Below is an excerpt from our 2018 annual letter. Unfortunately, I can't share the whole letter due to #compliance, but you can contact my colleague Rob (his email is [email protected]) if you're interested in seeing the whole letter or learning more about us.
Looking Forward to 2019 and Beyond
Zooming out a bit from our 2018 results, I want to spend some time talking about how we identify undervalued opportunities in today’s markets, and then expand on a few specific areas that we think are particularly attractive. But before we do that, allow me to veer off on a quick tangent.
A few months ago, computers beat gamers at the popular computer game “Dota 2”. Maybe that computer victory seems small versus the groundbreaking “Deep Blue” chess win in the 90s, but Bill Gates called the Dota victory a “huge milestone” because the Dota win represents a completely different type of accomplishment for computers. In fact, in many ways, comparing the computer victory in Chess to the recent victory in Dota is completely apples to oranges. In chess, there is only one opponent, each player has perfect information (i.e. they know exactly where all the pieces on the board are and what all of the possibilities are at each second), and each player plays in turn with no interruption (white goes, black goes, white goes, and so on until the game is over. Black can’t suddenly move twice or slap white’s hand away when white tries to move). In contrast, Dota is a team game that happens in real time with imperfect information. In order to win, players on a team need to cooperate with each other while responding to the opposing team in real time, so there is no perfect strategy or move that always works. The game is “hugely complex”, with two teams of five facing off on a giant map filled with obstacles and destructible buildings. There are pieces of the map that aren’t visible, so knowledge is limited, and there are more than 100 heroes and hundreds of items and skills that can boost abilities and change the tide of battle*.
(*I understand not all readers are familiar with video games, so here’s another way to think of it: playing a game of DOTA is like calling plays in a game of (American) football. You’ll need to make different calls depending on the time of the game, the game situation, and the strengths and skills of the players on the other team. However, in football, you have a basic idea of what the field looks like and where the other team’s players will line up, while in DOTA the field can change constantly and the other team can hide their players and “tackle” you when you least expect it. Also, the other team might summon a dragon to eat your quarterback at halftime.)
Computers beating humans at Dota is a fun story, but at this point you’re probably wondering what the Dota story has to do with investing and our portfolio? Simple: in many ways, I see similarities between the evolution of computers playing games like chess and Dota and computers competing in the markets. I think learning lessons from the gaming side helps inform us of where AI and machine learning are effective, where they fail, and where we can find alpha in the markets going forward as computer trading and quantitative models become an increasingly dominant portion of the market.
If we rewind to the 1960s/1970s, stock markets were inarguably much more inefficient than they were today. Consider:
Over the past few decades, all of those inefficiencies have been (mostly) competed away. Regulations and computers ensured that information would diffuse evenly. Investors overcame their fear of high-yield bonds, and today incentivized investors will buy anything that offers enough yield for its risk (CDO squared, anyone?). Black-Scholes was invented, and reams of quant funds spent decades capturing huge alpha by shorting expensive stocks and buying cheap ones. While there may still be some alpha to those strategies (for example, every now and then a piece of corporate news will leak slowly because it’s posted in the wrong place / at the wrong time, creating alpha for traders who see the data before anyone else, and it seems there’s still some alpha that exists from investing in “value” factors; it’s just much smaller than before), but in general the alpha to those strategies is gone and not coming back.
In many ways, I view the death of that alpha as the “computer winning at chess” stage of investing. Black-Scholes and CAPM value / size factors are simply mathematical formulas. Thirty years ago, a human could make money by applying those formulas (or things similar to them) on their own. Today, computer “quants” can apply those factors (plus thousands of others) at blinding speed and with ruthless efficiency. If you’re a human today investing on simple quantitative factors (stock Y trades at 10x earnings and stock X trades at 12x earnings, so I’ll buy stock X and short stock Y based purely on that valuation), you’re effectively a human trying to play chess against a computer: overmatched and destined to lose.
The alpha from those “chess” strategies was large, lucrative, and relatively stable / easy to achieve. If that alpha’s gone, what’s left for active investors trying to outperform?
I think you can see pieces of the answers by looking at how humans have beaten computers at games. For example, in the computer DOTA victory, it’s worth noting that the computers won two of their first three Dota games against the humans, but then lost a third “just for fun” match when the computers weren’t able to choose optimal conditions / get the specific heroes they wanted (in the words of one review, the “humans proved more adaptable” than the computers). In other words, the AI’s knowledge proves “very brittle” and as soon as the computer encounters something unexpected the AI breaks down. We’ve actually seen things like this for years; in the famous Kasparov vs Deep Blue chess matches, Kasparov would intentionally make ridiculous openings (“don’t ask me what to call that opening”!) to force the computer out of its prepared opening sets.
