by Keith Law
It is the nature of baseball culture to anchor on a number that’s presented as if it has actual meaning—even more so if the number is nice and round: 100 RBI, for example, or 20 wins. For decades, a pitcher had to win at least 20 games to even get due consideration for the Cy Young Award, regardless of how else he pitched. Between 1969, the start of the four-division era in baseball, and 2010, the year Felix Hernandez won the award with a won-lost record of 13-12, there were 68 starting pitchers who won the Cy Young Award in full seasons. (This excludes 1981, 1994, and 1995, when MLB teams played fewer than 162 games apiece because the billionaire owners didn’t want to pay the players more money.) Only 17 won fewer than 20 games in their award-winning seasons, even though the 20-game winners who did win those awards were frequently not the best pitchers in their leagues—just the guys who had the most help from their teammates.
The late Bob Welch’s win in 1990 is one of the most glaring examples of writers who vote on postseason awards and the Hall of Fame (a system so bad it only remains because it’s better than all the other possible systems), focusing so much on a win total that they ignored all the other evidence of how the pitchers actually pitched. Welch, playing for an Oakland A’s team that went 103-59 and eventually won the American League pennant, was credited with a 27-6 record that season, but wasn’t even the best pitcher in his own rotation:
Welch had five more pitcher wins and five fewer pitcher losses than Stewart did, but where’s the evidence that he actually pitched better than his teammate—that is, that he produced more value for the team by preventing more runs? Stewart threw 29 more innings and yet allowed fewer runs than Welch did. If we just make the basic assumption, valid for our purposes here, that a pitcher’s entire job is to prevent opposing teams from scoring, then Welch could not possibly have been better than Stewart in 1990, because to match his teammate Welch would have had to throw 29 more innings and “allow” negative six runs to un-score, or something like that. Welch had the season we remember, because of the award and because no pitcher has reached 25 wins in a season since then, but it should have been obvious, even to the phrenologists of the day, that Stewart was better.
If Welch wasn’t even the best pitcher on his own team, where did he rank in the entire American League? Well, the writers of 1990 were apparently about as dim as a 10-watt incandescent bulb, because one of the best pitchers in history had one of the best seasons of any starter in the last fifty years in 1990, right under their noses . . . yet finished second in the voting. Roger Clemens—you might have heard of him at some point—threw 228 innings and allowed just 59 runs, 49 earned, for a 1.93 ERA. His won-lost record wasn’t shabby at 21-6, but the gap in their ERAs means that for every nine innings Welch or Clemens pitched, Welch allowed one more run. That’s close to a run per start, with both pitchers averaging over seven innings an outing. I’m not relying on sabermetrics here, nor am I pointing to stats that weren’t available in 1990. I’m talking common sense: the pitcher who gave up way fewer runs probably did his job better. As it turns out, more advanced metrics, like Wins Above Replacement, mark Clemens’s 1990 season as the tenth-best pitching season in the last half century, yet it wasn’t good enough to beat Welch out because of those fancy, shiny pitcher wins.
Those A’s teams of the late 1980s produced some statistical comedy because they were so good offensively, yet people inside and outside the industry still didn’t quite catch on to the folly of the pitcher win. In 1989, Storm Davis had the season that first started the break in my mind between the pitcher win and a pitcher’s actual value when he went 19-7, tying for the third-highest win total in the league with teammate Mike Moore, despite a 4.36 ERA that was the seventh worst among the 39 pitchers who qualified for the ERA title by throwing at least 162 innings. The mere fact that a pitcher can be that far below the median and still have one of the league’s best won-lost records should have tipped someone off that the pitcher win was misleading. Instead, the opposite happened. We continued to see pitchers lauded and paid primarily on the basis of those records rather than their underlying performances, resulting in egregious awards and contracts such as these:
• Willie Blair, a thoroughly mediocre starter who had a 4.73 ERA through the first seven years of his career, went 16-8 with a 4.17 ERA in 1997 for the Tigers, signed a three-year, $11.5 million deal with the Diamondbacks that winter. He was so bad for Arizona, posting a 5.34 ERA with a 4-15 (not a typo) record in four months, that the Snakes dumped him off in a trade to the Mets that July.
