A Dollar and a Dream

Why P&L doesn't always tell the whole story

Everyone still loves a winner. Even in these dark times on the Street, as careers grind to a halt, deferred compensation plans wither on the vine, and politicians and editorial writers rail against bonuses, Wall Street remains proud of its strict merit pay model. We are rightfully disdainful when we read of how the teacher unions fight merit pay, terrified of being

individually judged by objective criteria. Wall Street, on the other hand, has long thought of itself as the polar opposite. Sought out by the ambitious as the ultimate meritocracy, it is a place where a rock star on the trading floor can make 10 or 100 times what the next guy gets. Firms brag about their “eat what you kill” culture: Make money and get paid a lot, lose money and get shot.

The Fittest

Traders choose their counterparties with a similar ruthlessness. Algorithms and crossing systems that produce regular good results get richly rewarded with flow, while weak algorithms are shunned like the money-losing traders, with volume gradually drying up as the business unit inevitably atrophies.

And mostly this system of “survival of the fittest” works, until you confront something known as the “reverse lottery” problem, and realize that there is a hole in the logic of paying people and allocating flow based strictly on results you can see. The hidden flaw is that in probability games, observed outcomes often obscure an ugly reality.

For example, let’s say I walk into my friendly local casino. I wander over to the blackjack table and begin playing, betting $50 at a time. I have a very simple strategy: As soon as I am up $100, I will immediately stop playing, get up from my seat, and walk out of the casino for the day. But if I am losing, I will doggedly continue playing, digging myself into a deeper and deeper hole until I either rally back and make my $100, or until I bankrupt myself by blowing my full $100,000 credit line.

The asymmetrical payouts in this game make it far more likely that I’ll be a winner. Win two quick hands and I’ve made my money for the day. But to lose, I need to lose 2,000 more hands than I win. So while my expected profit for the long run is exactly zero, on any given day, I have a 99.9 percent chance of walking out up $100 (making a few simplifying assumptions). If I played every workday, there is a 78 percent chance of wrapping up the year having made $100 each day for all 250 business days, producing $25,000 in profits. And since most days I will have barely dipped into my credit line, my likely return on average capital would be well north of 100 percent.

Reverse Lottery

To an observer that just looked strictly at my empirical results, and didn’t understand the huge risk lurking in the shadows, I would look like the ultimate producer. They would have seen day after day after day of $100 wins. My returns would be so high, and so incredibly consistent, people might even suspect I was pulling a Madoff. But there is no fraud in the above example-all I’ve done is create a lottery in reverse. People who play the lottery accept almost certain failure each week in return for a very small chance of extreme success. A reverse lottery, like the above blackjack example, is a trading strategy that offers almost certain success, in return for accepting a small risk of massive failure.

When you’re playing with your own money, the traditional lottery is the smart play: The steady loss of a few dollars has a small effect on your net worth, whereas the unlikely giant gain could get you a ski house in Aspen. But Wall Street thrives on the reverse lottery: The small but steady gain accrues over time, eventually getting you the house in Aspen, which is then yours to keep when the giant loss finally rears its ugly head, leading your firm to take a one-time charge due to “extraordinary market conditions.”

Trading algorithms can also be set up as reverse lotteries, with perceived steady success really the result of an asymmetrical payout structure. Algorithms that are set up to “lock in” a win whenever they are ahead of their benchmark, while being patient when they fall behind, can exhibit aspects of a reverse lottery. These algorithms will beat the benchmark by a small amount on the majority of trades, only to fail spectacularly on a few outliers.

Clients that tend to evaluate performance over short-term periods are at risk of falling into an asymmetric payout trap. To climb out of there, Wall Street needs to break some ingrained habits. Outliers should be included in performance analyses, unless there is a clear and obvious problem with the data point. Managers need to understand the underlying process that creates the outcomes they see, and not just look at the outcomes themselves. And most importantly, both traders and algorithms need to be evaluated over sufficiently long-term periods.

Land Mines

As Wall Street looks less like the street of dreams and more like the boulevard of broken dreams, playing the lottery grows more attractive to the Street’s denizens. Which means that it’s just a matter of time before a clever trader notices his colleagues throwing their cash away day after day, and then creates a leveraged swap instrument that re-creates the lottery’s payout structure in return for a lower premium than the state charges. After his colleagues probably don’t hit the big number, and he finishes the year with a nice P&L, here’s to hoping his manager won’t pay him based on his seemingly impressive results. Because when you’re running a reverse lottery, the game is not about a dollar and a dream. It’s about a dollar and a nightmare.

Dan Mathisson, a Managing Director and the Head of Advanced Execution Services (AES) at Credit Suisse, is a contributing writer to Traders Magazine. The opinions expressed in this column are his own, and do not necessarily represent the opinions of the Credit Suisse Group.

(c) 2009 Traders Magazine and SourceMedia, Inc. All Rights Reserved.

http://www.tradersmagazine.com http://www.sourcemedia.com/