High frequency trading? Ironic that the best long term returns came from short term strategies. Like most people, my favorite holding period is forever in theory. In reality it is rarely feasible and never optimal when economic volatility, creative destruction, business innovation and global instability are permanent market phenomena. The previous post was on space - people should only invest in good alpha opportunities anywhere. This post is about time - buy and hold and low frequency trading do not diversify enough so investors also need higher frequency strategies in their portfolios. Different holding period alphas reduce total risk.
Best fund manager for this new decade? The chances are that fund is not yet in existence though I have visited some interesting start ups recently. What about the best strategy? That strategy has likely yet to be invented. However some future trends are certain, like the growth of the alpha vendor industry AUM from the currently tiny $2 trillion to at least $10 trillion by 2020. Some $100 billion funds will appear, many not established today. High frequency trading was the best strategy in the 2000s so it will not be the winner again with so many entering the field dragging down "aggregate" returns. The obscure is now mainstream but the top exponents will thrive, extracting alpha out of the many more high frequency wannabes.
To hedge or not to hedge? That is the question. Exactly ten years ago I presented at a conference on "whether" institutions should invest in skill-based absolute return strategies. Conversely the key takeaways from some financial gurus speaking were that investors didn't need hedge funds and likely didn't even need long only portfolio managers to select securities since "cheap" index funds were bringing in +20% a year and could be "assumed" to return +10% in the long term. Fundamental and quantitative analysis were apparently a waste of time as was paying the "cost" of hedging. Be satisfied with "market" returns. Asset allocation was risk management!
Higher frequency of investment decisions matters at all levels. We know from failed investment policies that static beta asset allocation is not sufficiently reliable if you have retirement liabilities to fund. What does work is the dynamic triple alpha process of ongoing portfolio structuring, strategy selection and manager due diligence. You would not exist were it not for the triple alpha of nuclear fusion and the chances are your portfolio will not perform over the long term without triple skill fusion. Absolute returns fuse into spendable cash unlike the relative return universe.
Apart from that, the public markets were "efficient" and so there was no such thing as investment skill; here are some other "facts" I noted down in March 2000:
- the apotheosis of risky buy and hold to its exalted and unjustified status
- hedge fund capacity was limited and $1 billion was "too large" AUM in one fund
- "high" 1 and 20 fees were "certain" to drop. They are now typically 2 and 20
- after a poor 1999, CTAs, global macro, vol arbs and short sellers were "finished"
- long only private equity and real estate were independent of the stock market (!)
- short term trading didn't work and was unnecessary for long term investors (!!)
- large cap equities in major markets were always efficiently priced so offered no opportunity for sustained alpha capture, so forget about high frequency trading (?)
Of course all those financial beliefs were ludicrous and skilled high frequency trading went on to be the best returning strategy in the 2000s. Long term performance doesn't require a long term holding period. Most of the people in high frequency until a few years ago were good, but recently copycat rookies have started crowding into an area where they think they have the expertise to compete successfully. This creates more trading opportunities and capacity for the best players. Contrary to common advice, the more liquid and number of participants in a market the more inefficiently and wrongly priced the securities become. Irrationality does not cancel out so there are more anomalies and mispricings than ever before. Alpha is abundant but the skill to extract it is rare.
The evolution to very short holding periods is the logical progression of Grinhold and Kahn's Fundamental Law of Active Management equation: IR=IC.TC.SQRT(breadth). Translating the formula, if you have an investment edge then apply it as often and as widely as possible. The transfer coefficient is how efficiently active bets can be implemented and lower trading costs help increase that. Breadth is the number of securities where you can apply your edge. The more skilled bets you can make, the better the information ratio. Ipso facto an active manager delivers the most value to clients by trading as many securities as frequently as their competitive advantage allows, provided they have talent, excellent execution and are unconstrained on longs and shorts and what they are mandated to do. Hire managers to make money for you how they see fit not to give you unhedged exposure to the ups and downs of an asset class.
Despite being considered "new", temporal arbitrage has been utilized for centuries. There is nothing modern about exploiting time advantages. Didn't the Nathan Rothschild credit hedge fund make money out of slower investors in 1814 with early news of the Battle of Waterloo outcome? Munehisa Honma's managed futures hedge fund back in 1753 constructed a high frequency data transmission and execution platform by stationing village runners as information conduits from where rice was traded to where it was grown. Momentum, mean-reversion and statistical arbitrage have been around a long time and when applied skillfully work on numerous time frames.
