March 12, 2003
The Grand Experiment
Converting Graves and Kim to StartersSure, the games don't count for anything, and as we've pointed out before, any team relying on spring training to make important roster decisions is likely in for a long season. But there are exceptions, and none more important than in Sarasota and Tucson, where Danny Graves and Byung Kim are bidding to join their respective teams' rotations.
Starting pitching has long been the Bowden Reds' weakest link; 30 league average starts out of Graves would go a long way toward placing them in the midst of a Great American Wild Card Race. The situation is even more critical in Arizona, where PECOTA is expecting a big decline in the Snakes' offense, and the lack of obvious short relief replacements makes the move a higher-risk gambit. If Kim can make the transition successfully, it's not unrealistic to imagine Diamondbacks finishing 1-2-3 in the league ERA race; anything less, and Arizona could have trouble defending its title in what should continue to be a highly-competitive division.
Moving a top-notch reliever into the starting rotation is neither a new nor a particularly unusual idea. The Indians and Orioles experimented with moving Hoyt Wilhelm into a starting role in 1958, which culminated in a nationally televised no-hitter. Bill Lee and Wilbur Wood began their major league careers as relievers; Goose Gossage received a one-year trial as a starter; Rick Aguilera logged 89 career starts. Derek Lowe, of course, made a successful conversion last year, which may have helped assuage any misgivings on the parts of Bowden and Garagiola.
What's interesting about this year's guinea pigs is that Graves and Kim are radically different pitchers. Both were born in the Far East and allow very few home runs; that's where the similarities end. Graves, statistically speaking, is a finesse pitcher--perhaps the most successful finesse closer since the beloved Dan Quisenberry. Kim is a fire-and-brimstone submariner, striking out a quarter of the batters he faces, sometimes at the cost of allowing others too many free passes.
PECOTA has something of a blind spot when it comes to pitchers who switch roles. There's no reliable way to predict ahead of time which reliever might be tested in the rotation; in its effort to be objective, it assumes that Kim and Graves will keep on keeping on in the bullpen. We can, however, use some of PECOTA's approach to evaluate the effect that the transition might have.
By applying a fairly stringent set of criteria, I was able to identify 19 pitchers who made a clean break from relieving to starting. In order to be included in the study, a pitcher needed to have:
Graves has been used as a reliever from the earliest moments of his professional career. Kim was a starter in high school and in international competition, but has pitched almost exclusively in relief since coming stateside. Therefore, it's important to select those pitchers who made as neat a break to starting as possible. Wilhelm, who made his transition in the middle of the season, did not make the cut. Nor did pitchers like Jimmy Key or Pedro Martinez, who got their feet wet with a year of relief despite having been groomed as starters. (I'm kidding about the BPR bit, of course, although Will Carroll might well be trying to lock down the Spaceman for a forthcoming appearance as you read this).
In order to evaluate the pitchers, I relied on their baseline runs allowed average (RA) as derived by PECOTA from their three most recent seasons of major league performance. The purpose of the baseline calculation is to remove luck from a forecast line. Gossage posted a 1.84 ERA in 1975, the season before his conversion to the rotation, but did so on the basis of an unsustainably low home run allowed rate while walking a batter every other inning. As the baseline forecast accounts for, he was not likely to sustain that performance going forward, regardless of the role that he fulfilled; it was only after his long year as a starter that Goose learned to be an efficient pitcher, turning his command into a strength as he reeled off a string of nine consecutive dominant seasons. Conversely, Lowe and Scott Garrelts were somewhat better pitchers than their actual RAs would indicate, and their success in their new roles should not have been a huge surprise.
Yes, we're using RA instead of ERA; it's a more reliable predictor, although that's a subject for another article. Everything has been translated to a neutral park and league, using an analogue to the ERA+ convention established by Total Baseball in which 100 indicates a league-average performance and higher scores indicate superior performances.
