It is increasingly clear by insiders on Wall St that the preeminent multistrategy hedge funds (Millenium Partners, Citadel, etc) are set to consolidate the hedge fund industry - leveraging economies of scale and large investments in technology infrastructure to compete in every corner of global public markets. These firms seem to be achieving information ratios (risk adjusted returns) in excess of 3, while remaining market and factor neutral. This translates to almost never having losing years, while remaining decoupled from the performance of the market and many other known macroeconomic effects - an extremely unusual performance profile. This achievement can largely be attributed to their unique approach to portfolio construction and risk management.
Suppose our objective is to maximize the information ratio (IR) of an investment strategy. Working forward from the definition of IR, our options are to improve the return of the strategy or to decrease it’s risk. Unfortunately, this insight isn’t terribly useful in practice. It’s very hard to improve returns. Improving returns entails improving forecasts of the future and outsmarting the competition. Those that can do this (and it seems that the vast majority of financial professionals can’t do it) are paid well into the seven-figures for their work. Reducing risk on the other hand entails hedging, which is at best expensive and often not even possible.
There is however another way - diversify over strategies. Multistrategy hedge funds are broken up into teams called pods that are each responsible for trading their own assigned subset of public markets. Each pod is led by a portfolio manager (PM) and is required to be roughly market-neutral, factor neutral, as well as being subject to tight volatility and drawdown constraints. This leads to each individual pod running a responsible amount of risk and with a return stream that is uncorrelated / economically decoupled from all of the other pods. This creates an incredible opportunity to reduce risk by diversifying over the unique, idiosyncratic risks of all the pods. While each individual pod typically has a IR well below 1, the portfolio of pods can have an IR as high as 4. With the largest multistrats like Millennium diversifying over hundreds of pods, portfolio volatility can reach as low as 2-3% with returns around 10% - an extraordinary feat of financial engineering.
The diversification effect that the multistrats are exploring can be modeled simply using the , following formula (also implemented in python). Let’s chart this formula for a variety of values for n and p and see what we find. Perhaps this will help us build some intuition for what is happening behind the scenes.
Let’s think about what we’re looking at here. From this chart, we can immediately note that the scaling of portfolio volatility for all correlations is sub-linear in the number of strategies in the portfolio - each incremental strategy contributes diminishing marginal utility. We also find that the correlation p really matters. At the extremes, if p=1 then there is no diversification benefit no matter how many strategies we add. With p=0, portfolio volatility will eventually converge to zero with enough uncorrelated strategies.
The intuition is fairly simple - just find as many uncorrelated strategies as you can. Unfortunately, producing uncorrelated strategies is exceptionally intricate, primarily because all equities are substantially correlated with each other - long-only strategies are not sufficient. This can be shown by looking at the underlying risk decomposition of publicly traded securities:
What we find is that a bet on Microsoft is actually only about 50% a bet on Microsoft’s idiosyncratic (company specific) risk. The other half is a bet on the market as a whole, the software sector, and Microsoft’s factor exposures. The solution is to explicitly short out these other risks - or create a perfectly balanced portfolio of long and short positions that causes these shared risks to be offset. The latter is a complex task that requires the support of data & compute infrastructure, as well as a fair bit of intellectual property in portfolio optimization. However, if done correctly, the result is a market and factor neutral portfolio - the building block of a multistrat hedge fund.
The multistrat approach to portfolio construction and risk management really is the only practical way to reliably pull high information ratios for long periods of time. Perhaps to the surprise of those outside of the industry, this is not achieved by having prophetic stock picking ability, but rather by having a modest edge in a multitude of independent, idiosyncratic risk factors and exploiting this independence by diversifying away the majority of the risk at the portfolio level. This general architecture is supported by mathematical and statistical theory and in principle can also be applied to your portfolio.
-PCH