Nate Silver looks to have gone 49 of 49 in projecting states in this year’s Presidential election, and called Florida “too close to call” and within the margin for a recount in his last projection before the election. Even that is right on, as we still await the final tally from Florida.
Leading up to the election, he was pilloried by many in attacks that made Murray Chass vs. Bloggers look sane. There were ad hominem attacks about his appearance or that claimed he was too effeminate. The guy that “unskewed” the polls by making them match what he believed would happen, and who took those shots, has now admitted he was kind of wrong.
Nate Silver is not the only one. There were plenty of analytical types who had Obama as a heavy favorite, even as pundits on television kept proclaiming it a tossup and some on the right were calling for a big victory the other way.
So why are analytical types not popping up left and right to nail sports results so much across the board? There are plenty of us who like to analyze things in baseball, or football, or basketball. No one can claim anywhere near the success of going nearly 99% over two elections, though.
Well, the difference is what we are measuring. Predicting the outcome of an election based on polling data is not the same as trying to predict an athletic contest before it occurs. When Nate Silver got involved in the SPI (Soccer Power Index) before the last World Cup, his results were not the same. It was a cool rating system to try to incorporate factors like looking at the best rosters, and performance in professional leagues. It still only barely outperformed the FIFA rankings in the final tourney results, and the best team by SPI went 33-13-18 compared with 34-12-18 for the FIFA rankings.
If we are going to try to draw some better analogies to sports, we have to look at what would happen if we knew some of the in-game results, and could then try to project outcome. In Silver’s case, he’s relying on the state polls over the last several months. He has his internal proprietary formulas for how to tweak and which ones to put the most weight on, but the basis is those polling numbers. They consistently showed Obama with a lead in key swing states that would decide the election.
If you showed me a snap shot sampling of the plays from a game before it occurred, and I was confident there was no fraud or manipulation to deceive me, I could do a much better job of predicting who would win the game. Football’s not even the best sport to use here, though, because I might miss THE key play, like a 90 yard interception return, that accounts for a huge amount of the outcome. There are no 90 yard interception returns in politics, unless a candidate gets legitimately raked at the polls because of comments that swing states that went strongly for their party otherwise (looking at you, Mourdock and Akin).
Basketball is probably a better one, where knowing the outcome of possessions, all of which have roughly the same value, matters. Let’s say you gave me a sampling of 30% of plays from a Bulls-Pistons game and I knew the Bulls “won” 54% of those plays. Even though it sounds like a close margin, I could probably project that it is more than a toss up that the Bulls won the game. Maybe there was some systematic bias in my sampling, and maybe I just got unlucky with what I sampled. If I get to take small sample slices over and over again, though, and they are generally consistent, I can be a little more confident I am going to get it right.
Another sports example that more accurately fits what is going on here, rather than comparing what Silver accomplished to predicting the outcomes of future games, is looking at punditry and prognostication for the NCAA tournament selections. Joe Lunardi went 67 for 68 in projecting the field last year, and I broke down why that wasn’t as impressive as it sounds.
Just like elections, those at-large spots are like up or down votes on the candidates. Most of them are obvious. There are a few that are uncertain. Now, if I could actually sample some of the people voting on those spots, my predictions would be even better with that information.
The difference, though, is that the political arena has a lot of people braying about intangibles and gut feelings, and those like Silver stand out more. At least when it comes to the basketball selections, everyone knows about the RPI, even if they don’t believe in it. No one is predicting a team will make it just because they think the electorate is blowing one way or another like the wind without proof. Nate Silver did a great job, but he stands out even more because of the competition.
In sports, there are bounces, luck factors, and huge swings, and we don’t have any polling telling us how the plays will likely vote. In politics, if you are getting good data, the luck factors aren’t as big as those networks that want us to tune in want us to believe.
[photo via salon.com]