Keys to Success in a Run-and-Gun Basketball System Thomas L. Moore (Grinnell College), Benjamin Johannsen (Northwestern University), Eric Ohrn (University of Michigan) August 4, 2011 talk: Joint Statistical Meetings Abstract: Since 1991, Coach David Arseneault of Grinnell College has used a unique style of run-and-gun basketball that he calls “The System.” In a 1997 book, he outlines what he considers the keys to winning basketball games within “The System.” These keys are defined as meeting thresholds in various offensive and defensive game statistics. Using data from 9 seasons, we use classical statistical methods to challenge Coach Arseneault’s keys and, where possible, refine them. Introduction: Coaching men’s basketball at Division III Grinnell College has not been a way station toward the Hall-of-Fame in Springfield, Massachusetts. Of the 9 coaches who have held that post since World War II, none had career winning records until the current coach David Arseneault, who came on the scene in 1989. David’s first two years stretched a streak of 26 straight losing seasons into 28, before his 11-11 season got Grinnellians used to the notion that they might permit themselves to sniff success from time to time in the win-loss column. Since then, David has experienced just 7 losing seasons the past 20, has achieved a career coaching record of 275 wins versus 230 losses, has guided his team to 5 Midwest Conference championships, and has been the conference’s “coach of the year” 5 times. While by Grinnell’s standards—which never put preeminence on wins and losses—this record is excellent, it’s still not the stuff of Springfield, Mass, nor would it explain the list below, a list that Grant Wahl of Sports Illustrated put together in a 2003 article to answer a fan’s question: Who are the most innovative men’s college basketball coaches? Roy Williams Bill Carmody Rick Pitino Mike Krzyzewski Bob Knight Jim Boeheim Lute Olsen David Arseneault Tubby Smith Billy Donovan Why would Coach A be included on such a list? Consider this quote from Wahl’s article:“You've gotta see it to believe it. Arseneault's D-3 Pioneers make those legendary Loyola Marymount teams look like they were running the four-corners offense. Using an attack that emphasizes shooting the ball every 12 seconds, giving up 2-pointers so you can get 3s, and going up to 17 players deep, Grinnell has led all NCAA levels in scoring and 3-point shooting for the past 10 seasons. (This year's Pioneers, we might add, are 5-0 and averaging 133.2 points and 21.6 treys per game.)” Using a style of play that eventually became known as “The System” David transformed a moribund program into one that is fun for the players—where all members of the team have a role—and fun for the fans. Terry Glasgow, coach of conference rival Monmouth College, has described playing against Grinnell as like opening up a shoe box with 5 mice in it. Coach Areseneault said recently that he worries when a player dwells too much on winning; he wants them to enjoy playing with abandon and piling up weird statistics. Consistently the team leads the NCAA in scoring average per game, three point shooting, and steals. Grinnell has led the country in scoring 16 of the past 18 years, with a high-water mark of 126 points per game in 2002-03. This past year they led in all three categories averaging 103 points per game scoring, on nearly 18 threes, and 15 steals. In his 1997 book The Running Game, Coach Arseneault states a “formula for success” in terms of 5 statistical goals for each game: Make at least 150 trips up and down the court for the game; Grinnell takes at least 94 shots in the game; At least half of these shots are three-point attempts; Grinnell rebounds at least 33% of its missed shots; Grinnell forces the opponent into at least 32 turnovers. David derived these keys based upon another student project he had directed a couple of years into the system. By “trips,” Coach Arseneault actually recorded from game film how many times the opposing center traveled between the two hash marks of the court. This was the only statistic mentioned in his keys that could not be recorded or derived from the stat sheets and it was tedious to record. (Stat sheets are statistical summaries prepared by the sports information director and his crew during the game.) Moreover, he found it useful to compare his keys at the half-time of a game, simply by viewing the half-time stat sheets and this made the trips key impractical. Thus he substituted for his first key to success one based upon shot differential, namely: Grinnell should attempt at least 25 more shots than its opponent. The data and basic analysis: This JSM talk comes from a student project that Ben and Eric did for a class in the spring of 2006 that was taught by Tom. Ben and Eric are each now well into the pursuit of PhDs in economics; at Grinnell both were top students and excellent athletes: Ben in golf and Eric, tennis. Both were enthusiastic fans of our men’s basketball program, which led them to this project.They collected data on all 147 Midwest Conference games for a 9 year period ending in 2005-06. Using the stat sheets they created a data set with 147 cases and over 30 variables. We call this data set Hoops. Subsequently, Tom selected a second data set, which we call LateYears, comprising all 88 conference games from the 2006-07 season through the 2010-11 season. We use Hoops to build models and LateYears to validate them. Table 1 summarizes four possible approaches to modeling the data, all of which we considered and tried to some extent. We ultimately preferred the fourth of these—ordinary least squares regression, with the response variable being Point Differential and using for predictors the variables from which the keys to success are defined. The option of making Y binary and using logistic regression (the top, left-hand option) gave similar results to what our approach found, but there is some loss in precision with using a binary response, whereas Point Differential affords a more nuanced view of the game’s result. Moreover, Point Differential better captures the spirit with which Grinnell tries to approach athletic competition: a 2 point loss really is a better outcome—competitive game, excitement for fans and players—than a 20 point loss. The disadvantage of using