Image Credit: Liu Bolin
“The problems are solved, not by giving new information, but by arranging what we have known since long.” – Ludwig Wittgenstein
It turns out, there is a ton valuable data lying around, hiding in plain sight – just waiting for any fool to come along and turn it into something else, something greater. And, it also turns out that I am one such fool who is captivated – perhaps even obsessed – with solving puzzles (like this 8,000-piece beauty purchased – unassembled – at the Louvre, on the right); turning what usually appears as scraps of data into something much clearer, much bigger, and much more valuable – like a piece of art.
So, you might imagine my excited anticipation to finally work on unlocking some of the value contained in 13F holdings reports. (I know, right?! Who isn’t?) Here’s the gist: Basically, all institutional investors – whether they be public or private entities – with discretion over $100 million or more of US equity assets under management must file a quarterly, position-level report to the SEC.
Now, I am certainly not the first fool to dig into a 13F report or two. For instance, there are websites, like Whale Wisdom or Holdings Channel, that specialize in tracking, assembling and picking some of the obvious nuggets out of the latest 13F filings. However, like most everyone else these days, these services tend to focus on easy-to-harvest, quickly-decaying fragments of limited insight – such as, the top holdings of famous managers, like Warren Buffet, George Soros or Bill Ackman – and which also tend to be geared towards a more retail-oriented audience. Lots of interesting lists, but missing the bigger (strategic) picture – where Alphacution usually looks to plant its flag.
Of course, this type of limited curation can be useful for lots of reasons, but I can’t imagine why it would be all that useful to an institutional audience beyond novelty. For this data to be useful to an institutional audience, we first need to climb to a higher vantage point – and then apply some unique perspective. With that perspective, the nitty-gritty takes on a whole new meaning…
So, with that in mind, what follows is something we have never seen presented before: A “picture” of the entire 13F report for Goldman Sachs Group, Inc. – its most recent, from September 30, 2018 (which was filed at 14:02:40 on November 14, 2018 – a seemingly innocuous tidbit that will become more important down the road).
Now, this chart may seem a little odd at first glance. Of course, this is by design. I want you to stretch your perspective on this by first emphasizing the shape of the portfolio before we get into the weeds around the numbers, such as the total value of this portfolio, the position count, and other fascinating insights that can be harvested from this dataset when viewed by starting with a very wide lens. And, by the way, if we had not used a log scale on the vertical axis, the shape would have been far less illuminating (see chart on right) – except to imply a level of value concentration in the top few holdings.
Note also that this particular filing contains the point-in-time positions for the brokerage division (Goldman Sachs & Co., LLC), the asset management division (Goldman Sachs Asset Management, LP), and the market making division (Goldman Sachs Financial Markets, LP) among a few other sub-entities aggregated into the group 13F filings – the positions for all of which can be distilled from the broader filing.
Furthermore, and perhaps more importantly, consider that this same type of exploration can be extended and performed for other secretive and mythological managers, like:
- D E Shaw
- Two Sigma
- Susquehanna (SIG)
- and many, many others…
Besides being super cool for students of the game, why is this analysis so important? Well, to understand the value of this data at even slightly deeper levels – and share more pictures, too – we need to talk rather unpredictably about maple syrup first.
The process behind this type of analysis – like most everything else we do here at Alphacution – is similar to the process of making maple syrup. If you have ever been to any number of sugar shacks in places like New Hampshire, then you already know a bit of what I am talking about: Maple syrup is distilled sap from maple trees. Raw sap looks like water; clear, with viscosity closer to that of water, and devoid of the taste you would very much prefer to slather on your pancakes and waffles. The ratio of raw sap to maple syrup is usually 40:1. In other words, it can take 40 gallons of sap to produce 1 gallon of maple syrup.
Now, take this thought and turn it to the data. There is a lot of “signal” still embedded in the “noise” of the data that is lying around us; hiding in plain sight. We just need to find the patience, tenacity and experience to distill the “signal prototypes” – and then assemble the apparatus – to harvest this intelligence at greater scale and frequency.
In this case, we are opening a new can of delicious worms related to the data contained within the SEC’s Form 13F filings, and then beginning to add that to our existing dataset and ongoing analyses. Of course, for those of you who are likely to know anything about form 13F, let’s pause briefly to indulge a brief refresher to get everyone on the same page (or, skip down a paragraph or two).
