MORE ABOUT THIS BOOK
The Book on the Book
THE GAME ON THE FIELD
PREFACE TO PART I
When Albert Spalding wrote one of the first histories of baseball in 1911, he deliberately sidestepped any discussion of strategies. Said Spalding, "The opinions of up-to-date, scientific experts so widely and so honestly vary as to what really constitute important methods that there is no intelligent hope of bringing them together."
If he were alive today, Al Spalding couldn't get away with that dodge for two reasons. First, virtually every fan has an opinion on strategies, and fans are not at all shy about sharing their wisdom. But second and more relevant to Spalding's point, today we often do have the tools to determine which strategies work and which don't. We can tell through computer analyses precisely how frequently a stolen base must be successful in order to help the offense. We can develop statistical models to determine how much money is too much to pay Alex Rodriguez. We can determine exactly--if only in retrospect--how large a budget any team must have in order to be competitive in the pennant race.
In other words, we can gauge what works and what doesn't.
Spalding's reticence notwithstanding, the merits (or demerits) of particular strategies have always fascinated fans, players, and owners alike. Spalding himself was not immune from this reality; witness a comment from the 1885 edition of his Official Baseball Guide regarding baserunning:
Each season's experience only shows more and more the fact that good baserunning is one of the most important essentialsof success in winning games ... Any soft-brained heavyweight can occasionally hit a ball for a home run, but it requires a shrewd, intelligent player, with his wits about him, to make a successful baserunner. Indeed, baserunning is the most difficult work a player has to do in a game.
The importance of baserunning was one of the principal tenets of the original "Book"--by which is generally meant the common understandings that have governed baseball strategies. "The Book" became fixed in the minds of the game's cognoscenti over a period of decades, although in fact it has undergone a constant process of reform and revision. Over time, entire chapters governing on-field play--on topics as diverse as the wisdom of the sacrifice, playing for the long ball, the arrangement of a lineup, expectations for the starter, the use of relief pitching--have been written, erased, and rewritten. Just as managers of the 1950s would have viewed strategies of earlier generations as dated, some of their assumptions seem unfathomable today.
Those changes may be driven by alterations in rules or playing conditions, such as improvements in the manufacturing of the baseball around 1910 that made for a harder product capable of being driven farther. They may also occur due to external demand: When Babe Ruth proved that fans loved a slugger (and would pay to see one), the long ball became a marketing device as well as a strategic tool. And as often as not, the on-field "Book" is rewritten due to simple copycatism. If one manager tries a new strategy and enjoys success, it is a virtual certainty that within a short time the idea will be widely pilfered, whether the pilferage actually makes strategic sense or not. For evidence, just look into the nearest bullpen.
It is one thing to say that strategies have changed through the years--driven by wisdom, popular tastes, rules, legal judgments, or other factors--and another to assert that the changes have represented constant improvement. "The Book" as it is commonly understood todayrepresents nothing more than the collective judgment of the game's managers and general managers. The very fact that it has undergone constant revision strongly argues that "The Book," in the sense of representing the statistically ideal way of logically playing the game, has never been "written." And if that is the case, why would one deduce that "The Book" is "written" today, that strategic knowledge has reached its apotheosis, or that improvement has even necessarily taken place in a straight line?
What we do know is that, more than ever, we are equipped to assess "The Book," to judge its legitimacies and weaknesses, and determine the extent to which big league teams do (or ought to) govern themselves by it. Which is joyous news to long-suffering fans of numerous perennially hapless teams: Maybe those bums of ours can't win, but at least we can figure out why not.
A side note: Because the data do not exist for it to do so, neither this book nor any other explores one potentially vital aspect of modern "strategy" (if strategy it be), and that is the impact of performance-enhancing drugs. To the extent it is a factor, that factor is miasmic. We know from 2003 testing results that between 5 and 7 percent of players were using some type of improper, if not always illegal, performance enhancement. We don't know for how long, nor do we know the extent to which (if any) that use has affected the game.
