How improbable was the historical comeback we saw yesterday, when New England came from down 28-3 against Atlanta in the second half, still trailing by 25 until about 2 minutes remained in the 3rd? Pretty dang improbable.
If your reaction to the talk that it was a 99% chance of Atlanta winning is that the win probabilities are wrong, then I think you need to recalibrate. It was a special game, precisely because it was so improbable. I mean, it wasn’t mathematical models that were declaring the game over.
Win probability models are based on past events and similar circumstances. Over the last 20 years, for example, teams are 2-264 when trailing by between 18-20 points entering the 4th quarter. If you limit it to favorites trailing, it was still 2-66, less than 3%. So if you are like “those estimates are wrong!” you are quibbling about whether it was 1% or 3%.
And the counterpoint to Tom Brady being at quarterback is that Atlanta was the #1 scoring offense in the league, had a big lead, and simply needed one more field goal over the last 20 minutes to win it.
So many plays had to go the Patriots’ way: the Edelman catch, the forced fumble, the sacks and holds, and not only scoring but converting two conversions. That’s a lot of 50/50 (or worse) plays that almost all had to work out.
And the win probability models only say the likelihood at a given point. If I told you that Atlanta would have a 28-3 lead halfway through the 3rd quarter, and THEN AFTER THAT have two different possessions where they had the ball at the New England 32 with a 2nd and 1, and at the New England 22 with a 1st and 10, you would have guessed they would have ended the game with more than 28 points.
How improbable was it that Atlanta came away with no points from those situations on two of their last three possessions before New England tied it?
Let’s start with the scenario at the end of the 3rd quarter, after the Patriots unsuccessfully tried an onside kick. Atlanta gained 9 on first down to set up a 2nd and 1, already inside field goal range at the New England 32. In 2016, there were 63 times that a team had the ball with 2nd and 2 or less, between the opponent 30 and 35. Eight of those eventually ended in a turnover on downs, kneel down, or time expiring (situations inapplicable to Atlanta). In the other 55, the team with the ball scored 80% of the time. The breakdown was 44 scores, 3 missed field goals, 5 turnovers, and 3 punts (Rams, 49ers, Dolphins).
And then there was the situation, still up 8, that will haunt Kyle Shanahan, Matt Ryan and company for a long time. The Julio Jones catch, now destined to be a footnote, should have for all intents won the game, putting Atlanta at the New England 22-yard line.
In 2016, there were 108 times a team had a 1st and 10 at the opponent 22. Five of them involved a turnover on downs (near the goal line, by teams needing a touchdown) or a kneel down to end the game. The remaining 103 resulted in a score 88% of the time, with 86 scores, 7 missed field goals, and 5 turnovers. There were no other punts out of that situation.
Ending up with no points, not running enough clock, and punting it back to New England was a truly remarkable chain of events. Even after New England started the comeback, the likelihood of a win was still slim once Atlanta got that deep with a first down.
At that point, New England had to get a stop with no points (something that happens about 12% of the time, and not adjusted for Atlanta’s #1 offense), march the field and get a touchdown, then get a 2-point conversion, then hold Atlanta, then win in overtime. Even if you think they had more than a 50% chance at all those things, the Patriots still had a 3% chance or less as late as the Julio Jones catch.
It’s difficult to truly comprehend how disastrous that series was. Being the first team to end up punting in that situation all year–in the Super Bowl–is one way to show how crazy it was.