“Moneyball” Analytics
Last month my manager asked my group if anyone would like to read the book Moneyball. I have been interested in baseball statistics since I was a child, so I volunteered to read the story of how Billy Beane, the Oakland A’s general manager, used the writings of Bill James, a writer and statistician, to help the A’s win games while keeping a low payroll. But, what amazed me is how I have lived my own personal Moneyball story.
In the early 1970’s I was a twelve year old boy who liked to play the game “Strat-O-Matic Baseball.” This game had cards for dozens of the best baseball players of all time. The cards were used to play a game that consisted of two parts. In the first part the historical players were drafted by the people playing the game. The second part consisted of using the drafted player’s cards to simulate a baseball game. Dice would be rolled for each batter and a table on each card would be used to determine what the result was for each at-bat. I wanted to know which players were the best ones to draft, so I used a simple formula to calculate how good each player was. The formula gave one point for a walk or single, two for a double, three for a triple, and four for a home run. This value was multiplied by the chance for that cell in the table being selected. Then all of those values were summed up to give the player a score. I did this for every card. It turned out that a player I was not too familiar with, Honus Wagner, had the highest score and a few others players my friends and I did not know well also had high scores.
I started picking these players in the draft phase and I started winning most of the games. This is not how I became a stats geek, this is just one of the things I did because I am a stats geek. So, in the 70’s, I acted similar to Bill James, the person who took the initiative to analyze statistics that no one else cared about and was able to generate knowledge and understanding about the subject of the data.
In 1985, I graduated college with a degree in computer science and started working on a software training program at GE Global Research. The training program had about a dozen people a year for three years, so there were about 36 of us in the program. The older people, who were on their later years of the program, had a softball team in the corporate league. A bunch of us in the new class wanted to play, but they only let one or two join the team. The rest of us were not good enough. So, the rest of us formed our own team.
We played the older team twice that season. The first time we played them they beat us pretty good. So, the second time we were to play they were boasting about how they were going to beat us again and even told us they were having a competition to see who on their team can hit the most home runs in their second win. We decided to put in a new rule for our team for this game called the “one strike rule.” This rule said that no one could swing at a ball until they had one strike. Most people played softball to hit home runs, and no one celebrated walks. But, the pitchers in this league had played all season knowing that people would swing at any pitches, so they would almost never threw strikes. We walked and got deep into counts which forced the pitcher to try and throw strikes, just like the Oakland A’s did in 2002. And we beat the team that said none of us were good enough to play on their team. Then a few years later our team, with a few recruiting additions, won the league championship, partially because of the plate discipline we obtained from situational use of rules like the “one strike rule.”
One of the roles I had on the team was “stats man.” I would keep the book on the game (recording every at bat). Then, the day after the game, I would send out an e-mail with the on-base-percentage, average, and slugging for everyone in that game along with season totals. So, in the 80’s, we acted similar to Billy Beane, the baseball general manager who used statistics and analytics to help his team win.
In the 1990’s, I started teaching a class on applied intelligent reasoning systems for the computer science department of RPI as an adjunct professor in addition to working at GE. I wanted to create an in-class exercise that would be interesting for the students. At this same time it was baseball season and a young Derek Jeter was up for a new contract next season. I thought it would be fun to have the class determine how much the Yankees should pay Jeter. I was able to obtain data on players batting statistics and salaries from the espn.com website. Using this we selected the best stats for comparing players. The two stats we determined were the best were on-base-percentage and slugging.
Interestingly, the next year ESPN added a new stat to their lists. The stat they added is OPS, short for on-base-percentage plus slugging. OPS is now a single stat many people use to compare players. The class appraised Derek Jeter’s salary by calculating his OPS, finding three other players with similar OPS who have recently signed contracts (with some preference for players at the same position, same age, and from cities with the same market size), and using these three salaries to calculate what Jeter should be paid. The three similar players had their salaries adapted higher if they had a lower OPS or lower if they had a higher OPS. There were a few other adaptations made. Finally, the average of the three adapted salaries was calculated at $5M. That year the Yankees decided to pay Jeter $5M– the same as our analysis.
I am still using this exercise as part of teaching the Edison A class on computer algorithms. One of our projects at GE Global Research used this same appraisal technique to automate the appraisal of residential property. The papers and patents on the automated residential property system have been sited over a hundred times. So, in the 90’s we went beyond what was done in the book Moneyball and created intelligent systems based on the analytics!
