A brief note to help with orientation: We’re in Section Three of the book, which is called Dealing With Noise. So far in this section, we’ve had a look at dealing with noise in the personal domain, had a detour into what I call modes of functioning, and now we’re back to dealing with noise, but in the market domain. There’ll be two chapters in this subsection, Process Design and Process Discipline, and then we move onto dealing with noise in the people domain for two chapters, after which I’ll attempt an elegant dismount.
In this chapter of the book, which was published in 2019, I shared a few superficial thoughts on AI. As you know, things have fairly skipped along in the last few years, so my musings on the subject look rather quaint now. Anyway, let’s get to this process-design business, shall we?
Briefly: Process Design
Noise. The better the design of your investment process, the less noise you will generate and experience.
Missing. The design of your investment process is probably missing the most important bits:
Edge. An ongoing assessment of the existence and nature of your edge.
Pain. An accurate assessment of how much pain you can take.
Discipline. Don’t confuse design with discipline. To be disciplined about sticking to a poorly designed process is to be disciplined about destroying alpha.
Your Friday night poker game is going to be a costly form of entertainment unless you’re clear on three variables: the specifics of each hand that you’re dealt (obvious); the size of your bankroll (fairly obvious); and your skill relative to that of your opponents (not always so obvious). Your edge can be difficult to assess, but there’s a useful rule of thumb for that: every half-decent player knows that if you sit down at a poker table and you don’t know who the patsy is, you’re the patsy.
That’s not much different to your day job. As a professional fund manager, you also need to balance three sets of variables: the market conditions; your client’s mandate; and your competitive advantage.
Here’s my guess about how you do this.
Conditions. You spend virtually all of your time doing the important work of analysing securities, managing portfolios, and reporting to clients. That seems fair enough because that’s what you’re paid for, isn’t it?
Mandate. You spend a little time understanding your client’s objectives and constraints when they first transfer their business to your firm, and then the guys in Compliance keep an eye out for any transgressions. That should do it, right?
Edge. You spend practically none of your time consciously assessing the investment capabilities of your firm and team in relation to those of your competitors. You’ve got a team of people who are highly educated, highly experienced, and highly paid. That ticks the boxes, surely?
A poker player who plays without an edge is relying on luck, and that’s not a smart approach if you intend to be at the table for a while. Similarly, in the zero-sum game of trying to generate alpha, if you play for long enough without an edge, you will be carried out. There’s a big pool of money that’s chasing alpha, but most of the managers of that money fail to find it. Over time, 70% or more of active fund managers fail to beat the market averages after fees. Some significant proportion of the field can’t identify the patsy.
There are two secular shifts that account for the difficulty.
Institutionalisation. Over the past 30 years the easily-identified patsy has left the game. The fund-management industry is now largely institutionalised and populated by very smart and qualified people like you. The almost complete absence of uninformed private investors means that the average skill level in the industry has increased, which in turn means that some proportion of the remaining group of very smart and qualified fund managers, by definition, has no competitive advantage.
Machine Learning. Already underway is a second secular shift that forces you to ask with renewed urgency whether there is a whole new cohort of patsies. The confluence of big data, artificial intelligence and machine learning prompted Jason Karp, fund manager at Tourbillon Capital, to ask, “What do you know that a machine cannot work out?” If a machine can work out what you do, it’s a matter of time before it eats your lunch.
The design of your investment process, at its most fundamental, is about identifying your edge (whether you know something that other humans don’t know and that machines can’t learn) and then devising a system (a series of repeatable steps) that utilises that edge to convert data into alpha. The system for conversion can be purely quantitative, it can be qualitative, or some combination of the two. And it can involve the use of machines or not.
Many fund managers are not sufficiently clear on what their process really is, where alpha has really come from in the past, and where it might reasonably come from in the future. What they call their investment process, is not so much something that has been consciously designed, but rather more a hodge-podge of happenstance, history, and hype.
Happenstance. The guys from Marketing get their way for the launch of a new fund in a new area because they say it’s a hot story right now and they can sell a bucketload.
History. The way that clients’ money gets invested is an awkward compromise between the two teams that were thrust together after Management’s most recent attempt to achieve scale.
Hype. Your process is really just a story that you tell to soothe or excite your clients and prospects, but which is only loosely, if at all, connected to what you and your team actually do.