But simply because research into the impact of performance enhancements is not presently possible, we should not proceed as if those enhancements do not exist. In retrospect, the evidence of the impact of their use seems clear. For years experts have mumbled that there is "something different" about the modern game--generally this occurs in the context of the growth of offense during the most recent decade--and they have done so seemingly against all logical evidence. The flailing toward a theory at times sounds almost comical. Expansion? But why should expansion dilute only the pitching, while actually improving batting? Greater racial diversity? Oh, so what you'resaying is that minorities can hit but not pitch? Fitness regimes? As if pitchers don't work out. Each of those pseudo-theories tumbles one by one, and what remains is the one surreptitiously spoken theory that cannot be disproven because it cannot be factually addressed: juice. For the good of the game, let us hope that theory as well is eventually undermined by evidence.
THE COMPONENTS OF PLAYER VALUE
Even before the result was announced, conventional wisdom conferred the 1998 National League Most Valuable Player Award on Sammy Sosa rather than Mark McGwire.
The argument for Sosa and against McGwire could be heard that fall on the evening sports shows and read in the national newspapers. USA Today's National League columnist justified his selection of Sosa on the basis that McGwire's accomplishments--which, it should be noted, include setting the all-time record for home runs--came on behalf of a non-contender. In his mind this posed the question, "Valuable to what?" The previous night, an ESPN analyst backed Sosa's candidacy for the same reason precisely one breath before naming Ken Griffey, Jr. the American League's MVP. Griffey's Mariners finished nine games below .500 and in third place in the four-team AL West that season.
Another ESPN commentator said he would vote for McGwire as Player of the Year, but not MVP. This is the smarmy new parsing. In baseball there is no such thing as player of the year. But what of it? ESPN pays its on-air personalities to be advocative and entertaining; commentators are not required to be reality-based.
The cavalier dismissal of McGwire based on his team's nonstanding--which, by the way, received nearly unanimous ratification when the official vote was announced--logically reduces as follows: he's not the MVP because Ron Gant, Donovan Osborne, Jeff Brantley, and Kent Bottenfield stunk.
Yet in that befuddling reasoning, the vote also frames a fascinating and broader question: What precisely is value? As the baseball experts saw it in 1998, Sosa had value because his team won; McGwire lacked value because his team didn't.
The way we define "value" is the linchpin of much of "The Book" about baseball strategy for the simple reason that strategies are inevitably focused toward maximizing "value." Yet in the modern game, value analysis is often subjective and thus prone to error. Sports commentators commit this error as do team executives. As we consider the relevance of the modern baseball "Book," one of our foundations has to be an understanding as to what constitutes "value." In forming that understanding, we have to be careful that our trendy definitions don't confuse the ends with the means.
THE FORMULA FOR SUCCESS
One of the singular beauties of baseball is that the components of success can be quantified. Relating the value of a basketball player to scoring average is too one-dimensional; football rotisserie leaguers don't even try to assess the value of a pulling guard. It's hard to reduce to a numerical formula the components that make a great professor, a great lawyer or, for that matter, a great book. The measure of a corporation's success can be stated on a bottom line, but can one devise a formula accurately presenting the relative contributions of all the corporation's accountants, salespeople, engineers, production line workers, and officers to that bottom line?
This can be done in baseball because we have a very measurable result--namely a victory--which is created by means of a second entirely measurable occurrence, that being runs.
Thanks to the intervention of computers, we know--and I do not use the word "know" loosely--that in the modern game the average base hit produces forty-six one-hundredths of a run; the average double produces eight-tenths of a run, etc. Conversely, we know that the average out reduces production by one-quarter of a run.
All that being the case, and setting aside only the intangibles and uncountables, determining the MVP at season's end is a deceptively simple task. Count the singles, doubles, triples, home runs, walks, steals, and outs made by each player, apply the proper run-producing value of each event, add proper fielding and baserunning quotients, and you can define any player's value in the appropriate context of games won. (The same process works essentially in reverse, by the way, for pitchers--with equal validity.)
