January 12, 2005
Getting Inside the Numbers on the Dodgers' New Starter
For a contract that hasn't even been signed yet, Derek Lowe's rumored four-year, $36MM deal with the Dodgers sure has gotten a lot of press. Most of this has centered around two sides of one issue: Has Dodgers GM Paul DePodesta lost his mind and, assuming he hasn't, does that mean the rest of us have?
Let me be very clear about two of my primary assumptions: First, I think DePodesta is a very smart guy and that he has shown a tendency to be very bold in his short tenure as GM of the Dodgers; second, I don't think he's lost his mind.
The reasons people have found the Lowe contract so horrendously out of line with DePodesta are 1.) both its length and its amount, and 2.) the fact that it's being spent on a pitcher who hasn't just been trending downward, he's been spiraling. Given DePodesta's pedigree in Oakland and his willingness to trade players whose best qualities - according to the media - are intangibles, committing that much time and money to a player who has one - count them, one - good season as a starting pitcher on his resume and a few well timed outs in the post-season appears drastically out of line. Thus the conclusion that either DePodesta has lost his mind or he knows something that the rest of us don't. Since I'm assuming the Dodger GM is quite sane, I'd like to know what he was thinking.
On the surface, the Lowe contract is way out of line with the other major pitching contracts of the 2004-05 offseason. Look how Lowe's last two seasons compare to the rest of baseball's nouveau riche, using Clay Davenport's Translated Pitching Statistics (which adjust for park, league, and all-time):
Player Age W-L IP ER BB SO ERA BB/9 SO/9 ------ --- --- -- -- -- -- --- ---- ---- Pedro Martinez 33 37-14 472.3 136 112 474 2.59 2.13 9.03 Carl Pavano 29 37-21 500.0 185 91 298 3.33 1.64 5.36 Brad Radke 34 33-20 505.0 189 44 294 3.37 0.78 5.24 Matt Clement 30 32-22 453.3 176 178 353 3.49 3.53 7.01 Odalis Perez 27 29-18 451.0 184 92 270 3.67 1.84 5.39 Russ Ortiz 30 37-21 493.7 203 243 293 3.70 4.43 5.34 Kris Benson 30 24-19 357.7 165 98 212 4.15 2.47 5.33 Derek Lowe 31 26-28 450.3 218 156 238 4.36 3.12 4.76Lowe is the third oldest pitcher in the group; he has the worst ERA, the lowest strikeout rate, the third worst walk rate, and the lowest innings total except for Kris Benson. He's also the only pitcher with a translated losing record; while actual pitcher winning percentage is easily manipulated by usage patterns and run support, these translated numbers are much more immune to that kind of fluctuation.
So, as has been well publicized, on the surface, Lowe has more than struggled in Boston the last two years. But - as has been suggested - perhaps DePodesta is looking at things in a way which provides solid evidence that Lowe would find success with the Dodgers - more than he had found in Boston and certainly more than his horrible 2004 season.
Much of the speculation about Lowe's performance in a different venue has centered around his groundball-flyball ratio (GB/FB). Among qualified pitchers, Lowe had the highest GB/FB ratio in the majors in both 2002 and 2003 before finishing second to Brandon Webb last year. In fact, by posting a ratio of 2.87, 2004 became the first year since his brief rookie season in 1997 in which Lowe's GB/FB was below 3.00. He is without question one of the most extreme groundball pitchers in the game and - if historical data were available - would likely be one of the most extreme of all time.
In previous studies, GB/FB ratio hasn't been significantly correlated to a pitcher's performance with regards to runs allowed, but it is an excellent predictor of home run rate. The reasons for this are quite obvious (and are stated by Nate Silver in his article): with the rare exception, ground balls are never home runs. However, as Nate has also pointed out in a recent email exchange, "Groundballers have a notably higher BABIP than flyballers." Without getting into the entire argument of Defense Independent Pitching Statistics (DIPS), there is evidence that groundball pitchers allow higher batting averages with lower slugging percentages, a tradeoff that keeps their overall runs allowed very close to their flyball counterparts. Groundball pitchers also see a noticeable increase in double-play rates, as a quick look at BP's Double Play Rate for Pitchers report confirms.
