Verbatim: Stats guru Jeff Sagarin (Part 1)
Jeff Sagarin is one of the Godfathers of data-based sports projections, having started earning money for his work in the early 1970s after graduating from M.I.T. The New Rochelle, N.Y., native has provided rankings to USA Today since the mid-1980s and also has provided advisory services to multiple professional sports franchises. His Sagarin Ratings were one of the rankings used in college football’s BCS calculations and the NCAA tournament committee also has access to his work.
SI.com was able to catch up with Sagarin last Friday for a captivating conversation. In Part 1 of the discussion, Sagarin provides depth and color to how he got started in his profession. Part 2, more about current-day applications of projections in sports, will run on Tuesday.
SI: How did you initially get interested in sports statistics?
Jeff Sagarin: I probably got real proficient with arithmetic in elementary school. I was born in 1948, started kindergarten in the fall of 1953. So in the mid-50s, I’d say ‘Let me figure out this batting average. What if a guy got 106 hits in 279 at bats?’ stuff like that, and I would do long division with the decimal points, for fun. You learn, you know?
So I just used to play with numbers to just relate them to sports. I remember I started entering the New York Daily News — they used to have a weekly football contest back in the late ‘50s, that’s when I discovered it — and there were 15 college games to pick the winners of and they had you fill in an actual score for each team. The way they judged it, though, was who had the most winners and who had the most wrong.
I would get 12 right or 13 right and some grandmother from Brooklyn would go 15-0 because she liked the names of the teams. It was very frustrating as a 10-year-old to have that happen. There was a particular week where I went 13-2, it would have been the fall of 1961, and on the same day, I believe, Texas, which was undefeated and was at that moment ranked No. 1 in the country, lost [6-0] at home to TCU. I think they had three, four or five drives to where [Texas] had 1st and goal inside the five and they couldn’t score, and I had picked them.
My other loss that day was a strange game. Syracuse was visiting Notre Dame and I loved that Syracuse team. That was the first team I ever followed starting in ‘59, when I was in sixth grade, and the New York City papers covered them because they were the bigtime team in New York. Ernie Davis was a senior in the fall of ‘61 and they visit Notre Dame and they’re ahead [15-14] and time for the last play of the game and Notre Dame lines up for let’s say a 53-yard field goal, which was pretty long in those days, and they miss it and Syracuse wins the game, and I picked Syracuse to win.
Except, the referees blew the whistle and said Syracuse committed a penalty and fouled the holder or the placekicker, I don’t remember which. A) It was a stupid foul because the guy missed the kick. B) it was an illegal call under the rules at that time because it actually occurred after the gun sounded. It’s different now, the rules have changed, but at the time they totally acknowledged [afterward] it was the wrong call, it should not have been called. Syracuse truly, inherently had won the game.
(Here is footage of that game, including the roughing the kicker penalty that gave Notre Dame a second chance: http://blip.tv/classic-notre-dame/1961-nd-vs-syracuse-501928)
SI: What happened once you got to M.I.T.?
JS: I played intramurals a lot, I watched a lot of games on TV. You know, it’s funny, I don’t really care about games anymore. I mean, I still like the math, but back then I really used to care. I loved the AFL and I was a junior when Matt Snell went off tackle and for the first time in history, the AFL had taken the lead in the Super Bowl game. And it was just overwhelming that the Jets won that game. I still remember anticipating the first Green Bay-Kansas City Super Bowl, and now, it’s nothing compared to that. It was magic then, but maybe I’m just older now and that’s the difference.
There’s too much celebrity [now]. Singers coming into the press box, the halftime ceremony. The halftime ceremony at the first Super Bowl I think had maybe Grambling’s band playing down at the [Los Angeles] Coliseum. Everything is way too big now.
SI: At M.I.T., what prompted to start playing around with projections as far as sports, and why sports rather than another area?
JS: I’d been doing that for fun since I was younger. I’d probably been fiddling with formulas since I was maybe 10, 11, 12. As I got older, I got slightly better ideas. I first computerized what I was doing maybe in ‘71 or ‘72.
1972 was the first time I ever got paid for doing anything. Pro Football Weekly put out this little side thing, a tout sheet called Insider’s Pro Football Newsletter (Here’s an awesome newspaper ad for it: http://news.google.com/newspapers?nid=1356&dat=19720903&id=IIVPAAAAIBAJ&sjid=ZwUEAAAAIBAJ&pg=5146,562792). I was like the “handicapper” for them and I remember there was this guy named Charles Fujiwara who was one of their customers, and they’d forward me mail sometimes, and this letter says, “This guy is awful, he loses all the time, but make sure he keeps doing the picks the same way. All I do is go against him and I’m winning a fortune. One week, I went 8-0 and he sent in a letter that said “The kid wiped me out this week. I went 0-8.”