So what does that mean for investors? Increasingly, I think alpha will accrue to investors who can invest in unique situations that can’t be captured properly by computer models. An example may illustrate this point best: over the summer, Elon Musk tweeted “considering taking Tesla private at $420. Funding secured” (disclosure: we have a small short position in Tesla). This tweet eventually turned into quite the saga: for a while, some thought the best source of information on the buyout was pop star Azealia Banks’ Instagram account (where she claimed Musk sent the tweet while on Acid, and, yes, I am serious that this is a thing that happened), the SEC sued him and revealed that the $420 price was a marijuana reference (rather obvious to most people plugged into pop culture) designed to impress his girlfriend (less obvious), and the tweet caused countless lawyers and journalists to spend time wondering how Tesla could go private while keeping its public shareholders who wanted to remain shareholders (as Musk promised they could).Was the whole saga “highly” (pun very much intended) amusing? Absolutely! But it also represented a unique situation that broke the bounds of normal market functioning and that computers were uniquely ill-equipped to handle. After seeing Musk’s tweet, computers would immediately turn Tesla into a deal stock, and their math looks something like this, “he’s trying to take the company private at $420. Tesla was trading for $340 before the tweet, and there’s a 50% chance the deal goes through, so Tesla should be valued at ~$380/share.” Humans, however, could quickly draw a lot from the context clues around the offer (isn’t $420 a marijuana reference? How many times has a go-private offer been launched on Twitter during market hours, never mind that this would be one of the largest in history?) that computers might not be able to pick up on. In addition, humans might be able to think down the event path a little further than a computer running a quantitative model could (for example, if the human investor had read Ashlee Vance’s Musk biography, they might remember the story about Musk taking “risks that seemed like they could land him in jail” and bluffing about funding when Tesla had faced distress a few years earlier. Remembering that story, the human investor might think that this take private might be a similar “bluff” to cover up some other issues at the company, or the human investor might be able to start thinking further down the line about the potential for fines / sanctions / disbarment if Musk’s tweet wasn’t factual).
Obviously the Tesla saga is a unique and over the top example, but it serves as a really nice / extreme example of the types of things where I think investors will be able to generate value in going forward: investing in things that valuation / computer models can’t or won’t invest in, including things that are too complex for a computer to model / value (for example, holding companies with complex accounting, or companies that announce a transformative merger so that historical financials have little to no bearing on future valuation), securities too illiquid or small for computers to buy, companies where value can be created through something a computer can’t do (for example, running an activist process), or through quirky situations that involve human tendencies or histories that computers can’t pick up on (like the Tesla situation above). In my head, I consider these “Dota” opportunities: they’re opportunities that exist because of uncertainty or tail events that quantitative computer models can’t pick up on (or can’t invest in).
In a few years, I would guess computers are able to reliably beat humans at any video game, but the good news for investors is that, as long as corporate executives can tweet fake buyout offers / marijuana jokes to impress their girlfriends, I have a feeling that there will be some tail events that computers can never capture and some alpha that investors willing to be patient will be able to earn. Going forward, I think alpha will be smaller (as most of the easy / less risky “chess” profits have been picked away) and lumpier (with the “chess” profits picked away, a lot of the remaining profits will come from investing in things computers undervalue or can’t buy and waiting for a corporate transaction to unlock that value) than it was a decade or three ago, but I think alpha will still be there for people willing to do the work to find quirky / off-the-run / complex situations.
Let’s turn back to our portfolio and talk about how we’re invested and why I think we’re positioned to capture that alpha going forward.
The vast majority of our portfolio currently falls into one of four “buckets”: tracking stock and holding companies, publicly traded sports teams, publicly traded alternative investment managers, and special situations. We believe we have a significant edge in analyzing each of these buckets, and we believe each of them fits into the “Dota” framework we laid out above. So, what are these buckets and why do we think we have some type of edge in investing in them?
We also believe the major private equity firms have significant tail winds that will allow them to grow at above average rates for years to come, in part driven by a “nobody gets fired for buying IBM” effect at large pension / sovereign wealth funds. If you’re a pension fund employee in charge of private equity allocation, you have two choices: you can give the allocation to an unknown startup, in which case you’ll get fired if it does poorly and a pat on the back if it does well. Or you can give the allocation to one of the ~seven major private equity players, in which case you’ll get a pat on the back if it does well and a nod of understanding if it does poorly because every other pension fund in the world will have invested in the same fund and your results will look like everyone else’s. As we prepared to publish this letter, the Wall Street Journal published an article on Blackstone’s record $20B real estate fund that captured this line of thinking perfectly, “Investment officers working for pension funds, endowments and other institutional investors prefer firms with marquee names and long track records because they have less explaining to do if things go wrong.”