If you don’t want to be a patsy, you need to make a dispassionate assessment of the nature and extent of your competitive advantage, and you need to design a process that exploits your advantage. A lack of clarity about your edge will generate significant quantities of noise; it will make you do things that you shouldn’t do and make you not do things that you should, both of which will result in the destruction of alpha.
When you go through the inevitable fallow patch in performance, you will be tempted to blame it on something. Publicly, the prime suspect is usually an irrational market; privately, the target might be your irrational colleagues. Whatever the supposed cause of poor performance, the inclination to assign blame is a symptom of noise, and noise will nudge you into looking in the wrong places when your process-design is really the issue.
One of the classic errors that fund managers make when their performance is poor is to redouble their efforts at sticking to their investment disciplines. This sounds right and even heroic, but it is only right if you’re being disciplined about sticking to a well-designed investment process. If the process is poorly designed, in that it doesn’t offer the prospect of a better-than-even chance of generating alpha, sticking to that process is just being disciplined about destroying alpha. Better discipline will not fix a design problem.
Risk & Return
One way to think about process design is to do so using a variation on the traditional risk/return framework.
Return. To generate returns that are different to the market portfolio (hopefully better), you need to have a portfolio that is different to the market portfolio.
Risk. When you hold a portfolio that is different to the market portfolio, you will generate returns that are different to the market portfolio (possibly worse).
As an active fund manager, you need to design a process for identifying investment ideas that are different to consensus opinion, and for implementing those ideas in such a way that your portfolio generates alpha. By definition, alpha opportunities exist in the field of non-consensus ideas, and for those ideas to be harvested as realised alpha, a decent proportion of your calls needs to be right. For some other proportion of the time you will either look wrong or you will actually be wrong. Given that this may be too fine a distinction for your clients and other stakeholders to make, you will need to design a process for surviving when things don’t go your way.
The Wall Street legend, Michael Steinhardt, made non-consensus calls that were right often enough to enable his fund to compound at just shy of 25% a year over a 28 year career. He had this to say on what he called variant perception: “The hardest thing over the years has been having the courage to go against the dominant wisdom of the time, to have a view that is at variance with the present consensus, and to bet on that view...In an immediate, emotional sense, the market is always right. So if you take a variant point of view, you will always be bombarded for some period of time by the conventional wisdom as expressed by the market.”
Alpha and noise arise together. Noise brings about the mispricing of securities that creates the opportunity to capture alpha; and the very act of pursuing alpha will generate noise because your positioning will be dissonant with the consensus. Designing a process to help you cope with noise is an exercise in risk management; and designing a process to help you capture alpha is an exercise in noise exploitation. It’s going to end badly if you design a process that delivers you non-consensus ideas, but you have no process for managing the noise that will inevitably accompany those ideas.
When your P&L contains sufficient red ink your clients will doubt you, your boss and colleagues will doubt you, and even you will doubt you. Doubt is but one manifestation of noise, and when it crosses a certain threshold, it will make you do things that you later regret.
This is not adequately captured by the complex quantitative risk-management tools that you have at your disposal. You will add significant robustness to your process if you make a qualitative assessment of how much noise you can stand. If that seems too abstract, ask yourself this: How much pain can I take before I am forced to fold?
Don’t answer too fast.
Your Pain. There is ample research to indicate that we are surprisingly poor at predicting our future emotional states, especially with regard to their intensity and duration. Your pain, when it comes, is likely to be worse and longer lasting than you imagine, and you’ll probably capitulate at the worst possible time.
Others’ Pain. Even if you have exemplary intestinal fortitude, that might not be enough to keep you in the game. There are other stakeholders in your portfolio who can pull the plug, so their pain threshold matters too. Your survival is also a function of how much pain your clients and other stakeholders can take.
The blow-up of Long-Term Capital Management in 1998 is a good illustration of JM Keynes’ famous quote: “The market can stay irrational longer than you can stay solvent.” It also illustrates the point that you can lose control of your portfolio even if you can take the pain.
Example: Long-Term Capital Management
The LTCM team was formed in 1993 by John Meriwether and consisted of a group of elite bond traders from Salomon Brothers and a couple of future Nobel Prize winners. They launched their fund in February 1994 with $1 billion under management and over the next four years they delivered stellar performance, turning a $1,000 investment in the fund at inception into $4,000. However, six months after the fund’s peak in April 1998, its NAV fell to $330, with much of that 92% drawdown occuring in six short weeks during August and September of 1998.
