Scioscia and Gardenhire . . . so farby Chris Jaffe
April 12, 2010
Last week my column was based on a book I recently wrote: Evaluating Baseball's Managers, 1876-2008. In it, I took the book's major analytical tool, the Tendencies Database, and extended my book by a year to look at the 2009 season.
You know what, though? While the Tendencies Database was the main tool used in my book, it wasn't the only one. It provided some interesting details on particular managers, but a different tactic gave a mighty nice overall assessment of managerial value. This is a little something called the Birnbaum Database, named for its creator, Phil Birnbaum.
The Birnbaum Database
Giving an appropriate explanation of this is rather difficult. To do it properly, I feel I'd have to go on for too long for purposes of this column. Going short could leave the reader hanging with some questions.
First, it isn't a perfect tool with which to evaluate managers. Perfection is impossible, but fortunately flawed doesn't mean useless.
The Birnbaum Database consists of five components. The two most important are a pair of algorithms designed to tell if a hitter or pitcher under/overachieved in a given year. For hitters, let's say you wanted to see how David Ortiz over/underachieved in 2007. Figure out his Runs Created for 2005, 2006, 2008, and 2009, and create a weighted average based on that to determine what he should've done in 2007. (In the weighted average, count 2006 and 2008 twice as much as the other years). Adjust for playing time and regress to the mean. It's the same thing for pitchers, only use another Bill James stat - Component ERA.
One season's results for one season tell you more about luck and random happenstance, but over a period of time the signal to noise ration improves dramatically. (For example, it turns out that Earl Weaver's teams were lucky pretty much every year he was in Baltimore. Funny how that works.) There are five parts of the Birnbaum Database in all, but those are the two most important.
There's one obvious problem for using this in modern times: to know how a team did in 2009, you need info on 2010 and 2011. In the book, the Birnbaum Database could only take me up until 2006, but now I have information through 2007.
There's one other problem. If a team had a great year in 2008, it will hurt them in 2007. Usually this isn't a big deal. If a great 1954 makes 1953 look bad by comparison, that's OK - they'll get all the credit back when 1954 comes around. In this case, though, we don't have 2008 or 2009.
I actually performed a check to see if the Birnbaum Database indicates managerial skill. Sure would be nice to have some evidence instead of merely my say so. I divided all games played into four categories: 1) those managed by skippers with at least 2,000 games at the helm, 2) those by 1,000-1,999 game managers, 3) 500-999 gamers, and 4) the remainder.
The theory was that if there's any correlation between luck and skill, longer guys should do better. If it's luck, there shouldn't be any clear pattern. Short version: the results verified my idea that the Database indicates managerial skill.
There was an unexpected result: The other three components also came back largely verifying the skill thesis. Two of these components compared expected run scoring to actual run scoring (using Runs Created and Component ERA for offense and defense respectively). The fifth was the team's difference from their Pythagenpat project W/L. Here's what the results (per 162 games) look like through 2009 (in other words this isn't the version in the book):
Group Hit Pit Pyth Offen. Pitching Total 2000s 4.08 5.92 3.99 1.98 1.65 17.62 1000s -0.56 1.86 -1.88 1.38 -0.06 0.73 500s -1.16 -1.64 -1.05 -0.95 -0.33 -5.13 499s -7.84 -15.53 -5.71 -5.78 -3.61 -38.46
To clarify: The first two components are the ones Phil created his algorithms for. The third is pythag, and the remainders runs scored/allowed vs predicted.
The 2000s are always on top and the 499s always worse. There is only one step out of place in the entire chart: between the two middle groups with Pythagenpat. Even with that component, there is an overall tendency toward indicating some degree of managerial influence.
I believe the first two categories indicate coaching, the final two on strategic ability. The Pythagenpat group I'm shakiest on, but the manager can have some influence, based on things like bullpen deployment. Frankly, the first two components are the only ones I'm really sold on, but I'm willing to go along with the others.
Applying to modern managers
Originally I figured I'd write a column about how all current managers rank in the Birnbaum Database as of now, but I realized something: That's really boring. Even worse, it's at least a little too redundant of what I said in my book. Tony LaRussa is still the best manager since WWII. Bobby Cox still has an incredible score with individual pitchers. And so on.
