Interview with Bud Goode
Interview conducted in 1995.
"When the Green Bay Packers pay $14,000,000 to Reggie White, the ex-Eagle defensive end, they are getting a player who earns 16 sacks in a 16-game season."
Bud Goode lives in Studio City, California, and is active as statistical consultant to many of the National Football League coaches. His statistical services have been purchased by 21 of 28 teams. Goode has written for The New York Times, sports Illustrated, The Washington Post, Inside Sports, "PRO" magazine, and the Super Bowl Program. In August 1993, he was recognized by the American Statistical Association for his pioneering work in statistical analysis of sports and he is currently active in teaching statistics concepts with sports at Belmont High School in Los Angeles, California.
Aczel: How did you first become interested in statistics and what about the field caught your attention?
Goode: I took my first course in introductory statistics at Occidental College in 1941 as a freshman in the Department of Physical Education. The department head, Dr. Carl Trieb, turned out 9 of the top 10 scores in the annual test for phys-ed teachers (in the Los Angeles City Unified School District). Simple reason: His brother, Dr. Martin Trieb, ran the Boys' Phys Ed Department.
The course was called "Tests and Measurements." Our text was the little red book by Professor R. A. Fisher. I still have it. The concepts of central tendency, variation, and correlation were so appealing I thought of giving up religion for statistics. The war came along and when I returned, having been blessed with a full disability pardon from the service in 1946, I would no longer fill the physical requirements of phys ed. I used that as an excuse to move to another major (psychology - and more statistics).
Aczel: Did you ever expect that the ASA would honor your work as they did in August of 1993?
Goode: No. Although I had started my research circa in 1961 (the head writer on the Groucho Marx "You Bet Your Life" TV show was responsible for this new direction), I had, indeed, been doing basic statistical research, using an old 1403 computer and later the IBM 7094 at UCLA, but I did not know how much my work had penetrated the ASA hallways. It was a surprise when Stu called me.
Aczel: When did you find out?
Goode: ASA President Elect, Stu Hunter, called before the Boston convention in 92 and advised that he was considering some kind of luncheon with me speaking for 30 minutes … but I couldn't make it to Boston because of modest health problems. He hoped I could do both the Boston and San Francisco conventions. I obviously made the trip to 'Frisco.
Aczel: You are interested in statistics education at the elementary level. How did that get started?
Goode: My interest in early schooling/teaching statistics was part of my life: I had earned a teaching credential [Calif. State General Secondary Credential Science] circa 1950. I have always been a frustrated teacher. When we had the 50th anniversary of Belmont High School graduating class, my close childhood friend, three-star General Thomas Howard Tackaberry, III, flew out from his home in North Carolina. We were both class officers in 41. The principal, Dr. Martha Helm [sic] spoke to us and when she learned my clients (in what I call my retirement) were NFL coaches, she asked if I would visit the school as a speaker.
I did, and have been going back every Friday since, as a volunteer teacher. I teach introductory statistics, making the Monday Night Game of the Week a homework assignment. It works. We move on to the reasons why students are successful in school, and why some folks are more successful in life than others.
Aczel: Let's turn to sports - football in particular - when did you start analyzing football?
Goode: I started the sports/football analysis in 1960, with a prediction of the Rose Bowl in January 1961. This was a result of the Groucho Marx head writer invitation to bid on a marketing/promotion account for a computer service bureau …. When asked by Groucho's head writer (the late Hy Friedman) to bid on a marketing/promotion account for the computerized service bureau he was taking public, I suggested we do promotional pieces on "… computer looks at Rose Bowl statistics … and predicts the winner."
That suggestion sold the idea to the board of directors of a small holding company. The computer company went public, and I started to do the marketing and promotion.
What happened? Six of the seven local TV stations brought their camera crews to the computer company and did interviews; 13 radio station interviews and a page in the Sunday Herald/Examiner saturated the Los Angeles area.
It was an immediate success.
In all fairness, we did not use the computer at the time. We did scattergrams of the stats that came out of the Football News and used variables where the scattergrams pictured strong relationships with our three measures of success - points scored on offense, points allowed on defense, and the difference between the two (the winning margin). It wasn't until the company (Universal Data Processing) moved from the IBM 1401 level to the 1403 that we had a statistical package to work with.
