How they should pick the playoff

Average in-game win probability among metrics committee should use

Updated: April 9, 2014, 12:31 PM ET
By Brad Edwards | ESPN Insider

StanfordDavid Madison/Getty ImagesAverage in-game win probability shows Stanford was slightly better at controlling games than MSU.

Another NCAA basketball tournament has come and gone, leaving us something fun to consider: The next college playoff that will capture the attention of general sports fans across America will be for football. The season that most of us have been anticipating for years, if not decades, is finally just a few months away.

Last week, the College Football Playoff selection committee continued its preparation by meeting for a couple of days to discuss some operational details. We were told that much discussion about recusal took place, which is one more example of how the football committee is being modeled after the longstanding basketball committee.

There is one noteworthy area, however, in which the football committee is likely to take a different approach from its basketball counterpart.

When the basketball committee arrives for its selection meeting every year, each committee member is given something known as a "Nitty Gritty Report." It's basically a chart of data that compares the teams across different splits of win-loss record and schedule strength, and it allows basketball committee members to be primarily focusing on the same set of numbers for data-driven discussion.

Current indications are that the football committee will have no common data source, and it will be up to each member to decide which numbers, if any, are worth evaluating. Some people might argue that this will cause the football committee to be less analytical than the basketball version, but I disagree. I think less structure will better allow a room of intelligent people to make their decisions as informed as possible.

Basketball committee members are not restricted to the data found on the Nitty Gritty Report, but those numbers being provided to the group gives them a feel of being official and probably limits debate over things like whether Team A's schedule was more challenging than Team B's. Football committee members, on the other hand, will need to seek and share information that they deem to be insightful, and numbers from many sources are likely to be debated.

Fortunately, huge strides have been made over the past few years in the world of college football analytics, and some of the new metrics could be just as useful to the CFP selection committee as they are to a coaching staff or a TV production crew.

Let's examine a couple of recent creations by ESPN Stats & Information that I hope will find their way into the committee's meeting room in early December and could help the committee pick the most deserving teams for a playoff.

Average In-Game Win Probability

Don't be overwhelmed by the name. It's a fairly simple concept.

By using play-by-play data from more than 7,000 Football Bowl Subdivision games since 2004, any game situation (score, time remaining, possession, yard line, down and distance) can be compared to similar situations in the data set to estimate a team's likelihood of winning. For example, a team up 28 at halftime would have a win probability of close to 100 percent, while a team down six at the opponent's 25 with a minute to play probably has around a 50 percent chance of winning.

By estimating a team's win probability for every snap, we can not only generate an average win probability throughout a game, but we can also compile an average in-game win probability across a sample of games (such as a season). This becomes useful when attempting to measure how well a team controlled play, and it tells a much more accurate story than the final score.

• Analyzes college football and the CFB Playoff as part of ESPN's Stats & Information Group
• Analyst for both College GameDay on ESPN Radio and the ESPN College Football app