The firm’s main strategy was to arbitrage temporary mispricing between related fixed income securities, such as on-the-run and off-the-run 30-year US Treasuries. The fund seemed to be highly diversified with hundreds of positions in different asset classes and different geographic regions, but in reality the portfolio was almost purely a short-volatility convergence bet, albeit expressed in a myriad of different ways.
The pricing discrepancies between the long and short positions were usually very small, and therefore the firm used significant leverage to generate attractive returns. At the end of 1997, the fund was leveraged about 25:1, excluding its derivative exposure. This leverage juiced their returns while the convergence bets were playing out as planned, but because it allowed them to get very big in what turned out to be a crowded trade, their leverage was the source of most of their pain when their trades started to unravel.
The first tremors were felt by the fund in the middle of 1998 as risk aversion started to take hold in a delayed reaction to the 1997 Asian Crisis. The widening of spreads on their convergence trades caused them to lose 6% in May, 10% in June, and another 10% in July. Then in August 1998, Russia caused a major shock to global markets when it unexpectedly defaulted on its local currency domestic bonds. The ensuing flight to quality meant that LTCM’s liquid short positions were bid up while its relatively illiquid long positions experienced heavy selling pressure, precisely the opposite of what was required for the convergence trade. It turned out that many others, including the major banks’ trading desks, had been aping LTCM’s trades, so when the scramble for the exits began, there was very little space for a big player to manoeuvre.
Some of the LTCM partners wanted to raise more capital to add to their positions because they were adamant that spreads were unsustainably wide and they believed that the inevitable spread-tightening represented a rare opportunity for outsize gains. Unfortunately for them, the most likely source of fresh capital for the fund was the very group of banks that were issuing LTCM with margin calls; and these same banks would later be accused by LTCM of picking off their distressed positions. In order to meet the banks’ margin calls amidst an evaporation of liquidity, LTCM was forced to sell what it could at almost any price, which caused the fund’s equity to shrink, which in turn triggered fresh margin calls, and which ultimately put the fund into a death spiral.
It quickly became apparent that the LTCM implosion might have wider consequences. In addition to the on-balance-sheet liabilities now being about 250 times the fund’s dwindling equity, it also had derivative exposure with a face value of over $1 trillion, most of it in the form of interest-rate swaps entered into with almost every bank on Wall Street. If LTCM went under, each bank would face significant losses, which would set off a chain reaction of default that could result in systemic meltdown.
The Federal Reserve stepped in to safeguard the financial system and it orchestrated a $3.6 billion recapitalisation of LTCM with the firm’s major creditor banks. The banks and the Fed had reached a point where they were not prepared to tolerate any more pain, so they pulled the plug; the LTCM partners lost control of their fund even though they were willing to take more pain. The terms of the bailout were that the banks got 90% of the fund in exchange for their capital injection, and the LTCM partners got 10% as a way to retain them to oversee an orderly wind-down of the fund’s many arcane positions. The fund was finally liquidated and dissolved in 2000.
The partners’ share of the fund after the bailout was worth about $400 million, but this was entirely wiped out by the debts that the partners had incurred to leverage their personal investments into their already highly leveraged fund. Although they had an unusual tolerance for pain, they were not able to survive being wrong; they lost control of their fund, they lost vast amounts of money, and they suffered severe damage to their once-impeccable reputations.
Edge can be derived from any number of things, such as: depth of analysis; complexity of computation; or speed of execution. These are the things that you might typically associate with edge, but sometimes it’s more subtle than that. It had seemed from the outside that LTCM had some sort of magic sauce for delivering excess returns, but arguably their main competitive advantage was their ability to use their reputations and prestige to get highly preferential financing terms from their banks for their positions. The fact that most of their positions received no haircut from their prime brokers meant that they could simply use more leverage than their competitors. And perhaps the boffins at LTCM themselves failed to properly understand that this was maybe the firm’s real edge, and also how quickly it could disappear.
Edge can also be derived from the culture of the firm.
