Finding the Best Division in Baseball using BaseRuns

The best division in baseball is a title that is hotly disputed among baseball fans.  Oftentimes sportswriters claim the superiority of one division over another, however they rarely use empirical evidence to back up their claims, instead relying solely on anecdotal evidence.

There is, however, a way to find the best and worst divisions in an objective manner.  FanGraphs recently started publishing Team BaseRuns, a team statistic that gives a team's expected run differential and win percentage based on its underlying offense, defense, pitching, and base-running.  If you want to find more about BaseRuns, this article by Dave Cameron does a good job at explaining the basic methodology behind the method.  The system is also validated by recent research.

Now that we are familiar with BaseRuns, let's sort each division by expected win percentage:

AL East

Blue Jays - .530
Orioles - .513
Rays - .497
Yankees - .465
Red Sox - .441

AL Central

Tigers - .563

White Sox - .485

Indians - .481

Royals - .474

Twins - .463

AL West

A's - .626

Angels - .605

Mariners - .537

Astros - .458

Rangers - .380

NL East

Nationals - .565

Braves - .515

Marlins - .477

Mets - .475

Phillies - .434

NL Central

Cardinals - .555

Pirates - .522

Brewers - .513

Reds - .511

Cubs - .501

NL West

Dodgers - .576

Giants - .528

Rockies - .476

Diamondbacks - .426

Padres - .418

Next, let's sort the divisions by average winning percentage to find out the divisional strengths.  

1. AL West - .521
2. NL Central - .520
3. AL Central - .493
4. NL East - .493
5. AL East - .489
6. NL West - .485

This list is a good starting point, but unfortunately it doesn't paint the full picture.  The American League has been the stronger league for quite some time, and still is in 2014.  This is the AL's record against the National League, in both 2013 and 2014, as provided by Vegas Insider and Wikipedia:  (I include data from 2013 because the 2014 season is not yet complete.  Also I weight the two seasons equally (even though 2013 has twice as much data) because 2014 data is more relevant.)

2013: 154-146 - .513 

2014: 90-81 - .526

Averaging the two winning percentages, we can see that the AL going forward can be expected to have a .520 winning percentage against the NL, while the NL can be expected to have a .480 winning percentage against the AL.  By dividing these winning percentages by .500, we get a multiplying factor for each league. 

AL: .520/.500 = 1.04

NL: .480/.500 = 0.96

After multiplying each divisional average BaseRuns winning percentage by the League Multiplying Factor, this is what we find:

1. AL West - .542
2. AL Central - .513
3. AL East - .509
4. NL Central - .499
5. NL East - .473
6. NL West - .466

Clearly, the AL West is far and away the best division in baseball, despite being home to two of the worst teams in baseball, the Rangers and the Astros, because of the A's and the Angels, baseball's two best teams playing in baseball's best league.  

In addition, baseball's best division got even better yesterday with a blockbuster trade between the A's and the Cubs, sending talented pitchers Jeff Samardzija and Jason Hammel to Oakland for prospects.  

GO/AO ratio and HR/9 (Part 1)

One of the best predictors of Earned Run Average (ERA), Expected Fielding Independent Pitching (xFIP) uses a pitcher's Fielding Independent Pitching (FIP) and Home Run to Fly Ball (HR/FB) rate to provide an estimate of what the pitcher's ERA should have been with neutral defense, luck, and opponent quality.

Many people, including myself, believe that the quality of a pitcher can essentially be derived from a few basic statistics, specifically Strikeouts per Nine Innings (K/9), Walks per Nine Innings (BB/9), Hit by Pitches per Nine Innings (HPB/9), and Home Runs per Nine Innings (HR/9).  The rest is up to the defense.  Because these statistics are the only things a pitcher can control, it is unfair to penalize a pitcher for having a poor defense behind him, which is exactly what many more common pitching statistics like ERA do.

Does a pitcher really control all of those aforementioned statistics?  At first glance, you may think so, but consider that home runs are still batted balls.  The best power hitters in the league are the hitters that are able to hit lots of home runs without sacrificing their batting average by constantly flying out.  As a result, the best power hitters almost always sport high HR/FB rates.  The worst hitters, therefore, have low HR/FB rates.

This is the major reason why a pitcher's HR/FB rate is almost entirely random, and the variation of the statistic prevents pitchers from maintaining a steady rate from season to season.  In order to compensate for this statistical noise, xFIP was created, normalizing a pitcher's HR/FB rate to the league average.  This adjustment has made xFIP the best predictor of future ERA among the statistics available to the public, even surpassing actual projection systems like Steamer, ZIPS, and PECOTA.

Unfortunately, almost all minor league batted ball data is unavailable to the public, making it impossible to quantify HR/FB rate for minor league pitchers.  Indeed, the only statistic available is Ground Out to Air Out (GO/AO) ratio.  This has made it difficult to distinguish true minor league pitching talent from fluke seasons caused by irregular HR/FB rates.

For major league players, these two categories are easily distinguishable.  For example, in 2012, Oakland A's pitcher Jarrod Parker had a 3.47 ERA and a 3.43 FIP, on the basis of a 6.8% HR/FB rate and 0.55 HR/9.  The low HR/FB rate was clearly unsustainable, and as expected in 2013 his HR/FB rate jumped to a more reasonable 10.5%, right around major league average.  As a result, his HR/9 hiked to a staggering 1.14 home runs per nine innings, causing his ERA and FIP to elevate to unimpressive values of 3.97 and 4.40, respectively.

If Jarrod Parker happened to be a minor league pitcher, we would have been unaware of his unsustainable HR/FB rate, and would have falsely characterized him as a talented pitcher capable of a 3.43 ERA (based on his FIP).  In reality, with xFIP and HR/FB rates available for major league pitchers, any person familiar with saber-metrics could have told you following the 2012 season that Parker's 2012 was a fluke, and that he is more of a 3.97 ERA pitcher than a 3.43 ERA pitcher just by looking at his xFIP.

A little more than half a run in ERA may seen insignificant, but Parker was worth 2.2 WAR more in 2012 than in 2013 despite similar peripherals.  That is the difference between an All-Star and a borderline major league pitcher.

Clearly, a statistic similar to xFIP is necessary in order to determine the true talent levels of minor league pitchers.  With GO/AO being the only available batted ball statistic available for minor league pitchers, this seems like a difficult task.

It is common knowledge, however, that pitchers who induce more ground balls also allow fewer home runs.  What is not clear is whether the relationship between GO/AO ratio and HR/9 is significant enough to form a new ERA predictor for minor league pitchers.

In Part 2, we will find out whether or not this statistic, MiLB xFIP, is indeed possible.  Is there really a correlation between GO/AO ratio and HR/9?