It is an adage as old as time in any sport. What little piece of the game is the most important to fielding the most successful team? In football, there is rushing, passing and defence. In hockey there is offence, defence and goaltending. In baseball, the game can be broken down fairly simply into hitting, fielding and pitching. I wanted to find out which of these three factors was the most important to regular season success. Now I know that doesn't exactly mean winning championships, but without regular season success there certainly won't be any championships.
The first step in figuring out the answer to my query was deciding which stats would represent each of my three categories. For hitting, I went with wOBA. If it's good enough for WAR, it's more than good enough for me. For pitching I selected xFIP. I had to go with a stat that eliminated fielding, as fielding has its own section, and xFIP has proven to have more predictive power than FIP. For fielding there were no surprises, and I went with UZR/150. For the Y value in my analysis, I had to choose between Wins and Expected Wins, and decided to go with Expected Wins as it is a more representative of a team's actual ability. However, I ran the regression for both for comparative purposes.
I mined five years worth of data (2005-2009), compiling every teams wOBA, xFIP, UZR/150 and Expected Wins. I then ran a regression with my three representative stats against Expected Wins. I was rewarded with this formula.
Expected Wins = 2.66 + 495.44(wOBA) - 19.11(xFIP) + .52(UZR/150)
For those of you who are curious, my r-squared was .693. If I replace Expected Wins with Wins, the formula is fairly similar, the major difference being a coefficient of 471.13 for wOBA, and my r-squared is .615. Now, these r-squared values aren't great, but for the purpose of this research I felt it was good enough to move forward.
Alright, now let's delve into the formula. For comparative purposes, I think it is beneficial to see how much a given stat would have to increase in order to lead to the same number of wins. For simplicity, let's go with five wins. In order to increase their Expected Wins by five, a team would have to either:
a) Increase their wOBA by 10 points (.010)
b) Decrease their xFIP by 0.25
c) Increase their UZR/150 by 10
The best way to figure out which of these goals is the most reasonable would be to figure out which can be achieved for the least amount of money. This would be done by figuring out on average how much teams pay for each point in a given category. Unfortunately, it would take someone much smarter and more informed than me to confidently procure an answer to those questions. Instead, I am going to look at the variation in each stat, and see which stat is the easiest to increase by the given amount required to win five more games.
Here are the averages and standard deviations for each stat over the five year span I studied:
As we can see form the data, increasing by five wins in either wOBA or xFIP would mean increasing the team's stat by slightly less than one standard deviation. UZR/150, on the other hand, would require an increase of just less than two standard deviations. What this means is that relative to the other teams, it is more difficult to differentiate your team's wins by changing your fielding ability. Oddly enough, this is contrary to what is happening in baseball right now, as many teams are focussing on fielding as a way to win more ball games. Not to say this is a bad strategy, as it in fact may be the least expensive strategy.
Overall, I would opine that this is more food for thought than an actual strategy. It doesn't take into account the financial implications of increasing a given stat, simply how difficult it is to improve in a given stat relative to the other teams. Please leave a comment if you have any suggestions on how to improve the research, or if you have anything else to say.