Because of Lowe's extreme groundball tendencies and his low strikeout rates, changes in park and defense are likely to have a more significant effect on him than on other pitchers. The question then is, exactly how much can Dodger Stadium be expected to improve Lowe's numbers and does his style of pitching lend itself to Chavez Ravine more than other stadiums around the league?
Rather than looking at raw park factors, in Lowe's case it's more instructive to look at how Fenway Park and Dodger Stadium affect individual aspects of offense. These individual park factors are generated by looking at the totals in games at the home park versus totals in games on the road, totalling both home and visitor stats. For example, the number of home runs hit by both the Red Sox and their opponents in Fenway is divided by the total home runs hit in Red Sox road games, adjusted for playing time. Here are three-year park factors for Fenway and Dodger Stadium:
Park R H 2B 3B HR BB SO ---- --- --- --- --- --- --- --- Fenway 106.1 106.5 123.8 82.0 89.7 99.3 95.8 Dodger 89.5 94.3 79.3 49.5 109.6 93.2 105.2While it's commonly known that Dodger Stadium is one of the most pitcher-friendly parks in the majors, it's interesting how it achieves that notoriety. Dodger Stadium actually gives up 9.6% more home runs than a neutral park while Fenway - a hitters park - allows 10.3% fewer. The reason for Fenway's offensive tendencies and Dodger Stadium's defensive leanings is the vast differences in hit rates, excluding home runs.
This distinction could be the key to the Dodger's interest in Lowe. Simply adjusting Lowe's RA based on the Park Factors gives him an RA of 6.25 in 2004 if he had played in Dodger Stadium. Allowing 6.25 runs per game in Dodger Stadium certainly isn't worthy of $36MM, but there is another way to calculate runs allowed.
Because runs allowed in a season can be subject to the timing of hits and other key events in a game, players can have ERAs or RAs disparate from the individual statistics they allowed. Rather than simply dividing Runs Allowed by Innings Pitched, combining various rate statistics like hits or strikeouts per inning or AVG, OBP, and SLG allowed, we can calculate a more accurate predictive value of ERA or RA. Doing so with Lowe's individual numbers in 2004 reduces his RA from 6.80 to an expected RA of 5.04.
Another estimate of Lowe's 2004 RA can be calculated by looking at Lowe's more specific statistics and adjusting each of them by the specific park factors. This way, it will be clear how Dodger Stadium's particular run suppression characteristics relate to Lowe's style of pitching.
Lowe's individual stats for 2002-04:
Pitcher Year AB R H 2B 3B HR BB SO ------- ---- --- --- --- -- -- -- -- --- Derek Lowe 2002 787 65 166 30 3 12 48 127 Derek Lowe 2003 795 113 216 35 7 17 72 110 Derek Lowe 2004 748 138 224 41 4 15 71 105Adjusting Lowe's individual stats from Fenway Park to Dodger Stadium, they look more like this:
Pitcher Year AB R H 2B 3B HR BB SO ------- ---- --- --- --- -- -- -- -- --- Derek Lowe 2002 787 60 156 24 2 13 47 133 Derek Lowe 2003 795 104 203 28 6 19 70 115 Derek Lowe 2004 748 127 211 33 3 17 69 110Lowe's estimated RA in 2004 based on these statistics would be 4.62. Suddenly, instead of being a pitcher with a 6.80 RA, Lowe becomes - if not quite league average - much closer to an acceptable innings eater. But this does not mean that Dodger Stadium particularly suits Lowe's pitching style; if his estimated RA based on his statistics in Boston (5.04) is adjusted simply by the raw runs park factors, it comes out to 4.63 - almost exactly the same as the 4.62 RA above. Thus, while Lowe had an unlucky year in a tough hitter's park, Dodger Stadium's particular features do not especially lend themselves to Lowe's style of pitching.
