<
>

The gold standard for predictions

4/15/2013

For the past five years, we have conducted the ESPN Summer Forecast series, usually in August. (In 2011, it was called ESPN Winter Forecast because it took place in December, after the NBA lockout.)

From 2008 to 2012, ESPN.com has asked a panel of NBA contributors to predict how many wins each team will have the following season. In 2012, for instance, we had exactly 100 experts participating in the ESPN Summer Forecast.

Beating the competition

Each year, a number of different predictions and sophisticated forecasting systems, including ESPN Forecast, have "competed" to be the most accurate.

These results have been researched and compiled by NBA analyst Michael Wilczynski, and there is a surprising finding:
For those five years, when the results are aggregated, the most accurate predictions of all have come from the ESPN Forecast system.

This is despite one significant issue that the ESPN Forecast has had to overcome: It is usually conducted about 12 weeks before the NBA season starts.

For instance, in August 2012, when we published our predictions, we didn't know the Oklahoma City Thunder and Houston Rockets would shake up their teams with the James Harden trade shortly before the season. We didn't know that Andrew Bynum would go down before the season, shattering all the optimistic forecasts for the 76ers. We didn't know that Dirk Nowitzki would have surprise knee surgery in October and miss more than two months of the season. We didn't know Kevin Love would break his hand in October, sending the Timberwolves tumbling.

The other forecasting systems? In most cases, they were able to incorporate all of this news into their predictions.

And still, the ESPN Forecast is nearly at the very top of the heap once again. Leading the way this season among the 34 predictions compiled: John Hollinger, the online gambling site Bovada and ESPN Summer Forecast.

One note about Vegas: The sports books and online gambling sites put out a variety of predictive numbers, often in the form of over/under lines (which are set not as an NBA forecast per se but to encourage bettors to predict whether a team will go "over" or "under" the Vegas projection, with, it is hoped, an equal number of bettors on each side of the line). In many cases, those "predictions" turn out to be very accurate, because they, like ESPN Forecast, are set to gauge what a large number of people believe will happen. But in the four years for which we have multiple "Vegas" projections, the ESPN Forecast has been more accurate once and the "Vegas" numbers more accurate once, with two split decisions.

Setting the standard

For the past five seasons, there have been more than 100 predictions compiled, including about 30 that have been compiled for multiple seasons.
Overall, among the predictions that have been compiled for the last five seasons, ESPN Forecast is at the very top.

If this were the Olympics of forecasting, the medals would be awarded this way:

Gold: ESPN Forecast
Silver: John Hollinger
Bronze: Las Vegas

Welcome to the gold standard of sports forecasting: ESPN Forecast.

The following provided direction for ESPN Forecast: Microsoft Research economist David Rothschild, Microsoft Research microeconomist Justin Rao, TrueHoop senior writer Henry Abbott, TrueHoop editor Kevin Arnovitz, ESPN.com senior NBA coordinator Chris Ramsay and ESPN.com NBA editor Royce Webb.