Jeff Sagarin (born 1948)[1] is an American sportsstatistician known for his development of amethod for ranking and rating sports teams in a variety of sports.[2] HisSagarin Ratings have been a regular feature in theUSA Today sports section from 1985 to 2023,[2][3] have been used by theNCAA Tournament Selection Committee to help determine the participants in theNCAA Men's Division I Basketball Championship tournament since 1984,[4] and were part of thecollege footballBowl Championship Series throughout its history from 1998 to 2014.[2][5]
Sagarin attendedNew Rochelle High School, earned aBachelor of Science inmathematics from theMassachusetts Institute of Technology in 1970 and an M.B.A. fromIndiana University School of Business in 1983.[2][6] He grew up outsideNew York City inWestchester County inNew Rochelle, New York. In 1977 he moved toBloomington, Indiana. In 1986 he created the computer gameHoops withWayne L. Winston.[7]
Sagarin, like the developers of many othersports rating systems, does not divulge the exact methods behind his system. He offers two rating systems, each of which gives each team a certain number of points. One system, "Elo chess," is presumably based on theElo rating system used internationally to rank chess players. This system uses only wins and losses with no reference to the victory margin. The other system, "Predictor," takes victory margin into account. For that system the difference in two teams' rating scores is meant to predict the margin of victory for the stronger team at a neutral venue. For both systems teams gain higher ratings within the Sagarin system by winning games against stronger opponents, factoring in such things as home-venue advantage. For the Predictor system, margin of victory (or defeat) factors in also, but alaw of diminishing returns is applied. Therefore, afootball team that wins a game by a margin of 7–6 is rewarded less than a team that defeats the same opponent under the same circumstances 21–7, but a team that wins a game by a margin of 35–0 receives similar ratings to a team that defeats the same opponent 70–0. This characteristic has the effect of recognizing "comfortable" victories, while limiting the reward for running up the score.
At the beginning of a season, when only a few games have been played, aBayesian network weighted by starting rankings is used as long as there are whole groups of teams that have not played one another, but once thegraph is well-connected, the weights are no longer needed. Sagarin claims that from that point, the rankings are unbiased.[8]
Sagarin's ratings are particularly relevant in the world ofAmerican collegefootball andbasketball, where, with hundreds of teams inNCAA Division I competition, there is no way a team can play against more than a small fraction of its competitors. Therefore, in determining the participants in championship games and tournaments, it is necessary to distinguish between teams that have compiled impressive win–loss records against strong competition and teams that have defeated weaker opponents.
In addition, sports rating systems are generally of great interest togamblers. Gamblers use Sagarin's ratings as a source of "Power Ratings," traditionally used as a way to determine the spread between two teams.
Along with his formerMIT classmate Wayne L. Winston, a professor of decision sciences atIndiana University, as of 2004[update], Sagarin advised theDallas Mavericks about which lineups to use during games and which free agents to sign using a system called Winval.[9][10] Winval is modeled after hockey'splus–minus system.
His rankings, published weekly in USA Today, are part of the complicated formula that will be used to determine the pairing in the Fiesta Bowl. Sagarin also has been a consultant for BCS boss Roy Kramer, who's something of a math whiz himself.