I'll just focus on the guys with the most interesting results. Most notably, two managers with fairly brief careers rank among the top 25 skippers of all time. Mike Scioscia comes in 23rd place at plus-442 runs and Ron Gardenhire is in 25th at plus-389 runs.
Both rankings sound a bit high - and frankly they probably are, as other factors (such as luck) bleed into the data. Scioscia has only managed for 10 years - with only eight in the database. Gardenhire has a mere half-dozen in the database.
I suppose the real question is how impressive their standings are. The Birnbaum Database is something of a rate stat. If someone has a great stretch in which the teams don't underachieve, he'll look great. Then again, the longer someone lasts, the more likely that a down season will exist. Thus someone with a short and spectacular career can rank very high indeed.
Let's look at it this way though. How often has a manager gone six or eight years and posted a score as good or better than what Gardenhire and Scioscia have recently done?
I'll start with Scioscia, as he is the senior of the two men. Perhaps more importantly, almost all his score comes from the two most important portions of the Birnbaum Database, the algorithms that focus on getting players to perform their best. He scores plus-352 in those categories, which amazingly is already the 14th best career mark in history.
Scioscia's overall score of plus-442 runs is the 75th best eight-year stretch in history. Given that it's the 23rd best career total in history, 75th is surprisingly low. It makes sense when you think about it, though, because of overlapping eight-year periods. Bill McKechnie, for example, personally had a half-dozen eight-year stretches over plus-442. So did LaRussa, Casey Stengel, and Al Lopez. Heck, Joe McCarthy had 11 such marks.
How many managers topped Scioscia's plus-442 eight-year score? In all, 19 did. They are (listed in order of their best eight-year score): Frank Chance (+823 from 1905-12 is the best of all time), McCarthy, Stengel, Bill McKechnie, Fred Clarke, Billy Southworth, LaRussa, Connie Mack, Miller Huggins, Al Lopez, Ned Hanlon, Weaver, Walter Alston, George Stallings, Ralph Houk, Sparky Anderson, Joe Torre, Bobby Valentine, and Cox.
Then comes Scioscia, narrowly beating out Billy Martin and - in a shockingly low finish - John McGraw. (In career value, McGraw is one of the five best, but he was just really consistent over his five-decade long career.)
That's pretty good company for Scioscia, obviously. He's one of the most respected figures in the game, and it isn't hard to find people saying he has the makings of a Hall of Fame career, but this just amplifies it. Of the 19 men ahead of him, 11 are Cooperstown managers, and three more will be as soon as they retire. Clarke and Chance would have decent cases if they weren't already put in as players.
The only other ones are Stallings, Houk and Valentine. They combined for only four eight-year stretches better than Scioscia's total so far (Stallings had two). Please note Scioscia himself could easily have two or three eight-year stretches at least as good as his 2000-2007 period. It all depends how the Birnbaum Database ranks his 2008 and 2009 seasons, once his algorithms have the 2010-11 stats. Given how well the Angels performed in those years, he could do it.
Actually, older managers have a key advantage over Scioscia in that they have multiple eight-year stretches to choose from. How well did they do in their first eight seasons? Based on that, here are the best managers in history in the Birnbaum Database:
Manager First 8 Frank Chance 823 runs Ralph Houk 505 runs Joe McCarthy 500 runs Sparky Anderson 485 runs BillySouthworth 465 runs Al Lopez 453 runs Mike Scioscia 442 runs
(Actually, Ned Hanlon scores plus-566 in the first eight years I have info for, but that wasn't the first eight years of his career. The Birnbaum Database only operates effectively when SH data is known, and Hanlon's career predates that.)
Chance really smokes the field, but Scioscia's score is pretty damn impressive. He's got the seventh best start overall and fifth best since WWII.