In the mid-60's I realized that I was able to "plant" sports page stories on every major event where the stats were seen through the electronic eyes of the hardware. If I could give the idea away for free, could I also sell it? So I took the idea to the L.A. Times sports editor. He turned it down. "I don't like statistics features," he said.
Okay. There was still a second newspaper in town, the L.A. Herald/Examiner, where the then managing editor, Don Goodenow, said, "Great idea!" and bought it immediately. From there it was sold to the Long Beach Independent. Managing Editor, the late Miles Sines, also thought it was a novel idea to look at sports stats in an objective manner. He bought the weekly column and told the Des Moines Register and the Tribune Syndicate about it.
They sold it nationally in places like the Boston Globe, Cleveland Plain Dealer, Minneapolis Tribune….. and I sold the Washington Post, Oakland Tribune, Chicago Today, and later the New York Times. From 1963-1971, "Bud Goode's Computer Corner on Sports" was seen in 36 metro papers on a weekly basis (not just football, but baseball, basketball, golf, Indy 500 - wherever there was reliable data).
Parenthetically, my dear wife, Betty, kept another newspaper syndicate check for a given season - $3.77. Syndicates have a funny way of keeping books and find ways to add "additional expenses," which are deducted from the author's receipts. Three bucks for a season's work! It doesn't seem fair.
Aczel: What was the most "catchy" statistic you kept in that early period?
Goode: In looking at football I created a correlation matrix and used several columns as criteria - points scored (column 43), points allowed (column 44), winning margin (column 123), and points by offense only and defense only. In the beginning I had not "transformed" many (or any) of the variables to create efficiency measures. So when I first was able to produce a correlation matrix, I felt a great deal of pleasure (power!). Here was a 70 x 70 matrix of correlations, and among the largest correlates with winning margin I found the interception.
So I walked into Art Linkletter's office (Link was one of my three bread and butter clients in showbiz) and knowing Art to be an avid sports fan, I asked, "Which is the most important passing stat?"
He replied, "The bomb!"
"Wrong," I said. "It's the interception!" And I proceeded to explain the correlation matrix.
Sometime later, when I began using the BMDP statpack, and transformed a number of variables, the "bomb" - which measures yards per pass attempt - emerged as the significant variable on both offense and defense. I apologized to Linkletter at lunch. We see one another on his birthday almost every year, and laugh about the anecdote.
Aczel: Now, you indicated that NFL teams contact you regularly. Which teams and for what kind of data?
Goode: Twenty-one of the 28 teams have been clients at on time or another, Coach Knox has been the longest lived (Rams, Buffalo Bills, Seattle Seahawks, and Rams again this year). Coach Paul Wiggin, who was my contact at the Saints, and now at the Vikes, called one day and said, "Bud, I'm looking at the stats for three teams. Do you have team A, B, and C as clients?
"Yes, Coach."
"Bud, you're changing the game."
It may be true. The short passing game, which I have long thought filled the role of a running play, does add predictive and explanatory variance. So I encourage my clients to use the short ball control game and consider these dump passes as running plays.
Aczel: What particular NFL statistics did you come up with or originate that a fan would recognize?
Goode: First, we must consider the errors of omission and commission made by radio/RV analysts. For example, fans recognize that turnovers are important. They may not know, however, that the interception is more than twice as important as the lost or recovered fumble. The reason: When a team is leading they tend to run the ball more to move the clock and protect the lead. So there are more interceptions when a team falls behind and goes to the air to score in the shortest possible time.
By comparison, the fumble is linked to the running game, and the running game is linked to the team with the lead. So it is not so much discovering or creating new stats. Determining the importance of each stat vis `a vis our three key criterion measures adds understanding.
Creating unit values, for example, is one way to measure the importance of a stat. Each variable contributes some point value to the offense, defense, and winning margin. The unit value is computed by multiplying the correlation coef. by the ratio of the two standard deviations. It shows the impact of a unit change in the independent variable on the criterion measure. In this sense, the interception is worth (rule of thumb) five points in the winning margin, while the lost or recovered fumble is valued at two-three points (depending on the year).
Indeed, teams have gone to the Super Bowl and ranked last in the league on lost or recovered fumbles (Dolphins are one example).
When Vince Lombardi joined the Packers the run/pass ratio almost doubled in his first season moving from 1-to-1 (one running play for each passing play) to almost 2-to-1. And the Packers were winners.
