Mike Tomlin's decision to go for an on-side kick when the Steelers led 30-28 with just over 4 minutes left against Green Bay has been poked at and prodded -- though not nearly to the extent of Bill Belichick's 4th down gamble against Indy earlier this season, perhaps because of the teams involved in that game and also the multitude of other interesting NFL story lines that emerged from this weekend. But we wanted to take one more pass through it because, unlike the Belichick decision, we did not like the Tomlin decision soon after it happened (though we also wondered what the numbers might have to say about it). First, here is Tomlin's explanation:

"I'll be very bluntly honest with you, based on the way the game was going in the second half, first of all I thought with the element of surprise we had a chance to get it, but if we didn't get it and they were to score, then we would have necessary time on the clock to score or match their score. Plan A didn't work, we got the ball but we were illegal, that was the correct call, but it kind of unfolded the way you envisioned it. "We had 30 minutes of evidence that we could drive the ball on them, we also conversely had 30 minutes of evidence to show they could also drive the ball on us. That's why we took the risk when we did. We were just trying to win the football game. There was time left in that game that had we kicked that ball away and the half had gone the way that it'd gone, they were converting third downs. They would have moved the ball down the field on us, we wouldn't have had necessary time to respond. I'm just being honest, but it starts with feeling pretty good about the element of surprise and having a good chance to get that ball, but that part of it didn't work out."

And now, here is the upshot of a report from Advanced NFL Stats, which also advocated for the Belichick move:

The onside kick is the better decision by 0.64 to 0.54 WP [win probability]. These estimates are only league baselines, but they suggest it was probably a good call. What's most interesting to me is that a failed onside kick is hardly certain death--a 0.42 WP. There was plenty of time for anything to happen--a stop, a turnover, or a score. And sure enough the Steelers gave up a touchdown but came back with one of their own. Also, a successful onside recovery doesn't seal the game. The Steelers would still need at least two first downs to clinch the win. Essentially, the Steelers traded 30 yards of field position for the chance to keep the ball out of the Packers' hands.

Now, there are some things to consider. First, the Advanced NFL Stats computation was based on a 60 percent success rate for "surprise on-side kicks," vs. a 20 percent success rate for all on-side kicks. That said, even if the success rate for a "surprise" on-side kick (and we should be able to agree that the move by Pittsburgh came as a surprise) was merely 40 percent, it works out to a .56 win probability using the formula on the Advanced NFL Stats link -- still better than the .54 predicted by kicking deep. Perhaps most surprising is that even if the on-side kick fails, there is a predicted 42 percent chance the Steelers will win (the .64 WP is calculated based on adding the likelihood of winning with a successful on-side kick and an unsuccessful kick). As it turned out, the game transpired in one of the ways Tomlin envisioned: Green Bay scored, but the Steelers had time (just enough, we might add) to come back and get the winning score.

In any event, it was another interesting decision that caused us to rethink how we watch games and judge decisions. Generic probability shouldn't be the end-all be-all of decisions, but it can lend support to moves that buck tradition.