NAME: NFL Scores for 1998-2000 TYPE: Census (All values from 1998, 1999 and 2000 regular seasons) SIZE: 736 games, 7 variables DESCRIPTIVE ABSTRACT: The dataset contains scores for all regular season National Football League games from the 1998, 1999 and 2000 seasons. In addition to the points scored by the home and visiting teams in each game, the dataset contains a pointspread that handicaps each game. SOURCE: The pointspreads were taken from ESPN's Pigskin Pickem game (http://games.espn.go.com). Game scores are from personal records. VARIABLE DESCRIPTIONS: Columns 1 - 4 Year (1998, 1999, or 2000) 6 - 7 Week of the season (1 to 17) 10 - 27 Home team name 29 - 30 Home team score 33 - 50 Visiting team name 52 - 53 Visiting team score 56 - 60 Pointspread Values are aligned and delimited by blanks. STORY BEHIND THE DATA: The dataset contains scores for all regular season games from the 1998, 1999 and 2000 NFL seasons. The pointspread for each game is set (before the game is played) as a device to handicap the game. One may view the pointspread as a "predicted" margin of victory for the home team. To determine the winner "against the spread" the value of the pointspread is added to the visiting team's score. Note that a negative pointspread indicates that the visiting team is favored over the home team and will have points subtracted from its score. All pointspreads in this dataset are of the form (integer + 0.5) so there is always a pointspread winner for each game. PEDAGOGICAL NOTES: The NFL data can be used to investigate a wide variety of questions that might be generated by sports-minded students. Is there a home field advantage? What's a typical margin of victory in an NFL game? What does the distribution of scores look like? How well do the pointspreads predict actual game outcomes? Is there a good algorithm for picking winners (either outright or against the spread)? An article based on some analyses of these data appears in the Winter 2002 issue of STATS magazine. Data from earlier seasons (1993-1996) can be found in the Journal of Statistics Education data archive along with an article, "NFL Scores and Pointspreads", in the _Journal of Statistics Education_ , Volume 5, Number 3 (November 1997). SUBMITTED BY: Robin H. Lock Burry Professor of Statistics Department of Mathematics, Computer Science & Statistics St. Lawrence University Canton, NY 13617 rlock@stlawu.edu