In general my strategies aim to harvest well known and documented sources of return, that are driven by risk premia and advantages exploiting investor´s consistent behavioral inefficiencies – sources of return that could be called beta and alternative beta. These sources of return are more reliable, as they have a good reason for their existence and are likely to be persistent.
AQR´s Cliff Asness posted an excellent discussion on “How Can a Strategy Still Work If Everyone Knows About It?“, – well worth a read as a more in depth background.
Because it is so popular and so much rapid innovation and democratization is taking place in quantitative trading, it´s interesting to think about what trading an alpha factor system implies in practice:
Searching for alpha means, that you are trying to find fresh return sources that are known to only a small number of people. It can be found by using new, alternative data sets or higher frequency data with powerful software tools (or possibly using a unique discretionary approach). Classic datasets have been analyzed for so many decades, that all reliable factors have long been found and are no longer called alpha.
Finding a new edge is highly attractive, because it will show very high risk-adjusted returns. You can hear about strategies boasting Sharpe ratios of 3 to 5, yielding super high returns above 30% or even double digit sharpe ratios in high frequency trading.
Sounds great, but to me there are several practical problems with such a strategy:
- It´s unreliable. How do you know, that you are one of very few who use this alpha factor or how long that state will last? High potential returns attract sophisticated competition, which will quickly erode the edge – known as alpha decay. Deterioration of alpha strategy sharpe ratios is like a law of nature in investing – much like a hundred dollar bill, that will not lie in the street for long, even if it happens to be hidden underneath a bush. Therefore it is necessary to constantly discover new alpha factors. Recent development in HFT is an interesting example: rapid innovation and high investments in fast technology (we are talking about many millions), earned profits with double digit Sharpe ratios, until intensification of competition arbitraged most these profits away in the space of just a few years.
- It´s risky. Because you don´t know how much your system´s performance has already deteriorated and at what pace it loses its edge in comparison to your backtest period, you have to trade it conservatively to avoid overbetting and blowing up your account. This will make the strategy a lot less attractive in real life. For example a strategy with a Sharpe ratio of 4 that erodes to 0 in an unknown period of time would have to be traded at sub-optimal leverage to have a margin of safety and constantly be deleveraged further to adjust for the erosion. As you can´t determine exactly when to stop trading the system, because it has stopped working, the average performance will be much lower than backtest statistics suggest. You can see this effect taking place when looking at the performance histories of different trading strategies (for example at collective2, a platform that offers investors subscriptions to quant strategies): the older these strategies get, the lower their Sharpe ratio tends to become and their annual return drops steadily, with very few strategies managing to stay above market beating returns (and these usually closely follow the best market of recent years or are strategies with big hidden tail risks that just haven’t materialized yet – e.g. short volatility).
- It´s expensive. Your competition invests in very expensive data and technology and you have to do the same or face severe disadvantages because of asymmetric information. Recent democratization of tools and data through quant platforms have leveled the playing field, but you can be sure that the frontier is constantly being pushed outwards.
- It´s complex. Because high Sharpe ratios are most likely found at higher trading frequencies, it will make it harder to separate the signal from noise. Fine tuning and combination of several alpha factors are necessary to cut through that noise, but more moving parts and specific parameters make it hard to avoid data mining and the complexity will lead to a less robust, less reliable strategy.
These reasons would make it hard for me to trust such a strategy. The high pressure to constantly innovate to stay competitive is not very attractive to me – it goes against my conviction that patience and a long term horizon play a large role in winning the investing game.
In his 1991 letter to shareholders, Warren Buffett noted, “Our stay-put behavior reflects our view that the stock market serves as a relocation center at which money is moved from the active to the patient.”
This doesn’t mean it can´t be done very successfully. I think, most promising would be finding niche data that is quite unique and innovative, anything openly available is bound to disappoint as it is analyzed and results are implemented by thousands of smart people simultaneously.
To me still the greatest example, because of its very long history of constantly staying the best, is James Simons´s Renaissance Technology´s Medallion fund at 35% annual return over decades – and that is after subtracting the highest fees out there!
Know, who you´re up against (a powerful common sense principle):
- Renaissance has roughly 300 employees, about 90 of whom are Ph.D.s.
- “They are the pinnacle of quant investing. No one else is even close.” Andrew Lo.
- Renaissance’s computers are some of the world’s most powerful. Its employees have more—and better—data.
Back to reliable, alternative beta and why that is nothing to sneeze at.
Given the deterioration of alpha strategies´ sharpe ratios, the – at first glance – low sharpe ratios of beta strategies (usually well below 1) suddenly look superior in practice. The longer the time horizon, the more likely it becomes, that returns from beta beat an unchanged alpha strategy.
There are strategies, that make a journey from alpha to beta, where they are known, but still work and are called alternative risk premia, factors or smart beta.
Catching that ride can be immensely profitable, but is very hard to do and you will most likely face cyclical periods of losses on the way. You earn returns on risky assets over the long run because they occasionally go through long stretches of poor performance. Winning on average is the compensation you get for the times you lose – only if those are very painful. The highest risk premia can be expected for assets that do particularly poorly in bad times.
That means sticking with a strategy (even if it´s simply holding a diversified portfolio of assets) is hard enough, even if there is a lot of data showing that it is very likely to be profitable over the long run. Great investment approaches like Warren Buffett´s or the Yale endowment´s seem obvious only in retrospect.
I feel that it´s unlikely that I´ll be pioneering a new concept that works for decades.
A diversified portfolio of known strategies with a stable long term history of a combined Sharpe ratio of around 1 on the other hand, could quite safely use leverage to produce 20% annual returns with 20% volatility – if you can find it.