• Russ Ortiz went 21-9 for Atlanta in 2003 while leading the league in walks issued with 102, then went 15-9 for the club in 2004 while walking 112, before entering free agency that winter. He signed a four-year, $33 million contract with . . . oh, hey, it’s the Diamondbacks again, having learned nothing from running into the Blair ditch a few years earlier. Ortiz posted a 6.89 ERA in 2005, then was even worse in six starts for Arizona the next spring before they released him, which at the time marked an unofficial record for the largest salary still owed (about $22 million) to a player who’d been released.
• Bartolo Colon won 18 games in 2004 despite a 5.01 ERA that was well below the American League average; the next year, he went 21-8, winning the AL Cy Young Award even though he finished 8th in the league in earned run average. Minnesota’s Johan Santana was actually the best pitcher in the league that year; he threw nine more innings than Colon and allowed 16 fewer runs, but with a 16-7 record he came in third in the Cy Young voting.
• Shawn Estes was on what appeared to be his final shot to remain a major-league starter when he signed a $600,000, one-year deal with Colorado during spring training of 2004. He went 15-8 for the Rockies that season, with a whopping 5.84 ERA, leading the NL in runs allowed with 133 . . . and the Diamondbacks signed him to a $2.5 million deal for 2005. Even accounting for the difficulty of pitching at Denver’s altitude, there was no looking at that 5.84 ERA and seeing a million-dollar starter there, unless you were so blinded by the won-lost record that you couldn’t accept how badly he’d actually pitched.
The market has shifted since these examples, as awareness of the worthlessness of won-lost records has permeated front offices and started to leak out into media coverage of baseball, but the mentality that a starting pitcher’s job is to win the game still exists among fans, players, and even coaches. This belief, which requires a starting pitcher to record at least 15 outs (five innings) so that he’ll qualify for the win as long as his team is ahead, is a major obstacle to the ongoing paradigm shift in pitcher usage that has starters working less in each start and relievers handling more of the work in later innings.
The flaws of the pitcher win stat run much deeper than its failure to reflect the modern game or to tell us how well the pitcher prevented runs.
The fundamental problem with the pitcher win is that it is an inherent failure of logic. It takes something—the team victory or loss—that is, by the definition of the rules of the sport, the result of the efforts of at least nine players on each team, and ascribes all of the credit or blame for each side to one man. If a company handled its cost accounting like this, the CFO would be drawn and quartered at dawn. The invention of the pitcher win was a sort of brain death of baseball statistics, and it took nearly a century for the industry to recover from this early misstep.
Think about what goes into a team winning a game, and what portion of that you might then assign to the starting pitcher. The team in question must outscore its opponent, and if you’re wondering if you just paid good money to get that kind of insight, I’m sorry to report that yes, you did. But outscoring the opponent itself has multiple elements.
First, you’ve got to score—the job of the offense, and only the offense. That’s half the game, by definition, and it belongs entirely to the hitters who appear in the game. That may include the starting pitcher if the game takes place in a National League park—but even if that’s the case, the starter will get, at most, one-ninth of the plate appearances in the
game, and is going to be the least effective hitter in the lineup, so his portion of the credit we give to the offense will be quite small. If we think of the entire game as a pie, cut into, say, ten slices, we just gave five of them to the offense, and the starting pitcher probably doesn’t even merit a bite.
The other half of the game is run prevention, but that, as we know quite well today but didn’t fully grasp thirty or forty years ago, is the result of the interaction of pitchers, hitters, and fielders, where apportioning credit is not as easy as it once seemed. I’ll get into some of these effects—what we think we know now about fielding, and what we still don’t know—in later chapters, but for now, I’m going to offer some conservative estimates on the subject. We know that fielding matters: if the hitter didn’t walk, strike out, or homer, then one or more fielders were involved in the play, and may have had the chance to convert the ball the batter put into play into an out. How much fielding matters depends a bit on the pitcher, who may be a high-strikeout guy or a groundball guy or a guy who makes fans cringe when he’s scheduled to pitch, but historical estimates of the effect of fielding on the outcome of a game have ranged from 10 percent upward. Even if we leave it there, that’s another slice of the pie gone, with just four out of ten slices remaining.
