This recent article on ESPN cites the following:
How improbable was Sunday? According to Elias, teams trailing by at least 24 points at halftime were 5-617 entering Sunday night. Make it six.
For context: On 11/24, the Patriots were losing to the Broncos by a score of 24-0 halfway through the game. They ultimately rallied back to win the game.
No doubt, this was an exciting match. But calling the Patriots’ odds of victory at 5-617 is absolutely bunk.
To illustrate: Suppose that of those 622 games (5 + 617), 617 teams were losing at halftime by a score of 9,001-0. (Not an easy score to accomplish in the NFL…but bear with me.) Those 617 teams all lost handily — that’s an impossibly difficult comeback to mount! Further, suppose that the remaining five teams were each down 24-0 and all five of them came back and won their games.
In that fantasy scenario, it’s still technically correct to say that “teams trailing by at least 24 points were 5-617.” But you can see why the statistic is horribly misleading — in fact, without changing the world, you could also say “teams trailing by 24 points came back to win every single time” and be just as correct.
Back in our real-world Broncos & Patriots scenario: It’s worth pointing out that the Broncos scored its 24th point with over 6 minutes remaining in the 2nd quarter. So the Patriots actually had 36 minutes (15 minutes per quarter, plus the aforementioned 6) to mount their comeback, rather than 30. That’s 20% more time!
This sort of idiosyncrasy happens in sports ALL. THE. TIME. A slightly different example: In baseball, when the announcer says “Derek Jeter has gotten a hit in 3 of his last 4 at bats” you can all but guarantee that Jeter got an out in his 5th-to-last plate appearance, and the announcer could have said “Derek Jeter has gotten a hit in 3 of his last 5 at bats.” (Otherwise, of course, he’d have just said “Jeter has gotten a hit in 4 of his last 5 at bats” because that’d have made a more compelling point!)
Though I find this tends to happen in sports most abundantly, because sports fans find statistics like this to be fascinating, it’s just as applicable anywhere. When you hear something like “Facebook missed its earnings projections two quarters in a row,” you can make a strong bet that three quarters back, the company met or exceeded its benchmark.
Choose your own lesson, depending on how optimistically you view the world:
A. This is just a fun way that stats work. You can tease your friends about how harmlessly inaccurate the Derek Jeter statistic is and how dopey the announcer is for declaring it, and not put too much thought into things.
or
B. Be very careful of the biases and values of your sources — because even statistics that are based in fact can be warped to deceptively convey a very biased opinion.