The Probability Problem

By Robert Seawright, Proprietor, Above the Market

Investing is a probabilistic enterprise. Since certainty is even rarer than high risk-free returns, we’re left trying to make the best decisions we can based upon the knowledge we have. If we do that extremely well, we might be right most of the time, but still a long ways away from all of the time. The improbable — the highly unlikely even – happens and happens surprisingly often.

Take yesterday’s NFL action, for example. More specifically, consider the astonishing Vikings v. Ravens game in snowy Baltimore.

The Ravens and Vikings set several records yesterday, but perhaps the best of them was that the lead changed hands six times in the 4th quarter, That had never happened before. There were six touchdowns in the fourth quarter alone, with five of them – and four of them on plays exceeding 40 yards — coming during the final 2:05 of the game. The teams combined to score 36 points over that final 2:05, by far the most combined points late in any game over the last 50 seasons, per Elias Sports. In fact, before yesterday’s madness, the fastest five touchdowns (anytime) in one game over the past 50 years required more than 5:30. It’s almost incidental to note that Baltimore won the contest, 29-26.

The good folks at Advanced NFL Stats run a win probability metric throughout games to show the likelihood of each team’s winning after each play based upon score, down, distance, and time using an enormous data-set of NFL history. This analysis provides a good way to examine, monitor and evaluate the decision-making of NFL teams and coaches.

Take note of the win probability chart for Vikings v. Ravens yesterday. It’s wild.

win-probability

 

 

Source: Advanced NFL Stats

With 1:05 left and the Ravens ahead, 22-19 (following a Jacoby Jones 77-yard kick-off return), the Vikings had only a 6 percent chance of winning. After a 79-yard touchdown pass by the Vikes, the extra point, and a touchback on the kick-off, the odds had completely reversed. But after the Ravens scored again and the replay official upheld the score, the Ravens won despite the odds. In fact, over the final three minutes, the Vikings’ win probability flip-flopped from 87 percent to 20 percent to 76 percent to 6 percent to 93 percent before they lost. We should call it the “fat tail” game. Minnesota had a 93 percent chance of winning and travelled home disappointed nonetheless ( as a point of comparison, the 49ers had a lower win probability after going ahead of Seattle, 19-17, with 31 seconds remaining yesterday).

As Joe Flacco understated it, “It was unlikely that we win this game today, but we did.”

So the next time you run your various investment models and Monte Carlo simulations to conclude that you have — say — a 93 percent chance of success, remember the Vikings yesterday. Things can go south in a real hurry. The unlikely happens with surprising frequency. A 93 percent chance is not a sure thing, not by a long-shot.

-------------------------------------------------------------------------------------------------------------------

Got a comment or question about this post? Feel free to use the Ask Cullen section, leave a comment in the forum or send me a message on Twitter.
Robert Seawright

Robert Seawright

Robert P. Seawright is the Chief Investment & Information Officer for Madison Avenue Securities, a broker-dealer and investment advisory firm headquartered in San Diego, California. Its focus is the capital markets and personal finance from a data-driven perspective.

More Posts

  • http://www.fanbrowser.com/ Cowpoke

    Is there an Index that just invest in mankind?

  • rp1

    One solution is the FASB approach: declare them both winners of the game.

  • me

    I get the idea of probability, but in this case the offensive numbers teams put up now is a function of rule changes specifically designed to increase scoring.

  • http://pragcap.com/are-machines-really-taking-our-jobs Michael

    And that is why statisticians like to set the p-value at .05, i.e., because the associated 95% probability is better than 93%. 93% just doesn’t cut it.