Rage Against the Machines
August 08, 2011
– Comments (4)
Black Swan dives, flash crashes, irrational exhuberance, market manipulation - it all stinks to high heaven, and so many people see the root of all evil to be HFT. Well, I'm definitely not here to defend HFT, nor am I here to condem it - this is just to show a quick counter to the notion that we can't win against HFT. This is all about CAPS. All about the strength of community intelligence. All about the benefits of cooperative learning.
We spend a lot of R&D time on CAPS trying to measure the value of the information embedded in our community, measuring it against such traditional models as CAPM (capital asset pricing model) or its more sophisticated cousin, the Fama-French Four Factor model, to determine if we have any truly excess performance that can be found. We also work in modeling how individual players score and continue to analyze the systemic scoring structure of CAPS itself to look for better ways of extracting information or enhancing player activities so that we generate more unique or robust information. And we also do something far more practical: we model different portfolio creation strategies to find practical ways of using the CAPS information. Below are just a couple of examples of how we have built portfolios from the CAPS data to show that whatever arena the HFT traders are playing in, they haven't eaten our lunch yet.
The first example below shows the excess return of a long only portfolio built from CAPS. This does not use only the CAPS ratings, but also integrates some fundamental factor modeling and imposes liquidity requirements on the stocks to make sure they are tradeable for a portfolio ranging in the $200M - $500M size.
The graph above shows the cumulative excess return of this portfolio relative to the returns of the SPY. Though a little bumpy, the trajectory of this return is fairly consistently upwards, with an excess return on the market of ~9.5% annually. It definitely suffered in the past two weeks during the sharp market downturn that we've experienced, falling an additional -4.5% below the SPY, so at this particular instance it is losing in the short term, but the trend to date has been strong. It will be interesting to watch this model evolve if the market takes a downward turn, given that it actually managed to outperform the market during the worst of the downturn in 2009.
The next model is one that I find more interesting from a quant standpoint, which is a long-short balanced portfolio that is designed to be market neutral. This portfolio uses the CAPS ratings plus fundamental factors models plus some additional secret sauce modeling to select long and short candidates, plus some additional hedging instruments to keep everything in balance. Again the "realistic" constraints of liquidity and short availability are included to make sure that this is representative of a portfolio that could operate with $200M - $500M in assets (basically a medium sized hedge fund, which seemed like a good example for determining realistic stocks to use in a model).
One of the most interesting things about this portfolio is the robustness with which it has handled the latest market meltdown. While the SPY has fallen about -16.6% since 7/22/11, this model has dropped only about -1%. The strength of the underperform signal is sufficiently strong to have balanced out most of the sudden fall in the stocks held long, and the overall performance has in fact been essentially flat for more than 4 months now. One of the underlying design constraints for this particular model was volatility control, keeping the (return/volatility) ratio at or above 2, so we settled on a moderate rate of return (~15%) in exchange for significantly lower volatility.
So what do these portfolios have to do with raging against HFT? We're not trading in milliseconds here, looking for penny arbitrage opportunities with zillions of trades. We're harvesting the aggregate intelligence of 100K+ humans and crunching it all on a single PC, and consistently beating the market. Whatever rigging of the system is involved, whatever emotionless algorithm is out there trying to shave bits of the bid/ask spread while they try to eat each others lunch, we are still able to pool our understanding and knowledge to outmaneuver and out-invest in the long run.
Interested in crunching some of the CAPS data for yourself? Drop by http://developer.fool.com and sign up for an account to access the CAPS API and 4 years of CAPS ratings data.
-Xander
RSS Headlines
Fool UK