Early in the season, we explored projections for the 2013-14 NBA season, looking specifically at how analytics folks viewed things. For picks and more, I encourage readers to view that post here.
Because we are roughly at the halfway point of the season, I figured it was a good time to revisit those projections, and to identify which sites were off to the strongest starts. To do this, I scaled each team’s win percentage to 82 games, using current win percentages on ESPN.com.
Next, I calculated the mean absolute error between each site’s projections and those scaled wins. Here are the results:
|Site||Mean Absolute Error (MAE)|
|Las Vegas projections||7.85|
|Box Score Geeks||9.11|
On average, Team Rankings has outperformed Las Vegas projections by about half of a win (MAE = 7.29 vs. 7.85), and currently leads these projection systems, at least as judged by MAE.
The plot below depicts each site’s projections, arranged in order from predicted finish (low to high). Teams highlighted with green text are those which were, on average, accurately diagnosed by the stats-sites. Teams in red text indicate inaccurate projections.
Of the teams which are outperforming their expectations by several wins, including Phoenix, Portland, Oklahoma City, and Indiana, the stats-sites only anticipated Phoenix doing so, and, on average, missed on the others. Meanwhile, the sites foresaw the poor seasons, at least relative to Las Vegas’ expectations, of New Orleans, Cleveland, New York, Brooklyn, and Chicago.
On the whole, stats-sites hit on 19 of 30 teams, which is a pretty solid performance. Even better, 8 of the 10 ‘best-bets’ from the preseason post are on pace to win.
As examples of its accuracy, Team Rankings, shown on the graph with a light blue symbol, was one of the only sites with relatively close projections of Detroit and Utah’s seasons.
At the end of the season, I’ll update these numbers, likely with additional metrics. Until then, thanks for reading!