Looking just at the coaching components for a second, his score (as noted already) is plus-352 in his first eight years. That's the 32nd best eight-year stretch in history. Only 11 men have ever topped it: Billy Southworth, Joe McCarthy, Earl Weaver, Ned Hanlon, Walter Alston, Sparky Anderson, Gene Mauch, Casey Stengel, Jimy Williams (!), John McGraw, and Bobby Cox. That's pretty impressive company. Nine are or will be in the Hall of Fame.
Ron Gardenhire's score of plus-389 runs in his first six years is very impressive, but pythagenpat is his best component, with a score of plus-174 runs. That still leaves him with an overall very good score.
Also, while I definitely think luck plays a role in his pythag score, I don't think it's just luck. One factoid I found out while researching my book: He has the best record in extra-inning games of any manager I looked at. Now, happenstance plays a role in those contests, sure. That isn't the only factor, though. These games, especially in the 21st century, are bullpen endurance contests. Terrific bullpen work is the hallmark of Gardenhire's tenure.
In my book I went through a prolonged look at relievers who came to him, and by and large pitchers not only performed well for him but tended to do considerably better than one would expect of them. Like I said above, while the pythag component doesn't mean as much to me as the algorithm ones, I don't believe it's all just luck.
Anyhow, Gardenhire has a score of +389 in six years, which is the 64th best stretch in history. There are 20 managers ahead of him at least one time, more than were there for Scioscia. That's surprising, but I guess the smaller the year sample, the more fluxuation comes into play.
The 20 who topped Gardy's six-season mark are: Frank Chance (from 1906-11, he was plus-701 runs, the best ever), Casey Stengel, Joe McCarthy, Bill McKechnie, Tony LaRussa, Connie Mack, Mike Scioscia, Billy Southworth, Al Lopez, Sparky Anderson, George Stallings, Fred Clarke, Bobby Valentine, Ned Hanlon, Miller Huggins, Earl Weaver, Ralph Houk, Joe Torre, Walter Alston, and Bobby Cox. Again, almost everyone is either a Hall of Famer or a solid candidate.
Taking it a step further, let's look at how he compares to other managers in the first six years of their careers. Here's the leaders:
Manager First 8 Frank Chance 606 runs Sparky Anderson 492 runs Ralph Houk 415 runs Ron Gardenhire 389 runs
That's rather impressive. Given how well the Twins have done the last two years, odds are Gardenhire's score will go up. That said I don't think he'll maintain his fourth highest rank. As is, he's barely ahead of Billy Southworth (plus-379 runs), and Joe McCarthy (plus-365 runs), and those guys really pored it on in the ensuring years.
That said, if you throw out pythagenpat, his overall first-six-season score is barely in the top 200. Almost 50 different men have had a six-year stretch better than that of Gardy, including luminaries such as Gene Lamont, Hank Bauer, Lou Boudreau and Art Howe. Looking at managers only in their first six seasons on the job, Gardenhire is 17th best when pythag is put aside.
One random note I feel obliged to make: some might be surprised to see Ralph Houk's name pop up so often in this discussion, but it makes sense. He arguably had the greatest start to his career of any manager in history. He won 109 games in his first season, the most ever for a manager (Earl Weaver won 109 in his first full season, but he already had a half-season under his belt).
Houk won the World Series in his freshman and sophomore campaigns. He won the pennant in his third season on the job. Only one other manager ever went to the World Series in his first three seasons, Huhgie Jennings (and he lost all his Fall Classics). Then things went off-track for Houk. He went in the front office and though he eventually returned to the dugout, he never captured the old magic. He was well respected enough to manage 20 seasons without ever getting fired, an impressive achievement.
References and Resources
Phil Birnbaum's my main resource for this. He not only invented the Database and ran it a few times for my book, but he even updated for another year for this column.
For brevity sake, I left a lot out the description of the Birnbaum Database. (For example, I normalize all components for all leagues at zero, so the midpoint is always the same). A complete description is in my book.
History instructor by day, statnerd by night, Chris Jaffe leads one of the most exciting double lives imaginable; with the exception of every other double life possible to imagine. Despite his lack of comic-book-hero-worthiness, Chris enjoys farting around with this stuff. His new book, Evaluating Baseball's Managers is available for order. Chris welcomes responses to his articles via e-mail. Oh, and now he's on twitter.