Before I wrap this letter up, I want to spend time on two different points: where does our largest position, Charter, fit into the framework above, and what’s our outlook for 2019?
Let’s start with the first question: I just described how we were looking for areas of opportunity where we see value that computers don’t, so how does Charter, a mega-cap cable company, fit into that framework? Glad you asked! I think it fits in the following ways:
In sum, we think Charter fits well into our “invest in complexity to beat the computers” thesis, and we expect the company to perform well in 2019/2020 as they put their integration behind them and as the mobile business starts to ramp up.
What’s our outlook for 2019?
We’re micro-investors, not macro-investors, which means we generally focus on identifying and understanding mispriced securities on a one-off basis. So rather than say “rates are increasing so we should be long consumer staples and short utilities,” most of our time is spent looking for companies and situations that we think are misunderstood / mispriced. We believe it’s a great time to be a “micro-focused” investor, as the disaster that was equity markets in the fourth quarter of 2018 (as mentioned above, one of the worst quarters in the past ~50 years for public equities) has left a variety of companies and industries severely oversold / mispriced. For example, in Q4 we saw a wide variety of closed-end funds (publicly traded investment companies that generally consist of a basket of liquid equities, similar to a mutual funds) that traded at discounts to their underlying asset values that hadn’t been approached since the financial crisis, and plenty of financials (banks, lenders, etc.) traded far below their tangible book values (the estimate of what shareholders would receive if the company were to liquidate tomorrow) despite strong fundamentals, good management teams, and rational share allocation (i.e. aggressive share repurchases when shares trade below book value). While market declines like the fourth quarter of 2018 can be frightening, they create fantastic opportunities to buy good companies at a discount or interesting special situations at a great risk/reward. Currently, we’re finding more opportunities to invest than we have since launch, and we’re confident / hopeful that those opportunities will produce strong results in 2019.
Moving away from our micro focus, let’s take a second to discuss our overall outlook for the market in 2019 (as that’s generally what people mean when they ask what’s in store for 2019). I generally stand by everything I wrote in last year’s letter: overall markets remain relatively fully but fairly valued, though strong earnings in 2018 combined with the market declines in 2018 mean the overall markets are trading a touch cheaper than they were at this time last year. As I write this, the S&P 500 is trading at ~15x forward earnings (~7.5% earnings yield), its cheapest level since the 2012/2013 timeframe. With interest rates remaining at historic lows (10-year Treasury bonds continue to yield less than 3%), the stock market continues to trade much too cheaply however you look at the market (you can use forward earnings or a cyclically adjusted earnings measure like the Shiller PE; either way the market trades for a solid discount to bonds despite the fact stocks should generally grow their earnings over time while bond yields are fixed).
At current prices, I think markets are probably priced cheaply enough to power through a garden variety recession (and, given our current focus on non-cyclical stocks like cable companies and special situations, our portfolio would likely do significantly better than “power through” in a recession). My biggest worry is that markets tend to “crack” when a core assumption of the market proves false. For example, heading into 2008, everyone assumed that housing prices could never go down, and much of the financial system was built on that belief. When that belief proved wrong, we got our worst recession since the Great Depression. A core assumption for financial markets for the past ~100 years has been the stability of the U.S.: The U.S. honored international deals, supported its allies, had a (mostly) functioning government with (relatively) stable rules, and the Fed was generally independent of politics. My big worry for 2019 is there are a wide array of events that could quickly test any of those assumptions, and if any prove to be wrong then we will have a crisis that will take years to resolve (i.e. if the Fed is proven not to be independent of politics a crisis can quickly develop, and once an assumption is invalidated, there’s unfortunately going back. Warren Buffett often says it takes decades to build a reputation but it can be lost in a second; similarly, those types of core market assumptions take decades to form and can be undone in a second if institutions aren’t careful).
I don’t want to end on a sour note, so let me reiterate that any of those assumptions being tested remains a tail risk, and my hope is that if they were to be tested, the assumptions would prove correct / our institutions would hold up (for example, to date the Fed has seemed to largely ignore any attempted influence from the President).
No matter what the market throws our way, I’m thrilled with the way our portfolio is positioned today. We’re currently finding more opportunities than we ever have, and we believe that 2019 should have several positive catalysts for many of our large investments. We’re looking forward to a strong 2019.