After thinking about my last post
on the ability of baseball managers to properly leverage their best relievers, I realized that I had been unjustly critical of managers because of errors I had made.
The first error was in only looking at a half-season's worth of data because managers
are most likely unable to know what they have in the bullpen until relievers rack up enough appearances to give a clear picture. To correct this, I looked at the past four seasons (2002-2005) to see how managers do in a full season and also to see if any teams show a consistency in being able to put their best relievers in the highest leverage situations.
My second correction to what I did previously was to eliminate pitchers who pitched for more than one team because after transactions are made for relievers, sometimes right before the trade deeadline, managers feel pressured to use the newly acquired relievers in high leverage situations without thought to how good the new pitcher is in relation to the relievers the team already possesses. This leads to low wins above replacement level (WXRL) and high leverage scores, which is not indicative of the pitcher's true talent level. Eliminating these sorts of players also disregards the teams who do benefit from the relievers acquired mid-season, but the effect of that was something I was ready to absorb.
Also, trying to find a overall correlation like I did was an exercise in futility because leverage scores are not uniform for every team and so the results are skewed and end up looking worse than they really are.
However, everything else remained the same. I again looked at the correlation between relievers' WXRL and Leverage scores (only relievers who made no starts were considered), both Baseball Prospectus
statistics. The correlation between weighted on-base average (wOBA) and Leverage is also something I looked at, but I had already been convinced teams give no thought to the batters coming up in the inning and only care about what number inning it is when deciding which reliever to signal for out of the bullpen.
As for the results, I will be listing the top five and bottom five teams in terms of correlation between WXRL and Leverage of each year from 2002-2005 in reverse chronological order.2005Top 5
1. New York Yankees .995
2. Texas Rangers .953
3. Los Angeles Angels .926
4. Cleveland Indians .925
5. Minnesota Twins .920Bottom 5
1. Detroit Tigers -.311
2. Atlanta Braves -.115
3. Pittsburgh Pirates -.109
4. San Francisco Giants .203
5. Boston Red Sox .3172004Top 5
1. New York Yankees .934
2. New York Mets .920
3. Boston Red Sox .919
4. Texas Rangers .913
5. Florida Marlins .903Bottom 5
1. San Francisco Giants -.236
2. Colorado Rockies -.183
3. Cleveland Indians -.111
4. Chicago Cubs .008
5. Kansas City Royals .0712003Top 5
1. New York Yankees .990
2. Houston Astros .931
3. San Diego Padres .904
4. San Francisco Giants .893
5. Minnesota Twins .834Bottom 5
1. Kansas City Royals -.630
2. Philadelphia Phillies -.331
3. Colorado Rockies -.229
4. Boston Red Sox -.004
5. Cleveland Indians .1492002Top 5
1. Detroit Tigers .983
2. Atlanta Braves .970
3. New York Mets .928
4. Oakland Athletics .913
5. Chicago White Sox .902Bottom 5
1. Tampa Bay Devil Rays -.160
2. Chicago Cubs -.145
3. Washington Nationals -.003
4. Texas Rangers .172
5. Kansas City Royals .454
A few teams reappear on this list, most notably the New York Yankees who had the highest correlation three of the four years, but do not fall too much in love with their high correlation because two years it was a result of having only three relievers to look at after eliminating the relievers who made spot starts or were acquired from other teams.
One of the most interesting teams to me is the 2003 Boston Red Sox because their correlation between WXRL and Leverage matches what actually happened during their season. 2003 was the first season Theo Epstein was the general manager and had a plan to have a bullpen where the manager was suposed to use relievers based on pitcher-hitter matchups, handedness, and other strategies which would have maximized the bullpen. Unfortunately, the Red Sox manager was Grady Little, whose pedestrian IQ rendered him unable to process so much information at the same time and this was a failed experiment. The plan also suffered from having poor relievers to begin with. The overall correlation of -.004 showed how badly the plan failed.
A pattern I did notice was that the teams with the highest correlation are also the teams with good relievers overall while teams who have a couple good relievers and a couple bad ones do not have a high correlation. What I took this to mean is that managers do fine when the relievers they rely on are doing well, but when a closer is doing worse than his set-up man then managers are unwilling to allow the two players to trade places, perhaps trying to avoid bruising a closer's fragile ego.