Saturday, February 26, 2011

Puzzles, Clues and Poker

"What I like is solving puzzles. I think that what you are dealing with here is incomplete information. You've got little bits of things. You have facts. You have analysis. You have numbers. You have people's motivations. And you try to put this together, decode the puzzle in a way that allows you to have a way better than average opportunity to do well. That's the best part of the business. "

- David Einhorn, founder of Greenlight Capital

I love investing because it requires an open mind and a constantly evolving lattice work of ideas. What's fun is that frequently the ideas overlap with major concepts in other fields (economics, psychology, statistics, physics, law, poker, etc.) so incorporating one new way of looking at something can literally change how you approach questions / problems the rest of your life.

For example, my history of trades reflects how my mental models have developed over the past decade. One baby step at a time it has evolved from gambling on chart price fluctuations, to using liquidation as a pricing backstop, to low free cash flow multiples, to investing by valuing the intangible aspects of a business that can add exponentially to the enterprise value over time. My most recent development is about searching for "clues."

While searching for "clues" is definitely not something most investors will ever admit they're doing as it can create the impression of trivializing "serious" work, but the concept is actually commonplace in another field: Poker. In poker, players are constantly listening for "tells" or signals that may give away an event or behavior before anyone else realizes it.

Let's start with a straight-forward, first example on Netflix (NFLX). Investors have lost millions (perhaps billions) trying to short Netflix. I could argue that it's been for good reason as NFLX stock is certainly not cheap.

The day before NFLX released its 4Q'10 earnings it traded at ~$180 per share. This would equate to a $9.4bn valuation. To put this figure in comparison, their free cash generation in 2010 was $25mn (of course I'm benefiting from information that wasn't available at the time, but it's still effective for making my example for reasons you'll see shortly). If I exclude the cash outlays they made to improve their DVD and streaming libraries their free cash generation was $242mn. So just based on the math you can see why the stock is not cheap ($240mn divided by $9.4bn valuation equates to a 2.5% yield). In order to justify paying $180 for NFLX stock, you'd have to believe NFLX will grow its business exponentially from current levels (currently has 20mn subscribers vs. ~100mn U.S. households). So on a high-level this covers the fundamentals. What about the clues?

Given this lofty valuation, one hedge fund investor even went so far as to published a letter titled "Why We're Short Netflix." Shortly afterward, Reed Hastings, the CEO of NFLX, published a response piece called "Cover Your Short Now." Now, this is a monster of a clue for several reasons. First, if the CEO is telling you to not bet against their stock weeks before they report 4Q'10 results, you may be right in the long-term, but in the short-term the CEO knows more than you do. Second, this means the CEO is watching the stock and any bad news associated with it, so he's going to do everything in his power to manipulate it higher. Whats this all mean? Walk away. The clue is that Mr. Hastings wouldn't put out a positive letter weeks before earnings if the results wouldn't be amazing and impress Wall Street.

Later in January NFLX ultimately did report knock-out 4Q'10 results and the stock quickly shot up to $240 (30%+ higher). While its now even more expensive, a lot of short-sellers could have avoided the near-term pain had they combined their fundamental analysis with interpreting a very obvious clue. Hindsight is 20/20.

If you keep an open mind, you start to realize that there are clues all over the place. As another example, look at Southwest Airlines (LUV). Now this is a much more subtle clue with a less significant price fluctuation (and less extreme valuation divergence) but it's still worth mentioning.

On 1/20/11, LUV reported 4Q'10 results. During the call management talked about their expectations for the month of January:

"Our January revenue and booking trends thus far suggest a year-over-year improvement in our January PRASM [a proxy for pricing] similar to or possibly slightly better than the 5% year-over-year improvement in December 2010 PRASM."

But then, only a few weeks later on 2/3/11 management provided the following comments at an investor conference:

"Our January 2011 revenue in Traffic will be reported early next week. But at this point in time, we are estimating that the year-over-year increase for January will exceed the 5% year-over-year increase that we reported in December."

Do you notice the difference? If not, I would suggest you reread each of the quotes.

What you should see is that the tone changed. January went from should "possibly" be "slightly better" than 5% PRASM growth to "will exceed the 5%." It's a subtle difference, but if you were a shrewd trader you would have bought the stock just by observing the difference. What's even more telling is how the stock dropped 12% from 1/5/11 to 2/3/11, the same day as the investor conference. In the week following the conference, LUV stock rallied 7% and on 2/7/11, management reported ~8-9% PRASM growth for January.

Learning this clue-based model has reinvigorated my joy of investing because I know the clues are out there, I just have to be receptive to them. That said, investing still requires a lattice work combining short-term data points with long-term valuation techniques. Just b/c you successfully interpret a clue (like that NFLX is going to report a good quarter or LUV is going to report a solid January figure) doesn't mean you should make the investment. The business can still be overvalued (ala NFLX) or susceptible to external shocks (LUV subsequently traded back down b/c of the current oil spike) so you must weigh the separate variables accordingly.

No comments: