The Only Truth in Markets is Surprise

  • Posted by
  • on April 26th, 2014

“Blessed is he who expects nothing, for he shall never be disappointed.” Alexander Pope

That phrase has stuck since I was a kid, written under a puppy in a framed picture in my grandparents’ foyer…who knew they were preparing me for a life in the markets? Fast forward to the present, where Brian Portnoy shared it as part of a wonderful section on expectations management; his book The Investor’s Paradox is an instant classic on the inputs to consider in choosing fund managers.

I love his use of the term expectations management, and think it deserves a spot right next to risk management and money management in the minds of market participants. To me, it’s a better lens than behavioral finance or investor psychology; it conveys a mindset towards solutions vs. just describing the problems and biases we all face. By acknowledging our blind spots and embracing the ambiguous nature of markets, we have a chance to create our own market experience regardless of the curve balls regularly thrown at us.

Soon-to-be Hall of Fame pitcher Greg Maddux had a theory that no hitter could tell the speed of a pitch with meaningful accuracy; change the words to investor, direction, and market and you immediately skip the naive stage on the learning curve. So if markets are unpredictable, yet probabilistic, should we then rely on those statistics that give us the highest chance of winning? On a long enough timeframe, buy and hold has a 100% success rate…go with that?

We’re all on our own here, but for my wiring it’s an unequivocal NO. Probabilities serve two roles, one potentially positive and one potentially negative. To the good, win rate is an input to expectancy; to the bad, it’s an input to expectations. HUGE difference. The former sets us up to be realistic, the latter sets us up to be disappointed. The former allows us to act uniquely, the latter puts us in with the herd. To me, the misuse of probabilities and expectation to be “right” is a toxic mix to be avoided at all costs.

The problem with market probabilities is context; as a complex adaptive system it is impossible to fully set the context of historical data. I don’t care if something happened 29 of the last 30 years; that sample size just doesn’t cut it for defining a true edge. All it does is prompt other participants to position themselves for 30 of 31, morphing the true underlying condition further by failing to account for all of that anticipatory positioning.

Go back to sports for a minute. Yes, it’s true that an 8-2 NFL team is highly likely to beat a 2-8 NFL team repeatedly. But we’re not able to make that bet at the price we want, are we? Let’s say our incredible vision gives the favorite an 80% chance of winning. In creating a (mostly) efficient market, Vegas forces us to bet $400 for every $100 we win. In expectancy terms, that makes:

80% Win Rate * $100 Profit

                                                ————————————            =     Expectancy of 1.0

20% Loss Rate * $400 Loss

Not very compelling, right? Market speculation may not offer the excitement of live sports, but it offers endless opportunities to prosper in a boring way. So how do we differentiate between useful probabilities and not-so-useful ones? How about sample size?

No matter how we torture the data, it is nearly impossible to simultaneously a) define the entire context of a market condition and b) acquire a sample size large enough to accept the result with confidence. There are exceptions that do provide useful context, including constructs like this that provide an estimate of likely range, based on using 252 data points per year for many years and using the entire distribution. Another method is applying probabilities across a universe of stocks to gain a useful sample, vs. simply applying it to an index like $SPX. A third way is to isolate noncorrelated tendencies in creating a summed probability that acts as a deeper context.

Besides demanding a large sample, my demands on a decision input have me managing the expectancy vs. expectations divide as follows:

1) Asymmetric Payoff Potential- 90% historical win rates don’t get me excited, but a large sample with 50% wins yet high expectancy draws my attention. The fallacy of a 90% “hidden” edge is so laughable it likely carries both high expectations and low expectancy through its other hidden component…its yet to be seen black swan blowup.

2) Quick Feedback- I actually don’t care as much about the success rate going in; what’s more important to me is a process that will alert me to failure as soon as possible. Back to Brian Portnoy’s book, he contrasted strong vs. weak feedback using examples such as a musician(where a failed note stands out) vs. a parent(could take years to decide whether we made the right decision). I use clearly defined breadth data for a reason; it clearly defines when I’m wrong. When can I make that determination with a “in the past 18 mid-term election years, $SPX has returned X” type of stat? June 30? August 15? It’s a narrative, just a binary bet with no opportunity to act on new information…save those for sports or not at all.

3) Logic- going beyond the standard correlation/causation question, does this source of edge make sense? Is there a behavioral or structural reason why this source of edge should persist? Be it odd-lot stats, the 3% dividend yardstick for valuation, 3 Fed hikes and a stumble, etc., a successful history is just that, history. We should demand more than that. Besides giving me the asymmetry and quick feedback I seek, I am very comfortable that there will always be a role for market breadth analysis. The inputs and lookbacks may change with the market climate, but the underlying supply and demand environment as measured by the universe of stocks is, to me, a bedrock principle of the appetite for risk. Without that appetite, we have no one to accept our offer to sell.

Trust me, I love statistics and probabilities, and it’s simultaneously fun/scary to now hear it in my soon-to-be 9 year old son. I learned the hard way that, to paraphrase Donald Rumsfeld, there are known probabilities(expensive because everybody knows), unknown probabilities(umm, unknown), and the unknown unknowns that masquerade as “expert” forecasts. As Richard Thaler perfectly described in this ode to Daniel Kahneman, “As a result(of the representativeness heuristic), we can expect forecasters to be predictably surprised when they draw on small samples.” And we, as market participants, should expect to be surprised and build that into our trade plan upfront.

From Buffett to Soros to Tudor, asymmetric payoffs are the mother’s milk of alpha creation. We may differ on whether value or momentum are the best paths to get there, but without the opportunity to make more on our winners than we lose on our losers, we have no shot at sustainable success. The combo of high expectations and low expectancy is a guaranteed destroyer of capital, with misuse of probabilities among the leading causes. To quote Charles Wheelan, “Statistics cannot be smarter than the people who use them. And in some cases, they can make smart people do dumb things.” Remember that if you decide to put it all on “Sell in May and go away” because it worked a few times.

The information in this blog post represents my own opinions and does not contain a recommendation for any particular security or investment. I or my affiliates may hold positions or other interests in securities mentioned in the Blog, please see my Disclaimer page for my full disclaimer.

blog comments powered by Disqus
Derek Hernquest Blog