The Most Valuable Player Award is a frequent backdrop for value-related arguments. Sometimes it takes the McGwire-Sosa context. At other times, the debate focuses on whether a pitcher should be eligible to win the award. None has since Dennis Eckersley won with Oakland in 1992. In the case of pitchers, the disqualifier tends to be twofold: First, they have their own award, the Cy Young; and second, they don't play every day. Of the first, little need be said in refutation beyond that it is plainly silly. So what if pitchers have their own award? In the NHL, defensemen have their own award ... did that stop hockey writers from voting for Bobby Orr for MVP all those seasons?
The second argument, that pitchers' value is diminished because they are not everyday players, is superficially more appealing. Yet given a tinge of thought, that argument, too, quickly collapses. The essence of baseball is the pitcher-batter matchup, and it is not a vast oversimplification to say that the fellow who wins the greatest number of those matchups over the course of a six-month season is the most valuable player. There are, of course, degrees of "winning" and "losing" such matchups: While both a walk and a home run qualify as "wins" for the batter, one carries a substantial number of bonus points. But we can adjust for those variations.
It is certainly true that an everyday position player has the opportunity to impact four or five times as many games as, for example, a starting pitcher. At the same time, a starting pitcher's impact on the game in which he appears is several times more significant than the impact of any individual position player. This is so for an obvious reason.In the typical game, a position player gets four or perhaps five opportunities to "win" or "lose" a pitcher-hitter matchup--and by doing so, influences the game's outcome. In a stretch of five games, that might amount to participation in between twenty and twenty-five such confrontations.
A starting pitcher would only pitch one of those five games, but he would be involved in every one of the pitcher-hitter confrontations for the duration of his tenure in that game. Pedro Martinez may have only made 29 starts in 2000, but he pitched 217 innings, averaging seven and one-half innings per start. Since we know that Martinez allowed 160 base runners by either a hit or a walk, we can state the proposition in another and more precise way: Martinez influenced 811 pitcher-hitter confrontations in 2000, and he won 80 percent of them. The average American League pitcher won 65 percent; Martinez's edge over the field was 15 percent.
Jason Giambi, the writers' choice for the 2000 MVP Award, certainly had a fine season. But Giambi was a part of only 647 pitcher-hitter confrontations--influencing the outcome of games 164 fewer times than Martinez. He did win 47 percent of those matchups, 12 percent more than the 35 percent league-wide average for batters. But still, his advantage was less substantial than Martinez's advantage
(A side note: Seen in this context, the 1992 MVP Award to Eckersley--or to any relief pitcher--is plainly silly. In 1992, Eckersley won 77 percent of his pitcher-hitter confrontations ... but he only engaged in 313 of them. More on the frivolousness of considering relievers as MVP candidates in a subsequent chapter.)
Our cursory examination of Martinez versus Giambi has not thus far taken into consideration the supplementary factors we cited above. Giambi, for example, hit 43 home runs; Martinez allowed just 17. Giambi drew 137 bases on balls; Martinez issued just 37. Giambi piled up 73 extra base hits; Martinez allowed 36.
We can't accurately weigh those numbers in our minds, but statistical analysis can. In fact that's precisely what Pete Palmer's BatterFielder Base Stealer Wins and Pitcher Wins1 does: Weigh all those factors and reduce them to a comprehensible sentence. When you do that, this is the result. In 2000, Pedro Martinez improved the Red Sox by a factor of 8.4 victories, the highest total in the American League. Rodriguez actually stood second, improving the Mariners by 6.8 victories. Giambi was the league's third most valuable player, improving Oakland by 5.2 victories.
SPINNING THE TRUTH
The point of this exercise isn't solely to criticize the thought processes most voters apply in considering their MVP selections--although that alone would be a constructive exercise--but to underscore the true concept of value. Recognition of the rough equilibrium between the importance of pitching and hitting--as expressed in the MVP voting--would be a step in that direction. A second step, illustrated by the debate between the credentials of Sosa and McGwire in the 1998 NL race (or the mirror debate involving A-Rod and Miguel Tejada in 2002), would involve recognition of the importance (or lack of same) of winning in the context of individual awards.