Example: Bridgewater Associates
The leaders of the firm, as well as the firm’s clients, say that Bridgewater’s culture is the single reason that it has become the biggest and most successful hedge fund in the world. Their way of doing things is unconventional and they have received extensive criticism in the press, but the firm says that the pursuit of radical truth through radical transparency is their key differentiator, because it allows them to operate as an idea-meritocracy.
Although they themselves do not say so, it seems to me that the culture of the firm is oriented around noise, directed at reducing the amount of noise in their firm and exploiting the amount of noise in other firms and in the markets. Noise arbitrage, so to speak. Almost always, you have to do something unconventional to get an edge.
But sometimes what goes for edge can be illegal. If you over-reach for edge, you can cross an important line and end up in jail.
Example: SAC Capital Advisors
SAC, run by Steven Cohen, was one of the best-performing hedge-fund firms in the world, averaging returns of 30% a year, net of their hefty fees (3% management fee and 50% performance fee) for the 15 or so years after its founding in 1992.
Those sorts of numbers are a compelling indication of the existence of an edge. However, the nature of some of that edge came to light after the firm and a number of its employees were charged by the SEC with insider trading. It later emerged that within the firm there was a colour-coding system for identifying the different types of information that would give SAC’s traders and fund managers an edge; inside information was known as black edge. SAC paid a big price for its ethical violations: the firm pleaded guilty to the SEC’s charges and settled with a $1.2 billion fine, and Cohen was barred from managing outside money for a number of years.
Even if you are fortunate enough to have developed a legal edge, you are not exempt from the need to innovate. The alpha that arises from your edge will attract new capital from your clients and your competitors, which will nullify your initial advantage. The more rigorous you are about identifying the nature of your edge, the more accurate you can be about quantifying its existence and extent, as well as whether it is under threat from cyclical or secular factors. Decaying alpha is your invitation to innovate, to renew your search for areas where you may have an edge, and to improve the design of your investment process.
The timing of a change to your process is important, both because of the optics involved and because of the underlying reality. If you change your process during the acute phase of a period of underperformance, that change will look like, and may well be, a function of poor investment discipline. If you are having a bad performance patch, you don’t want noise and its associated pain to cause you to abandon a well-designed investment process. But it’s really hard to know in the moment whether this is a transient patch of underperformance from a well-designed process or the onset of fatal underperformance from an obsolete process. One piece of data can help you decide which one it is: how much noise you are experiencing. If you are tempted to make changes to your process under conditions of elevated noise, it’s likely that this is a temptation to ill-discipline. If you are inclined to make changes to your process when the noise levels are low, it’s probable that this is a sensible impulse to improve the design of your process. Always, you need to be careful that the changes that you make to your process revolve around your competitive advantage.
Depth & Breadth
Another way to think about process design is as an attempt to optimise the intersection between depth and breadth.
Depth. Your skill in a given area, or how well you play.
Breadth. The number of independent bets in the portfolio, or how often you play.
This is sometimes referred to as the fundamental law of active fund management, which goes something like this: If you and your competitor are equally skilled, but she has access to more investment ideas or opportunities, she will deliver the better results. It seems to follow, then, that your process design should simply attempt to maximise your opportunity set. Keep depth constant, maximise breadth, et voila!
Except this is one of those ceteris paribus arguments where all things are not actually equal. There is a trade-off between depth and breadth, and the consequences of misunderstanding or misjudging the trade-off are not always self-evident and not always trivial.
Example: Steinhardt Partners
Much of Steinhardt’s success was attributable to the firm’s depth of expertise in US equities. As an illustration of the firm’s focus, Steinhardt had half-jokingly said that he would never invest in a place where he did not know the postal code.
The expansion of the firm’s existing capital base through excellent investment performance, combined with significant interest from new investors, made it one of the largest hedge funds in the world in the early 1990s, with around $5 billion under management. But the increase in assets under management began to nullify the firm’s depth of expertise in US equities, especially in relatively illiquid small- and mid-cap issues. They simply couldn’t deploy the larger sum of capital in the way that they had before, and the logical response was to create additional breadth by identifying new opportunity sets.
Steinhardt later said, “Having been successful in the markets that I had ventured into over time, I had confidence that the quality of my investment judgment was applicable worldwide. Perhaps rapid success had bred complacency.” The firm began to take positions in the large and seemingly liquid international bond and swap markets, and they also began to invest in emerging-market stocks. The firm’s swiftly expanding breadth is illustrated by the fact that they had positions in securities that six months previously Steinhardt had never even heard of, let alone knew the postal code, and his daily P&L report was now 30 pages long.