Samurai trading: the time between the decision to trade and executing that trade must be minimized. The quicker and better you are at information gathering and analysis then the higher the performance. The edge in high frequency is often slippage minimization and better transaction technology. Bid offer spread capture blurs the line between market makers and market takers. The fractal nature of markets means that the main constraint on capturing opportunities from microstructure and macrostructure were trading costs.
Technical analysis supposedly doesn't work so people employ semantic arbitrage and refer to it as pattern recognition instead. The seminal studies are correct: publicly disclosed "charting" methods are useless. However proprietary predictive black box models continue to perform outstandingly as many quantitative hedge funds have demonstrated over the long term.
The move from millisecond latency to measuring execution in microseconds was inevitable but now people are even talking about nanoseconds. Light takes more than one nanosecond to travel from the screen you are now looking at to your eyes. Picoseconds next? At least Planck time and Einstein relativity put a physical floor on how fast trading can ultimately get. With co-location competition, the time arbitrage arms race is reaching its zenith which puts the emphasis on developing better intellectual capital.
There is still plenty of money to be made out of the unskilled in high frequency strategies and capacity is expanding. Very liquid ETFs like SPY, QQQQ, EEM, IWM, UNG, EWJ and XLF already have most of their volume from shorter term strategies. The E-minis and KOSPI futures are probably the best trading vehicles on the planet and being ultra liquid, are, of course, the most wrongly priced, which creates a lot of alpha opportunities for talented traders.
When you buy a security you might hope to hold the stock for decades or the bond to maturity but the reality is that a short term outlook is usually optimal for risk management. Commodities and currencies are fantastic trading vehicles but never for buy and hold. The many gold bulls out there might recollect that GLD and SLV remain mired in a six hundred year old bear market and there are wiser ways of hedging inflation or for difficult times nowadays.
There is a strange idea circulating that short term trading serves no economic purpose. The investors that did allocate to high frequency are today better funded than those that concentrated solely on long term stocks and bonds. Alpha always has superior risk-adjusted returns than beta. Surely added liquidity is good for everyone. Those markets that heavily tax trading or ban short selling have deeper drawdowns and higher volatility than those that do not. Investors gain from lower transaction and slippage costs. The events of 2008 would have been worse were it not for the liquidity provided by automated and systematic traders. If you must make a fire-sale during a crash, the presence of buyers is essential. High turnover of a portfolio isn't bad and is often essential to control risk.
Whether carbon-based or silicon-based, sapient beings of all kinds can succeed in quantitative short term investing if they work hard enough and spend many years building core expertise in the hard sciences without the luxury of a proper salary. Natural language processing and compressed sensing can help determine the probability of near future moves when you have models to analyze recent historical information and identify order embedded in chaos. Sparsity of data is an occupational hazard in the prediction of financial markets but tick data provides large information sets to detect hidden structure. Hidden to the ridiculous random walk ranters, that is. Low frequency managers need to invest for years before we can be sure it wasn't luck. The more trades you do, the shorter the track record needs to be to demonstrate skill.
Rightly or wrongly, the world increasingly functions on short term factors. Therefore as an investor you have the choice of fighting the trend or accepting the high frequency attention span of most market participants and mainstream media. We live in a Twitter world where what is hot today is not tomorrow. Stock trading is already a level playing field. Algorithmic execution systems are arbitraged by better algorithms. Flash orders and sniper, guerilla or ninja algorithms are available to anyone prepared to pay the high price of access, hardware and software development costs. This also creates opportunity for long term investors that have the ability to find good securities amid the fluctuations. Make money from the volatility (HFT) or through the volatility (LFT)? Both, but one firm cannot be good at everything which is why broad manager diversification is necessary.
There are very few long term winning securities and price predictability declines sharply with time horizon. Consistently accurate forecasting is extremely difficult but investing with a 5 millisecond outlook has more probability of success than 5 years. Amazingly brilliant are the clairvoyants that "know" the stock market will "definitely" be higher in 2040 than today. Wish I also had a time machine that could look ahead that far. Considering the 1910-1940 and 1810-1840 eras I wonder why they are so confident this time around. They obviously know something I don't.
I have heard it said you need 10,000 hours to get good at something. To get good at investing probably takes 10,000 separate trades or at least the thorough analysis and due diligence of 10,000 investment ideas. As a researcher at Renaissance Technologies recently noted, "We try to find these very obscure patterns hidden in a lot of noise". There is also a vast amount of noise in portfolio construction and fund selection but one signal is clear: Strategy diversification with different managers whose holding periods range from femtoseconds to decades. The solution for consistent capital growth already exists and every investor needs high frequency trading strategies in their portfolios.