By Alexander Joenks | Miscellaneous | July 27, 2015 |
By Alexander Joenks | Miscellaneous | July 27, 2015 |
There are three types of lies, the old saying goes: lies, damned lies, and statistics.
Trust me on this. I’ve got enough graduate level statistics that I can get in pedantic arguments about Nate Silver’s techniques while using only six syllable words that don’t appear in the dictionary. The trick to statistics is not just that they’re only as good as the data that are put in, but that the approach itself is only as good as the statistician’s ability to identify where assumptions lead to breakdowns.
In other words, when your model tells you something unexpected, that’s interesting and a starting point for possibly figuring out something we didn’t realize. But when your model spits out bullshit, that’s a sign that the model is just shit in the first place, not that it’s picking up on something revolutionary that no expert has ever considered before.
For instance, the Harvard Sports Analytic Collective did some stats based on an aggregation of individual rankings of players and figured out which teams were the strongest in the NFL in this coming season. It is the funniest sports article that I’ve read in quite some time.
For instance, the third best team in the NFL is going to be the Dolphins, and the fourth the Kansas City Chiefs. That’s right, the Dolphins are going to win the AFC East and battle with the Chiefs for the AFC Championship.
Sweet Namath’s ghost, that’s like running a bunch of statistics and predicting that the Republican nominee for President will be Che Guevara’s chainsaw wielding zombie.
And that’s not even the punchline. The third best team in the NFC this coming season? Drumroll please?
Ladies and gentlemen, the 2015 Detroit Lions!
And that’s what happens when your model divides by zero.
(source: Harvard Sports Analysis)