The other major question of Lowe's transition to Los Angeles was the quality of the infield defense. Boston's infield glovemen were hardly worth the dirt they were standing on. While Kevin Millar, Pokey Reese, and Bill Mueller played well in the infield with Fielding Runs Above Average (FRAA) of 0, 4, and 6, respectively, the rest of the infield left much to be desired. Nomar Garciaparra (-6), Orlando Cabrera (-6), and Mark Bellhorn (-9) were definitely not in the lineup for their gloves. The Dodgers, on the other hand, were one of the league's better defensive infields. Adrian Beltre (13), Cesar Izturis (2), and Alex Cora (8) managed to compensate for Shawn Green (-10), saving more than a win over the course of the year just with their gloves.
Totaling the differences, the Dodgers' infield was 24 runs better than the Red Sox's in 2004. While on the surface Lowe would only stand to gain about 3-4 runs from the upgrade (assuming he pitched 200 of about 1400 innings), given his disproportionate number of groundballs, his share would be much larger. Dodger pitchers last year allowed 1807 ground balls while Lowe allowed 405 all by himself. Disregarding the possibility that the Dodger infield's 24 run advantage on the Red Sox is due to a better ability to catch the infield fly, Lowe would be expected to save about 5.3 runs in front of the 2004 Los Angeles infield.
However, the Dodgers have also revamped their infield defense, swapping out Beltre, Cora, and Green for Jose Valentin, Jeff Kent, and likely Hee Seop Choi. Valentin has averaged 6 FRAA over the last 3 years and, as defensive numbers from short appear to carry over to third, Valentin should be expected to post similar numbers in 2005. Kent has been all over the map defensively, but his last three seasons averaged out to around 6 as well. Choi has seen very limited playing time, but has shown himself to be about average. Assuming that Izturis maintains his play at short, the new Dodgers infield should actually be about the same as last year's with the loss of Beltre compensated for by the loss of Green. Thus, the better defense can still be assumed to give Lowe a healthy five run boost over the course of the season.
There is also the question of errors. Lowe's defense allowed 19 runners to reach on errors (ROE) behind him in 2004, yielding one of the league's highest unearned run totals. The Dodger defense, however, allowed only 29 runners to reach all season. Assuming those rates affected all pitchers equally, Lowe would be in line for about 4 errors instead of 19, another drastic difference. Those 15 runners are likely already accounted for in the earlier analysis of expected RA simply because ROE is not captured by any of the individual metrics used to calculate expected RA. They are the main reason Lowe's expected RA was so much lower than his actual RA and, while the Dodger defense seems more likely to return Lowe to his expected RA than Boston's, removing them a second time would not be accurate.
Lowe's performance in LA can be expected to improve for three reasons: the inherent run suppression of Dodger Stadium (rather than any specific park factor related to his pitching style), the fact that he was unlucky in 2004 based on his secondary statistics, and the improvement of the individual infield defenders in LA. As mentioned, based on his secondary numbers from last year, Lowe would have had an RA around 4.62 in Dodger Stadium. Adding in the expected 5.3 extra runs of improvement from the Dodger infield, Lowe would have an RA of 4.38 in 200 innings and an ERA slightly below that.
Finally, there's the question of an improvement from Lowe's performance independent of park and defense. While there have been few pitchers with his extreme groundball tendencies, his 2005 PECOTA projection is a nice place to start the discussion. PECOTA has Lowe pegged for a nice improvement in his ERA, but most of that is based on his rather unlucky 2004. His 2005 H/9, BB/9, SO/9, and HR/9 are all very similar to 2004 and, adjusting for playing in Dodger Stadium and calculating his RA based on his secondary stats, Lowe's projected RA is 4.66, again very close to the 4.62 above. For a very quick and unscientific estimate of his upside, let's assume he were to improve to his 2003 form; in that case, his estimated RA would drop to 4.12 with a little more improvement to come from the LA infield defense. Though that may sound impressive, an RA of 4.12 is just slightly better than league average considering his league and park.
Even considering the positive effects of the improved defense behind him and the specific offensive suppressions of Dodger Stadium, Los Angeles is still on the hook for $36MM for a pitcher with a projected higher than league average adjusted RA. The improvement expected as a result of the park and superior infield defense are not enough to overcome Lowe's steadily mediocre pitching. All things equal, Lowe's 2005 could look quite a lot like Jeff Weaver's 2004: not that bad, but certainly not worth the money. Unlike Weaver, Lowe will be keep being expensive for a long time to come.