In 2003, the Michael Lewis book, Moneyball, opened the eyes of a lot of baseball fans, insiders, and analysts as Lewis described the thought processes of GM Billy Beane and the 2002 Oakland Athletics in their quest to remain competitive in the American League despite a payroll significantly less than the Yankees, Red Sox, and Angels of the world. By valuing offensive traits such as on base percentage, slugging percentage, and walk rate, Beane was able to identify players that could be had at a tremendous value to the team through the draft, trades, and free agency. Using this philosophy of valuing skills many teams overlook (Moneyball emphasizes runners moving from station to station via hits, walks, etc; getting on base is valued more and expected more than moving runners along at the expense of outs), Oakland was able to make the playoffs every year from 2000-2003 and again in 2006.
Fast forward to 2005, and using a variation of smallball (sort of the anti-moneyball) called Ozzieball, the Chicago White Sox swept the Houston Astros to win the World Series. In smallball, teams use the hit-and-run, stolen bases, sacrifice bunts and hits, productive outs, and runner advancement to score runs whenever possible, and rely on quality pitching to always keep their teams in games. The 2005 White Sox finished with the best record in baseball, and a team 3.61 ERA on their way to the title.
Needless to say, the debate over which style of play is better has just begun - with countless numbers joining each side. Or is having a little bit of both the best? In 2007, teams such as the Boston Red Sox, Oakland Athletics, and the Cincinnati Reds continue to use the Moneyball philosophy, while others such as the Los Angeles Angels, St. Louis Cardinals, and Seattle Mariners rely less on the homerun and extra-base hit and move runners along however they can.
Looking at 2007, the White Sox are now one of the worst teams in baseball and the Athletics are struggling to reach .500 in a division that has had two teams run away from them. So while both teams that experienced such great success with their respective style of play struggle this season (something that was probably expected due to natural regression, especially for the White Sox), other teams using the different styles of play have risen up to take their places and are primed for the postseason.
So I thought an interesting study might be to look at 2007 offenses (pitching at another time) and see which teams are using the two separate paths, which teams utilize some of both, and how those correlate with winning percentages across Major League Baseball.
Then the question becomes how do you define teams using Moneyball, and how do you define teams using Smallball?
On the ESPN website, Rob Neyer concocted a fun little tool called the Beane Count (after Billy Beane) to tell which teams are playing the Moneyball style using hitting and pitching numbers that are reflective of that philosophy. Using reverse rotisserie style (the team in first place in a certain category gets one point, the last team in the NL would get 16 points, for example), he uses total HR and total BB for team offenses, and total HR allowed and total BB walks allowed for team pitching staffs to rate teams who at least unintentionally use Moneyball tactics the most. As of Sept. 3, Boston, Cleveland, and Oakland lead the AL, while San Diego, Colorado, and Cincinnati lead the NL.
What I tried to do was take that Beane Count concept and expand it a little bit, run the data for only offenses, and create an alternate set of comparative statistics for smallball stats. What I ended up with was a set of data that I call Money Count which ranks each team’s HR, BB, OBP, and ISO (Isolated Power; SLG-AVG); and Small Count which ranks each team’s Stolen Bases, Sacrifice Hits, AVG, and POP (productive out percentage; productive outs being any out that advances a runner with the first out or scores a run with the second out). I then ranked each team’s offense by these four categories the same way Neyer does in Beane Count, reverse rotisserie style.
The results for the separate spreadsheets are below (click on the link to enlarge, numbers are as of 9/2):
And here is the chart with each team’s winning percentage as of September 2:
From the two charts comparing moneyball to smallball, you see a few expected and a few unexpected things. In Money Count, slugging, slower teams like the Yankees, Red Sox, Reds, Indians, and A’s all show up in the top 12 or so. Similarly, faster, less powerful teams such as the Angels, Mariners, Cardinals, and Dodgers all show up in the top 12 of the Small Count chart.
What is interesting to see is teams where it is apparent that they have a combination of both sets of offensive traits. Teams such as the Phillies (with Howard, Rollins, Utley, Burrell, Victorino, et al), the Mets (Reyes, Wright, Beltran, Delgado, etc.), and the Devil Rays (Crawford, Pena, Upton, Young, Iwamura) have all developed players and qualities of being able to succeed offensively on both sides of the debate. They can beat teams with speed, out production, and the one-run inning; or they can mash the ball, reach base in a number of ways, and pile on the runs.
So looking at these two charts and then looking at team’s winning percentages. Can we compare the two and expect them to give us a significant correlation towards winning? Can we look at just 2007 offensive numbers and ask them to tell us how teams win or which way will contribute more towards a team winning? Offensive numbers from just 2007 is still a small sample size even when you are talking about five months worth of data - so a model built around 10 or 20 years of data would be better, but not having access to a database like that has limited me to 2007.
With the help of the brilliant Dr. Jared Benge, we ran multiple statistical tests to try and prove or disprove the theory that moneyball or smallball can correlate to or be a predictor of wins and winning percentage. What we found looks like this:
Trying to directly correlate moneyball style or smallball style to winning percentage proved difficult. Neither philosophy proved to be statistically significant based on 2007 numbers, but with the goal being p<=.05, Moneyball came in at .149 while Smallball came in at .912. Again, not enough data to make it significant, but Moneyball clearly comes closer to the goal of correlating to winning percentage.
So what we decided to do next was run a regression model to try and determine the explained variance of each of these philosophies to winning. The way it was explained to me was to think of a win or winning like a whole pie. By running a regression test for each style of play, we should be able to see what percentage of the variance or outcome (read:win) comes from each style. In doing this, we found that Moneybal received a variance of .073 or 7%. Smallball’s was essentially zero. The conclusion from this is that using this small sample size, 7% of a win can be explained by Moneyball, while Smallball essentially can not explain any part of it.
The bottom line is that regression is used to try and determine how much a predictor (the two different styles in this case) predict an outcome (which we want to make winning percentage). Trying to use either of these sets of data to predict winning percentage proves to be futile, and we see very little significance from Moneyball and even less from Smallball.
Our hypthesis is that over 10 or 20 years, if these numbers remained somewhat constant, Moneyball would eventually become statistically significant in determining a higher variance for winning percentage, or in predicting wins for a team. I will save that project for another day, but if someone wants to undertake the legwork, please feel free.
But in looking at individual teams, a pattern begins to show when comparing teams that populate high spots in both sets of rankings. Some of the best teams in the majors (Philadelphia, New York Yankees and Mets, Detroit, etc.) seem to have grasped an overall offensive approach: one that serves them well when needing to play a specific type of game that the circumstances demand.
So the debate will continue to go on until further proof can be produced. Perhaps someday we will undertake that. Someday we will collect all the data, modify our Count charts to more exact specifications, and run more tests. But for this year, the slightest of edges is given to Moneyball. For whatever that’s worth.
__________
If you want the full set of tests run for this project, let me know and I can find a way to get them to you. RK
