Indianapolis Colts' Advanced Stat of the Week: Win Probability

Nate Dunlevy@NateDunlevyGuest ColumnistMay 23, 2012

The ugliest possible picture for a Colts fan to see.
The ugliest possible picture for a Colts fan to see.Andy Lyons/Getty Images

This week, we turn our attention to stats that measure the odds that a team will win a given game.

Win Probability and its player-driven cousin Win Probability Added (WPA) are stats advanced by Brian Burke of Advanced NFL Stats. Win Probability calculates the likelihood that a team will win the game before each and every play. WPA adds up the impact that an individual player has on those odds.

For instance, a team with a 40-point lead with 45 seconds to play has a Win Probability of 100 percent. Win Probability is a stat that explains what happened after the fact. Swings in Win Probability illustrate which plays had the most impact on a game.

These plays can be charted to produce a graphical representation of the swings of NFL games. For games with twists and turns, these charts can be dynamic.

The Colts' most embarrassing loss of 2011 illustrates that nicely.

Against the Chiefs in Week 5, Indy jumped off to a lead. After sacking Matt Cassel to end the first quarter, the Colts had a seven-point advantage and the ball. At this point, they were 75 percent favorites to win.

Curtis Painter pushed the lead 17 points with a 67-yard strike to Pierre Garcon with 11:23 to play in the half. Teams with 17 point leads at home win 93 percent of the time.

The teams traded touchdowns and after the Colts scored to go up 24-7 with just over a minute to play in the half, their Win Probability still hovered at better than 90 percent.

The game would start to unravel from there. The Indianapolis offense stalled, and the Chiefs methodically put up points. Their comeback wasn't punctuated by big plays, but rather a slow and total collapse by Indianapolis. Still, after three quarters of play, the Colts had a three-point lead, the ball, and a 69 percent chance of pulling out their first win.

What's interesting about the graph is that it illustrates a counter-intuitive truth about NFL games. Teams trailing by three mid-way through the fourth quarter are more likely to win than teams that are ahead. The reason is that most NFL coaches are too conservative with the lead and make sub-optimal decisions. Coaches that are behind get desperate and tend do things that help them win like pass more and go for it on fourth down. In this case, the Chiefs crossed the 50 percent threshold with 8:04 to play, down three, and with the ball.

After Kansas City scored to take the lead, they had a Win Probability of 77 percent with five minutes on the clock and a four-point lead. The Colts would push that number down to 60 percent after a long pass to Austin Collie, but Painter would not complete another pass the rest of the game.

Unlike other games with a big swing in probability, Painter wound up with a solid WPA score for the game. He was at 0.2, in part because the Chiefs' comeback was so gradual. He had a good passing day, and the Colts should have been able to win based on the things he did. The collapse fell squarely on the defense, rather than a big turnover by Painter.

WPA isn't necessarily predictive, but it is a fun way to look at how big of an impact each play has on the final outcome.