Another narrative example from sports

I can learn to get along \ With all the things I don't know

I read an excellent example of explaining advanced statistics to a lay audience in an ESPN college football article (behind ESPN+ paywall) on what to expect from upcoming football games. This isn’t the first time I’ve seen sportswriting as a good example for us to follow. In this one, Bill Connelly wants us to know that,

while [the University of Southern California] USC's transfer-heavy offense has been every bit as good as advertised […] the Trojans have also been absurdly lucky.

How have they been lucky? Statistics can tell us. I’ll keep most of this section in here verbatim so you can see how the author wrote it. Bolding is mine, for emphasis.

USC's incredible current run of sticky fingers gives us good reason to have a quick conversation about turnovers luck.

Fumbles: While odds shift a bit based on where a fumble happens and how many of each team's players are near, your team can generally expect to recover 50% of all fumbles over a longer period of time.

In two games, USC recovered both of its own fumbles and both of its opponents' fumbles.

So, where you would expect USC to have recovered about two out of four fumbles, they recovered all four. The author doesn’t go through all the details and nuances here, just the main takeaway. I love that for this audience.

Interceptions: While odds can shift based on how much zone or man defense you play (and, therefore, how many defenders you have either facing the QB or close to the ball), you can generally expect to end up with a ratio of about one interception to every four pass breakups.

In two games, USC has intercepted six passes and broken up seven, while opponents have broken up one and intercepted zero.

So USC has had six interceptions when they would have expected about two.

Compare the average output to what USC has generated thus far, and you see that USC's expected turnover margin should be about +2.5 -- opponents should have committed about 3.7 turnovers on average to USC's 1.2. That's good! There's randomness baked into the turnovers cake, but generating turnover opportunities isn't totally random. Alex Grinch's USC defense wants to be aggressive, and this is a good sign that it's working to a degree.

On the other hand, USC's actual turnover margin right now is +8, far better than expected.

I actually like that the author didn’t break it down like I did above, and rather cut to the chase — USC has generated +6 in turnover margin, at least in part from luck. This is probably sufficient to tell an interesting story, but Bill Connelly goes a step further:

Based on field position value, turnovers are generally worth around four to five points. The differences in last week's 41-28 victory over Stanford alone (+1.6 expected turnover margin, +4 actual) were worth about 10 to 12 in a 13-point win. […] when the turnovers luck dries up, USC is going to find itself having to win quite a few 49-45 games without immediate improvement.

What does all this luck mean? It means that, without the luck, USC’s win last week just about vanishes, perhaps suggesting that USC isn’t as good as it looks, and will need to improve in other ways “once the luck dries up.”

Of course, with the benefit of a few days waiting, we know that USC actually stomped Fresno State in that upcoming game, which shows what can happen when we try to respond to stakeholder requests based on limited data.