Recognising that his existing team lacked sufficient expertise in these new markets, Steinhardt rapidly expanded his staff from a handful of people located in New York to a headcount of over 100 people situated all around the globe. But, he said, “I was reliant on new people with whom I had never before been in the trenches. It was a recipe for disaster.”
The firm’s positions in non-US bonds were leveraged through the repo market to create exposure of $30 billion on their capital base of $5 billion. This degree of leverage meant that each single basis point move in bond yields had a $10 million impact on the P&L. These bond bets were enormously profitable and were responsible for most of the firm’s 60% return in 1993.
But in February 1994 the Fed surprised the markets by raising rates, which triggered a major sell-off in the global bond markets. Steinhardt Partners tried to reduce their position size but there was simply not enough liquidity for them to exit, and they were down 20% in a flash. Steinhardt said that he seriously misjudged how crowded their European bond trades were when he discovered that many of the banks’ trading desks were simultaneously trying to offload their long positions.
The firm ended the year down 30%, it’s worst year ever, in which they saw $1.5 billion go up in smoke. Steinhardt said, “But I would be wrong to characterise myself as simply a victim of a liquidity squeeze or an unexpected market turn. If I was a victim of anything, it was of hubris and unjustified confidence in my own abilities - perhaps the result of too many too-easy successes.”
This overconfidence had given rise to the belief that they could reach for breadth without compromising depth. But Steinhardt later said, “...I did not have a competitive or intellectual advantage.” To which he added, “I had lost sight of my own limitations. I was shocked and humbled by my failure.” He was candid about the firm’s attempt to quickly manufacture depth in new areas of the market by going on a hiring spree: “With our new and relatively inexperienced team, we simply were not on top of the game.”
Where before Steinhardt’s ego had been stroked by prospective investors who were eager to get a piece of the magic, now he was peppered with calls from journalists wanting to pick through the wreckage: “Answering reporters’ questions was unmitigated pain, not to mention a huge distraction.” The experience of 1994 was devastating for Steinhardt: “It had taken a part of me that could not be retrieved.” He fell into a state of depression: “I remember being so badly dispirited...that I could hardly function.”
He also revealed the risky existential proposition that many fund managers half-consciously adopt, where you equate who you are with your investment performance. He said, “I could not avoid feeling that my very worth as a human being depended on my continually making money. What was I worth when I lost?” Steinhardt limped on for another year before he closed down his firm and withdrew from the industry.
There are a number of lessons to be learned from Steinhardt’s experience, but the most germane to the subject of process-design is that to get the best results you want to optimise the intersection between depth and breadth, but there can be catastrophic consequences if you over-reach for breadth. When your investment process leads you to act as if you have an edge when you no longer have one (or never did), you will have your head handed to you.
References
Article in the Financial Times on 2 May 2018 by John Authers: ‘The question investors must ask: “Do I know more than a machine?”’
Steinhardt, M. (2001) No bull: my life in and out of markets. New York, New York: John Wiley & Sons, Inc.
Wilson, T.D. & Gilbert, D.T. (2003). ‘Affective Forecasting.’ Advances in Experimental Social Psychology. Volume 35.
Lowenstein, R. (2001) When genius failed: the rise and fall of Long-Term Capital Management. London: Fourth Estate.
Dunbar, N. (2000) Inventing money: the story of Long-Term Capital Management and the legends behind it. Chichester, West Sussex: John Wiley & Sons. Ltd.
Dalio, R. (2017). Principles. New York: Simon & Schuster.
Article in the Wall Street Journal, 22 December 2016, by Rob Copeland and Bradley Hope: ‘The world’s largest hedge fund is building an algorithmic model from its employees brains’.
Article in Bloomberg Markets, 10 August 2017, by Katherine Burton and Saijel Kishan: ‘Dalio’s quest to outlive himself’.
Kolhatkar, S. (2017). Black edge: inside information, dirty money, and the quest to bring down the most wanted man on Wall Street. London: WH Allen.
Grinold, R. (1989) ‘The fundamental law of active management’, The Journal of Portfolio Management, Spring edition.