Yet another NBA stat gone bad at FiveThirtyEight

Following the Golden State Warriors’ collapse last week, FiveThirtyEight bungled some numbers and suggested they were the third best team in NBA history. They came to this conclusion by taking the peak, average, and end of season ELO rating, calculating a blend of those, and simply ranking the blend numerically. This bungling of numbers tells you weird things. Like that this year’s Spurs, who didn’t escape the second round, were the sixth best team in history. Or that this year’s champion Cavaliers — televised evidence to the contrary — were not better than the Warriors. The blend method also suggests these Cavs weren’t even as good as the 2009 Cavs, who like the Spurs, were not only were not champions, they didn’t even make the finals.

A simpler, better method of ranking the “best” teams would be to rank them by end ELO. You might actually get the people who play, coach, and manage basketball teams to agree that they are trying to attain the highest end of season ELO rating — probably not a true statement in regards to the statistical shenanigans behind the blend method. Simply re-ranking by the highest end ELO results in a list with eventual NBA champions representing the 13 highest ratings. Which tells us what best-of-seven series have told us the past few decades: the best teams win.