Table 1: Comparison of Expected and Actual RA+ for relief-to-starter converts
RA+ Pitcher Year Baseline Actual Delta --------------------------------------------------- Turk Farrell 1962 136 126 -10 Gary Bell 1966 119 110 -10 Wilbur Wood 1971 164 175 +11 Tom Timmermann 1972 127 110 -17 Bill Lee 1973 126 150 +25 Rich Gossage 1976 94 96 +2 Doug Bird 1976 112 97 -15 Calvin Schiraldi 1988 94 85 -9 Scott Garrelts 1989 162 148 -14 Greg A. Harris 1990 103 104 +1 Greg L. Harris 1991 131 169 +38 Bill Swift 1992 121 165 +44 Craig Lefferts 1992 94 93 -1 Kenny Rogers 1993 109 100 -9 Rick Aguilera 1996 114 103 -11 Willie Blair 1997 116 116 0 Darren Dreifort 1998 111 103 -8 Hipolito Pichardo 1998 103 91 -12 Derek Lowe 2002 188 181 -7 --------------------------------------------------- MEDIAN 116 110 -6
The first thing to notice is that, irrespective of the change in roles, a pitcher's previous performance was generally a very good predictor of his future performance. While the majority of the pitchers exhibited a slightly worse RA after making the conversion to starting, the correlation between baseline (predicted) and actual RA+ for our group of pitchers is a remarkably high .83. As we're going to discuss in a moment, there are certain characteristics that are associated with marginal advantages when a pitcher changes roles; that doesn't mean that a great pitcher suddenly becomes a poor one, or the other way around.
But what about those other characteristics? In addition to tracking RA, PECOTA generates baselines for several peripheral statistics, most importantly strikeout rate, walk rate, and home run allowed rate. Like the RA baseline, these metrics are normalized to a neutral league and park. Here are the correlations between those baseline metrics and the difference between predicted and actual RA+ (RA Delta):
Table 2: Correlations between RA Delta and peripheral statistics
Strikeout rate: -.46 Walk rate: -.20 Home run rate: -.39
Heck, even I'm a little bit confused at this point, so let's try a couple of off-speed pitches:
Table 3: Pitchers sorted by TTO%
RA+ Pitcher Year TTO% Delta ---------------------------------------- Bill Swift 1992 19.9% +44 Wilbur Wood 1971 24.8% +11 Bill Lee 1973 26.0% +25 Craig Lefferts 1992 26.1% -1 Tom Timmermann 1972 26.2% -17 Hipolito Pichardo 1998 27.0% -12 Derek Lowe 2002 27.6% -7 Willie Blair 1997 28.1% 0 Rick Aguilera 1996 31.0% -11 Greg L. Harris 1991 31.5% +38 Kenny Rogers 1993 31.7% -9 Greg A. Harris 1990 32.0% +1 Doug Bird 1976 32.3% -15 Turk Farrell 1962 32.5% -10 Gary Bell 1966 33.6% -10 Darren Dreifort 1998 34.6% -8 Scott Garrelts 1989 36.9% -14 Rich Gossage 1976 38.1% +2 Calvin Schiraldi 1988 39.5% -9 ---------------------------------------- CORRELATION -.48
The three pitchers with the lowest TTO% all performed much better than expected upon their move to the rotation. Conversely, most of the high-TTO% pitchers performed worse than expected. With a relatively limited sample, it doesn't qualify as categorical proof, but there does seem to be some relationship between the style of pitcher and his likelihood of making a successful conversion to starting. Pitchers who allow the ball to be put into play and let their defense do their work for them tend to do better.
It's always easier to follow statistical evidence with intuition than the other way around, but this result, if it's real, makes plenty of sense. It simply isn't possible to pitch like Calvin Schiraldi for seven innings at a time; the pitch counts rise too quickly, and the strain on the arm is too great. A pitcher like Schiraldi needs to change his approach when pitching for extended stretches in a way that a pitcher like Billy Swift does not. Inevitably, that means allowing a few more runners to reach base. If the pitcher is also vulnerable to the long ball, as Schiraldi was, his RA can rise very quickly as solo shots turn into three-run jobs.
Graves, who has a very low TTO%, and is very efficient with his pitches, is exactly the sort of pitcher who has adapted well in the past to a starting assignment. The news is a mixed bag for Byung Kim. While he's a strikeout-dependent pitcher, he's also become better over time at avoiding home runs and generating groundball outs, which reduces his combustibility if a few extra runners reach base.
The other part of the equation, of course, is how well a pitcher's arm holds up to the increased mileage. Because it's a self-selected collection of players, the group that we've chosen for this analysis doesn't do the question justice. Two of the better statistical comparables for Kim--Garrelts and Darren Dreifort--had Tommy John surgery within a few years of their conversion; whether we have any business comparing a sidearmer with overhand pitchers is another question entirely. Kim has cut down on his pitchers thrown per plate appearance in each of the last three seasons, and he'd be wise to continue to do so as he adapts to his new role. He is likely to be effective so long as his arm stays intact, but his conversion remains a high-risk, high-reward strategy.