The stat that Sports Illustrated emphasized in their 1974 profile, yards per pass attempt, is my favorite Goode stat. Joe Marshall, once pro football editor at SI, who wrote the piece, said, "Bud wants his tombstone to read, 'Here lies Goode. He told the pro football world about the importance of Yards per Pass Attempt.'"
When the Saints were a client, they improved from 4.3 yards per toss in 1977 to a league-leading high of 7.1. Adding one yard per attempt increases the winning margin by 3 ˝ points per game (unit value). So the Saints' increase in their winning margin (3.5 x 2.8) should add a theoretical 9.8 points to the margin (they actually moved from -7.4 to +.6, a change of 8 points)… and went 8-8 for the first time in their long history. (See the New Orleans Saints charts.)
The qb sack is also a stat which I emphasized and which fans are familiar with. One sack is worth 3 points in the winning margin (plus or minus a fractional change from year to year). This is a heavy impact on the margin and probably makes the sack the most underappreciated by the fans.
When the Green Bay Packers pay $14,000,000 to Reggie White, the ex-Eagle defensive end, they are getting a player who earns 16 sacks in a 16-game season (or one sack per game). This sets the performance limit for the best man at the position. Some of my research shows results that are counterintuitive. As an example, the correlations between fumbles and the criterion measures are smaller than the correlations between interceptions and the criteria. The reason: The lost fumble correlates with the number of running plays. And since running the ball (a clock-eating stat) correlates with being ahead on the score-board, the fumble, for a strong running team, may be a sign of strength.
Aczel: What types of statistical techniques do you use?
Goode: We do a weekly summary of stats. We use basics like the mean, median, trimmed mean, standard deviation, coefficient of variation, maximum, minimum, range, correlation and correlation squared (can be interpreted as a percentage relationship between the dependent and independent variables), the significance of the difference between two means (t-test). Other reports compute a correlation matrix, regression equations using Monte Carlo simulation.
When there are rules changes which impact these descriptive stats, the changes are reflected in the correlations - when the range increases, for example, as a result of clock changes (more plays per game), the range and standard deviation change.
Aczel: Have you tried to correlate winning records with the franchise data levels?
Goode: Franchise data level? I'm not sure I understand.
Aczel: Well, which teams do the best and the worst?
Goode: The St. Louis Cardinals (then Phoenix Cardinals and now Arizona Cardinals) have historically been weak on defensive passing efficiency (defined in a statistical sense by variable 83, opponent yards per pass attempt allowed). Passing efficiency on offense and defense is one of the vectors in a seven-dimensional space. If a team is consistently weak in an important area it is impossible to hide the weakness from the scouting report or from Goode's computer-added stat analysis. The weakness shows up.
In their wisdom, the Cardinals, puny on pass defense, in one recent season, drafted a punter, not a defensive back. In addition, after three seasons of continued improvement under their recent coach, they fired him (yet Cards were improved every year under his direction).
Aczel: What kind of "errors" do you have to deal with regularly in the data you get? How have you resolved those?
Goode: I have two editing programs that clean the data. It takes times, but most of the changes are caught.
Aczel: Are there better and worse sources of data?
Goode: Data come from Elias Sports. In the beginning, data for college football came from NCAA and other sources.
Aczel: Are there statistical lessons for students to learn from your sports experience?
Goode: Absolutely. It is the main reason I have been a volunteer teacher at Belmont. Belmont is the largest high school in California, 90 percent Latino. I have written the coaching LEGENDS in Spanish so the non-English speaking students can understand the reports.
My approach is to teach them introductory statistics tied to a hot button subject (sports - making the Monday Night Game of the Week, for example, a homework assignment). Two Hispanic students entered the North American Rockwell Annual Computer Sciences Competition this semester and won one of the 15 awards. This is the first time in the 70-year history of Belmont that Latino kids entered and won anything.
As a result the school and the L.A. Unified School District/Industry-aided group have funded the program. They are buying us a new 486 with wings, and I've asked BMPD to load their package (done) that we will begin using this fall.
They learn that statistics are ideas about numbers: First, the idea of an average; second, the idea of hot and cold variation around the average; third, the idea of the difference between averages; and finally, not a difference, but a likeness - how are two variables associated, linked, correlated?
From these concepts they learn that there is a correlation between education and income; an average performance in school and strong or weak performance; a change or difference in performance if they do more than 15 minutes of homework; a difference in the interest rate they pay the bank or credit card and the interest the bank pays them.