Wikipedia offers as good an overview definition as any for our purposes now:
Form 13F is a quarterly report filed, per United States Securities and Exchange Commission regulations, by “institutional investment managers” to the SEC and containing all position-level equity assets under management of at least $100 million in value with relevant long US holdings. All US-listed equity securities (including ETFs, equity options, and convertible bonds) in the manager’s portfolio are included and detailed according to the number of shares, the ticker, the issuer name, etc.
The full list of securities currently covered by form 13F includes more than 17,500 securities. Short positions are not required to be disclosed and are not reported. Form 13F covers institutional investment managers, which include Registered Investment Advisers (RIAs), banks, insurance companies, hedge funds, trust companies, pension funds, mutual funds, among natural persons or entities with investment discretion over its own account or another’s. Form 13F is required to be filed within 45 days of the end of a calendar quarter.
From here – and still staying at a high vantage point – there are two important directions we can go: “Absolute-relative” and “relative-relative,” where absolute-relative means comparisons with the entity’s historical reporting and relative-relative means comparisons with the entity’s peer reporting.
Let’s start with the second of these (relative-relative) first: What does the chart from above look like when we add Citadel, Millennium, AQR, D. E. Shaw, Two Sigma, SIG, and Virtu (for the same time period) – and then add the axis numbers? Can you guess which manager belongs to which curve?
A FAR more fascinating picture, no? With minimal training of the eye we can certainly see the diversity of “shapes,” which is influenced most by position count, portfolio value, and concentration. A slightly more trained eye might detect a sensitivity to portfolio dollar neutrality (i.e. – “flatness”), albeit with more risk of false positives given what we are able to see and show here. Add the manager names and what we might know about the underlying strategies and business models, and all of a sudden what we thought we might never know about each of these players is, at least, partially illuminated. (When we dig down a layer to the level of option positions, for instance – even though we can’t see short option positions – that “illumination” becomes quite a bit brighter – as we will detail in an upcoming post or two.
Ok, so how does the absolute-relative perspective work? Here, we have taken our analysis of a time-series of Goldman’s 13F reports and aggregated the top level attributes for the 42 quarterly reports beginning June 30, 2008 (see below).
Among the more fascinating clues from this illustration is the lack of variance in the position count given the growth in portfolio value. Over a 10.5-year period, the max portfolio value change is 257% while the standard deviation of the position count is 5.5%. Clearly, we are not even scratching the surface of what we might find and learn from a broad assembly of this data…
Now, when we widen our lens even further, we might discover at least one reason why it appears there are guard rails on Goldman’s position count: In the chart below, we present the same analysis template as above for the 52 quarters beginning March 31, 2005.
Notice anything different? A rare picture of what frozen markets look like…
The Journey to the Capacity of Alpha
As a hint of foreshadowing for where we go from here, position-level data from these and many other asset managers – even incomplete and periodic position-level data – creates a new zone of modeling between market micro-structure analysis (i.e. – the stuff of liquidity formation, liquidity fragmentation, and product selection) and market macro-structure analysis (i.e. – the stuff of players, their business models and their trade strategies). In short, this data is the beginning of our journey to quantify our hypothesis that the capacity of alpha is finite, and therefore, more accurately define the lines between the structural, active and passive management zones of our global asset management map, the background on which can be found in our recent post, The Privatization of Alpha.
Together, these and other datasets will come together to strengthen Alphacution’s framework for how the global trading and asset management ecosystem fits together – and shifts along with the changing landscape of fundamental drivers. With our internal plans for workflow automation, increasing scale and frequency of signal harvesting will yield increasingly detailed, accurate and timely “maps” of this ecosystem.
It will work kinda like a 3D printer – although perhaps more holographically. Alphacution’s unique dataset, scaled and sped up, to perform like a datafeed; delivering uncommon intelligence, as the original vision for this project presented itself…
Looking forward to sharing more.
As always, if you value this work: Like it, share it, comment on it – or discuss amongst your colleagues – and then send us firstname.lastname@example.org.
As our “feedback loop” becomes more vibrant – given input from clients and other members of our network, especially around new questions to be answered – the value of this work will accelerate.
Don’t be shy…