Win Probability is a statistical method that “determines” the chances a team has of winning a game at any point in the game. A team ahead by two runs in the ninth inning has a better chance of winning than if it were leading by three runs in the first inning.
The Tampa Bay Rays trailed the New York Yankees, 7-0 after seven innings in their final game of the 2011 season. They had a one-third of one percent chance of winning ,or one chance in 333. Tampa won.
The problem with Win Probability is that it applies data from the past to a situation that fails to fit into its model.
The Yankees used 11 pitchers in a 12 inning game. They scored seven runs over the first five innings and then were held scoreless.
Could the fact that the game was meaningless to the Yankees (right, as if the Yankees would ever lay down to Tampa to hurt the Boston Red Sox) have contributed to their lack of scoring? Could the fact that most of the Yankees regulars were replaced relatively early in the game affected the potency of their offense?
The Rays were highly motivated to win. They were attempting to finish a miracle comeback that rivaled that of 2007 Philadelphia Phillies and the 1969 New York Mets. Manager Joe Maddon pulled out all the stops.
But despite replacing their regulars, the Yankees were still ahead by seven runs going to the eighth inning. The real advantage for the Rays, a tremendous variable that Win Probability does not take into account, is the fact that the Yankees used 11 pitchers.
Joe Girardi didn’t follow the modern formula of having the starter go six or seven innings, use the bullpen and then bring in the closer. He didn’t use the old method of having the starter to pitch as long as he was still effective.
The combination of Boone Logan and Luis Ayala allowed six runs in Tampa’s eighth inning. Here is where statistics fail.
Based on the data used for Win Probability to determine a team’s chances of winning, the Yankees should have brought in their best relief pitcher to close out the Rays in the ninth inning. That would fit the Win Probability model.
Girardi had Corey Wade relieve Luis Ayala to close out the game. No one can fault him for not using Mariano Rivera with the start of the playoffs looming.
Does anyone thing that Tampa had a better chance of scoring a run off Wade than off Rivera? What are the chances of Rivera giving up a two-out, two-strike game tying home run?
Then Win Probability really failed.
Scott Proctor was the last pitcher Girardi was willing to use, regardless of how long he had to pitch. The Atlanta Braves released Proctor and the Yankees, some say as payback for Joe Torre almost ruining his career by using him in 83 games in 2006, signed him.
With the Braves, Proctor had a 6.44 ERA, which ballooned to 9.00 with the Yankees. He pitched effectively against Tampa for two and two-thirds innings or until Evan Longoria won the game with a home run in the 12th inning.
The key point is simply that Win Probability is merely a guide that fails to include many variables, such as the Rays desire and need to win, the Yankees preparing for the playoffs, the Yankees not really worried if they won or not and the fact that the Yankees didn’t use their closer.
The Rays had a Win Probability of 0.003 percent after seven innings. We all know that’s not true.