What’s left still can’t all go to the starting pitcher. Today’s starting pitcher rarely throws a complete game, with one or more relief pitchers—somewhat of a novel concept compared to baseball in the early twentieth century—finishing the game after he leaves. Not only does the starter therefore not fully control the pitching portion of the run prevention variable here, but he doesn’t even control how many runs he gave up: in the Enronian accounting of baseball, if a runner scores, the pitcher who allowed him to reach base is said to have given up that run, even if the event that let him score occurred after that pitcher left the game. Say that David Price starts a game for Boston, begins the ninth inning on the mound, and walks the first hitter, after which he’s pulled for closer Craig Kimbrel. Kimbrel immediately gives up a home run to the next batter, allowing the man Price walked to advance three more bases, from first to home. Kimbrel is assigned one run allowed, and Price is assigned one run. Whether that’s fair or not—hint: it’s kind of not—is a little beside the point here; the point is that Price didn’t give up that run entirely by himself. So of those four pieces of the win-loss pie remaining, the relievers get to eat some; maybe it’s one piece, maybe it’s three, but suddenly we don’t have very much left for the starting pitcher.
Even if the starting pitcher throws a perfect game—27 batters faced, 27 batters retired—he’d still have gotten some help from his defense. The most dominant pitching performances in history, such as Kerry Wood’s 20-strikeout one-hitter in 1998, still involved a little help from the fielders behind those pitchers. Giving those pitchers the full credit of a win or a loss is idiocy. Not only is it inaccurate, giving us the impression that the pitcher had more to do with the team result than he did, but it has the effect of reducing our understanding of the game.
“Win = pitcher pitched well” isn’t even true; you can get a win and still pitch poorly. Russ Ortiz (yep, him again) made a start on May 21, 2000, against the Brewers where went 6⅔ innings, gave up 10 runs—that’s one run for every two outs he recorded—and still got the win because the Giants scored 16 runs that day. It was at least the 34th time in MLB history that a pitcher gave up 10 or more runs and was still handed a win. (The record, for the morbidly curious, appears to be 12 runs, set by Gene Packard for the St. Louis Cardinals in 1918. The play-by-play data are only accurate from 1913 forward, and is available on Baseball-Reference.com’s Play Index.)
On May 26, 1959, baseball saw one of the most egregious pitcher losses in history, as Harvey Haddix threw nine perfect innings in Milwaukee for the Pittsburgh Pirates . . . then another perfect inning . . . then another perfect inning, retiring the first 33 batters he faced, only to have his offense unable to scratch a run across through their first thirteen times at bat. In the bottom of the 13th inning, third baseman Don Hoak (of City Slickers fame) made a throwing error on what would have been the first out of the inning, allowing the first Braves baserunner of the day. After a bunt and an intentional walk (of Hank Aaron), Joe Adcock doubled to deep right-center, ending Haddix’s no-hitter, shutout, and the game itself. Haddix surrendered the only run of the game, and therefore was given the loss. It’s not the worst by line score—four years later, Milwaukee lefty Warren Spahn went 15 innings and got the loss in a 1–0 defeat—but it might be the most heartbreaking.
So why do we still do it? Why do we persist in handing out a win to a pitcher in every game, and a loss to another pitcher, and pat ourselves on the back for a job well done? Statistics in sport can do one of two things for us: describe what happened, or interpret what happened. Pitcher decisions (wins and losses) do neither. They obscure the truth while adding no pertinent or useful information. The statement that such-and-such a pitcher “got the win” is the canonical tale told by an idiot, full of sound and fury, and signifying nothing. Yet they’re still there on the stat sheet, on sortable pages of pitching stats, and in just about any news coverage of a baseball game, where the winning and losing pitchers’ records will appear in parentheses after their names, or a trade or signing, where any pitcher involved is reduced to those two numbers, separated by the hyphen of ignorance.
One last note: You may hear the rationalization that pitcher wins are bad for evaluating a single start, or a season, but good for evaluating an entire career. It is somewhat true that a pitcher who sticks around long enough to win 250 or 300 games is probably a good pitcher, period. (Teams make dumb decisions, but letting a bad pitcher get that many decisions would be a whole new level of stupid.) However, using pitcher wins on the career level to evaluate performance merely compounds the errors of using them in single games or seasons. The factors that make pitcher wins useless in smaller samples do not just even out in larger samples because you want them to.
