Ray Dalio which pretty much sums up the fundamental objective of portfolio investment and diversification: “uncorrelated return streams are the holy grail of investing.” Traditional portfolio theory states that to achieve true diversification adding more assets or return streams to your portfolio leads to a better Sharpe ratio and more diversification only to the extent that the assets you add are truly new and not more of the same. The traditional model fails to achieve this because
- Positive Sharpe streams are a moving target which require on-going research
- Most investment teams lack diversity in terms of expertise and background making it challenging to surface new uncorrelated strategies
- Large multi-manager firms are successful BUT the barriers to entry to these large firms industries are extremely high
To overcome these challenges, Quantopian developed a non-traditional approach to pursue uncorrelated return streams through the use of crowdsourcing. The advent of the internet age has made crowdsourcing a viable option since it makes it possible for large groups of people with diverse backgrounds to come together and share their thoughts. As it is a relatively recent coined term I had to trek through a few scholastic articles to discover what fellow academics defined it as and through my trek of discovery I managed to find this definition:
Crowdsourcing is an online, distributed problem solving and production model
Daren C. Brabham
Or by the” tried and tested” Wikipedia, (which Dr Stauth says better explains the Quantopian model):
Crowdsourcing is the process of obtaining needed services, ideas, or content by soliciting contributions from a large group of people, and especially from an online community, rather than from traditional employees or suppliers.
Using this model, the Quantopian team developed an approach to investing that digressed from the traditional approach and made use of some of the major benefits crowdsourcing has to offer. In essence they took the platform and the vast human capital offered by crowdsourcing (versus only using employee capital). To encourage engagement from the online community, Quantopian put together a paper trading competition called the Quantopian Open as an incentive to engage the online community and, with their help, discover new ways of achieving uncorrelated return streams.
THE QUANTOPIAN OPEN
This is a paper trading competition. Paper trading was chosen because it has no barriers to entry. At the end of every month Quantopian awards a $100 000 prop trading account to the algorithm that comes top of their leader board. This algorithm runs against Quantopians prop trading account of $100 000 for a 6 month time period. At the end of the 6 months the author will keep 100% of the profit earned.
For the initial ranking, ALL algorithms in the competition are measured against 7 measures:
- Annual returns
- Sharpe ratio
- A stability measure
- Beta to S&P500
- Calmar ratio
In the early stages of the competition, Quantopian realised that the majority of algorithms selected were correlated to the market, and therefore to each other. A larger pool of uncorrelated return streams was needed for the fund in order for there to be more diversity and thus a greater likelihood for uncorrelated return streams. Beta then became a large disqualifier in terms of algorithms and which ones they wanted to invest in. They adjusted the rules and Beta badge of trailing 12 month beta of -0.3 were considered as the top ranked algorithms.
ATTRIBUTES QUANTOPIAN LOOK FOR WHEN SELECTING ALGORITHMS FOR THE FUND
- Low exposure to the market (algorithms with a market beta of +/- 0.3)
- Consistent Profitability. Algorithms with consistent Sharpe ratios and with Sharpe ratios over 1. Algorithms with volatile returns or that are losing money are not helpful for their portfolio selection.
- Active Trading, i.e. how often the portfolio turns over its entire dollar value for example annually. They showed a preference of once a month at the lower end and twice a day at the higher end.
- Low correlation to peers. Preferably a -30% correlation to other algorithms in their portfolio in order for the algorithm to make a contribution to their portfolio. The goal is for a diverse portfolio and not so much of the same.
- Strategic Intent driven by some underlying economic reasoning versus correlations developed by a big data mining exercise.
ATTRIBUTES QUANTOPIAN DO NOT LOOK FOR WHEN SELECTING ALGORITHMS FOR THE FUND
- Data snooping
- Spurious correlations
AREAS TO EVOLVE
Whether the back-test results created on a given strategy are matching up to an indeed seam predictive of the out of sample results. One thing they looked at was fitting distributions to the daily returns of the back test and the out of sample period. Something to look at was that your back-test is only as valuable to you as it is predictive of how that algorithm is going to do in the future. Going forward they are going to continuously develop in-house tools to evaluate algorithms.
… SHOULD YOUR ALGORITHM GET SELECTED TO THE FUND
In the event that an algorithm gets selected to the fund, the owner has the option to licence Quantopian for use in their fund in return for a 10% royalty based on the profits earned by the licenced algorithm. At the end of the day though, the owner of the algorithm retains ownership of their intellectual property. Quantopian will take responsibility of operating the algorithm, allocating capital and managing the risk.
It is hard to break tradition, but a sure way to do so is by breaking the bounds of conformed thought. I for one am definitely looking forward to seeing whether Quantopians “New Millennium” approach could be one of those ideas that break the bounds.