SI: If I recall correctly, the first time you computerized all of this, you got a fairly odd result?
JS: There was a story when I came to Indiana, when it was still decks of cards in the late 70s, and I submitted the deck — back then, you submitted the deck over a counter [to be run on the mainframe]. One of the operators said, “Jeff, I think you have a mistake here. You’ve got Montana State rated as your No. 1 team in either basketball or football.”
I said, “What?” So I looked at the deck and somehow it got mixed around. It must have been dropped or something. I just assumed the teams would stay in the same order [in the deck of data cards], but the names had gotten out of order. Imagine taking a deck of cards and cutting it, and taking the first half and putting it at the end.
Back then, I didn’t have a check on that. Now, there’s no cards, but you need to line them — think of them as cards — and the team’s code number is on the line along with the name of the team. So even if for some reason I got the electronic cards out of order, it wouldn’t matter because the team code numbers are on the same line as the team names. But you learn from oddball things happening, how to protect yourself against accidents like that.
SI: Is it true you still code in Fortran?
JS: Yeah, what’s wrong with that? It’s a good language. Fortran is real good for doing mathematics and running it real quickly. I’m not doing Photoshopping or anything like that. I’m just running numbers.
SI: How hard was it for you to adapt your projection methodologies across different sports?
JS: It wasn’t that hard. Really, if you’re talking about Team A is playing Team B and they’re playing to a score, it’s really the same system. The difference would be a sport like tennis, where Roger Federer would beat me 6-0 and would beat you 6-0, but if you and I played, on average you’d beat me 6-1. He can only beat either of us 6-love. It would be different if they said ‘Why don’t we play until each person has won one game?’ Somebody has to win, but you keep playing until even the loser has won a game. So Federer might win by 6,000-1 against me and only 600-1 against you, which would show that you and I were different in ability.
SI.com: So as you’re playing to a score, as you continue to grow your data for that particular season, you just get more and more context for each result, and that’s essentially how you rate the teams?
JS: The thing I have discovered over the years — I run an adjunct program for my own interest that will run different ways of doing things, run the system up through Day 5 and make predictions for Day 6. Then run it through Day 6 and make it run predictions for Day 7, and then store how accurately it’s doing each day of the season. And it turns out that the most accurate method to pick the games is Bayesian. Even late in the season, you’re still better off using the method that does not throw away the starting ratings.
With things like the BCS — I’ll kind of be glad I won’t have to be involved [after this year] and worry about the political correctness, the “How could you rank so-and-so ahead of so-and-so? They beat them.” Well, when I’m not responsible for the BCS, I don’t have to worry. For example, last year, if you looked at my Predictor rating, everybody in the know knew that Alabama was better than Notre Dame, but in my politically correct column, I was forced to have Notre Dame No. 1. It was a joke.
SI: One of the biggest talking points [the NCAA] has continued to resist is including margin of victory in their own ratings systems, the RPI or whatever else they look at, in order to accurately judge which teams are better. Why do you think that is?
The NCAA wanted me to talk to them in October 1988, to their men’s tournament committee. I said, “Well, I don’t like to travel, so I’ll talk for free if you just let me talk over the phone.” They said, “No, we want you here.” I said, “Well, then you’re going to have to pay me. I don’t feel like traveling. And I don’t even have a car, so you’re not only going to have to pay for my trip, my plane tickets, but you’re going to pay me another $100 so I can pay a friend at 4 in the morning to drive me up to Indianapolis to take a plane to Kansas City, and then pick me up at 10 o’clock at night when I come back the same day.”
So they paid for the tickets and I think I got $500 out of it, and they gave me another $100 for my friend. So when I was there, I remember the Dodgers won a key game that night. I think I stopped at a bar on the way back to watch it. I’ve improved my system over the years. I couldn’t even reproduce what I did then. But we were talking about scores and the RPI and not scores, and they said “You know what, Jeff. We like that you’re using scores in your formula. Keep doing that. Because we cannot officially encourage scores, but we know. All of us know the sport and we know scores tell you who the better team is. We know that.”
Back then, there wasn’t really anyone else doing any ratings. There wasn’t real internet or anything, so they would look each week and see what I had in USA Today. They said, “You keep using those scores. We can’t do it, but behind the scenes, we’re looking at it and we know they’re better.”