Take Bert Blyleven, who pitched most of his career for bad teams. Blyleven won 287 games, and for more than a decade he was denied entry to baseball’s Hall of Fame by writers who pointed to his failure to earn 300 pitcher wins in his career. Yet Blyleven was a remarkably unlucky pitcher, spending most of his career pitching for bad—specifically, low-scoring—teams, and that lack of run support meant he “lost” a lot of games that starters for other teams would have won. Blyleven pitched 75 complete-game losses, the second-most in MLB since 1957, one behind Hall of Famer Gaylord “it’s a hard slider” Perry. Blyleven made 40 starts in his career when he went at least seven innings, gave up two earned runs or fewer, and got the loss. In 35 of his losses, his team lost by just one run. In 41 of his losses, his team was shut out entirely. He didn’t reach 300 pitcher wins because his teams stunk, not because he didn’t pitch well, or because he lacked some sort of special pitcher woo that made his teams win when he pitched. And he nearly missed the Hall of Fame because of it, earning election in his fourteenth year on the ballot, one shy of the limit of fifteen years (since shortened to ten).
In a grand bit of irony, Blyleven, now a popular announcer for the Minnesota Twins, has been outspoken in his disdain for smarter baseball analysis. He said in 2010 that Hernandez didn’t have enough wins to deserve the Cy Young Award and has dismissed the use of advanced statistics as “cybermetrics.” Former ESPN color analyst Joe Morgan, one of the greatest players in MLB history, was an ardent critic of any attempts to use statistical analysis to value players, as well as the book Moneyball, in which author Michael Lewis chronicled Oakland’s attempts to find undervalued players by using such methods, spearheaded by General Manager Billy Beane. (Morgan also claimed on television that Beane himself wrote the book, but refused to read it before criticizing it.) Morgan was a sabermetric darling of sorts as a player, because he did so many things well on the field, including posting high on-base percentages, leading the NL in walks drawn twice, and stealing bases at a high
rate of success. So even beneficiaries of the new way of evaluating performance can’t escape the tyranny of traditional stats, whether they’re for pitchers or hitters, because batting average and pitcher wins and RBI are the way they’ve always looked at players.
3
RBI:
Baseball’s Unreliable Narrator
Branch Rickey, the general manager of the St. Louis Cardinals and Brooklyn Dodgers who is best known for signing Jackie Robinson to his first MLB contract, called runs batted in “a misleading statistic” in a well-known piece he wrote for Life magazine in 1954:
As a statistic, RBIs were not only misleading but dishonest. They depended on managerial control, a hitter’s position in the batting order, park dimensions and the success of his teammates in getting on base ahead of him.
Sixty-two years later, the RBI remains a useful tool for measuring offense in the minds of fans, writers, MVP voters, and at least one major-league manager (since fired), Atlanta’s Fredi Gonzalez, who told the Atlanta Journal-Constitution in March 2016:
I know in the stat-geek world RBI is not a big number, but it sure is. Because you can have all the on-base percentage you want, if you don’t have somebody driving anybody in, you’re not going to score runs.
Unlike most of the other “basic” baseball stats like batting average or home runs, the run batted in, or RBI, statistic only entered the baseball lexicon in 1920, the invention of New York Press writer Ernie Lanigan, who began tracking the stat several years earlier, along with the statistic we now know as “times caught stealing.” Lanigan and his editor, Jim Price, introduced the RBI into their paper’s box scores and later convinced the National League to include RBI totals in the league statistics. By 1920, it was an official stat for MLB, and over the years many researchers have gone back to tally and verify RBI totals prior to 1920. It was adopted as an accounting of something that happened, but if Lanigan saw the RBI as holding some greater meaning, it’s been lost to time. Of course, when he created the RBI, the Most Valuable Player Award didn’t exist, but RBI totals have long been used by voters as one of the main criteria; if you lead the league in RBI and your team makes the playoffs, you’re already a favorite to win the award.