Because baseball is a team game, not an individual one, it is impossible to fairly overlay team performance onto matters of individual accomplishment. No single player is good enough to merit that sort of accolade. The Sosa versus McGwire scrum illustrates the failure ofsuch logic. Sosa's 1998 achievements added up to a Batter Fielder Base Stealer Wins of 5.0. To put it another way, a team comprised of 24 completely average players plus Sammy Sosa should (pending the intervention of luck) have finished the season five games above .500 (86-76). But McGwire's achievements equate to a BFW of 7.1; a team of 24 average players plus McGwire should have finished 88-74 and beat out Sosa's crew by two games. To argue anything else is to visit Bottenfield's sins on McGwire.
These debates concerning value are a fascinating and ongoing part of baseball, and they are also--hopefully happily--much of what this book is about. Perhaps it is his personality as much as his performance, but in a recurring way, Sammy Sosa often finds himself personifying one aspect or another of that debate.
In June of 2000, when the Chicago Cubs briefly put Sosa on the trading block, the issue was all about his value to the team.
That seemed a silly argument to Sosa, and to a lot of Cub fans. Consecutive seasons of 60-plus home runs, nearly 300 RBIs along with an MVP Award and a spot in the playoffs (in 1998), and the guy needs to defend his production? "I don't know what more I can do," complained Sosa when Chicago management sniped at him.
His manager, Don Baylor, along with some of the Cub brass, had a few suggestions. The critique as management saw it: He struck out too much, didn't use his speed, had become increasingly one-dimensional, and was a liability in right field. "He drives in 160, but lets in 45 on defense," went the now-famous anonymous remark from the unnamed team official.
The Sosa Affair represented an especially prominent instance--only one of many in baseball (as in life)--of both sides engaging in selective truth-telling. When they put you under oath in court you are sworn to tell "the truth, the whole truth ..." but nobody takes an oath during meetings with the press, certainly not ballplayers or club executives. Therefore what often emerges is a process of limited truth-telling. In the matter of Sosa's value to the Cubs in the summer of 2000, not a single word presented by either side could have been arguedto have been spoken in error. But neither was either side's version accurate in the sense of being a full, honest, or complete assessment: "the whole truth," as it were.
The facts are that by June 2000, Sammy Sosa had become everything both his supporters and his detractors said he was. Sammy Sosa could make a clear and compelling case for his accomplishments. One does not devalue a home run; there is no such thing as a cheap or empty one. Runs batted in are a bit more problematic--the situation dependency of their nature can compromise their value as an index of production. But even so, 158 of them (in 1998) followed by 141 of them (in 1999) is pretty tough to argue around.
Nor could there be any question of Sosa's credentials as a gamer. Between 1997 and 1999, he missed only four of Chicago's 487 games. He also started every Cub game in 1995, and would have done the same in 1996 had not a fracture sustained when he was struck by a pitch cost him that season's final six weeks.
As might be expected, Sosa's Batter Fielder Base Stealer Wins--the best single-number gauge of all-around contribution--was superb in both 1998 and 1999. Following his 5.0 rating of 1998, his BFW slid in 1999, but only to 4.4, still a substantially positive figure.
Yet to Sosa fans, there ought to be something unsettling in that 4.4 BFW rating. It came, after all, during a season in which he delivered 63 home runs, a figure at the time surpassed by only McGwire in all of baseball history. More than that, Sosa counted for a league-leading 397 total bases, adding 24 doubles, 2 triples, and 91 singles to his 63 blasts. That sounds like it ought to produce a mega-mega season, not merely a very good one.
For purposes of comparison, here are some of the other players who rated within a few fractions of a 4.4 BFW in 1999: Craig Biggio (4.3), Brian Giles (4.1), Mark McGwire (4.2), and Robin Ventura (4.5). The best mark in the National League, by the way, belonged to Houston's Jeff Bagwell (5.3).
Well, the company Sosa kept at that rating level in 1999 certainly is nothing to be ashamed of. At the same time, when a similarly productiveplayer is Brian Giles (24 fewer homers, 26 fewer RBIs, 77 fewer bases), it is fair to suspect that there are, indeed, gaps in Sosa's all-around play, and that some of those gaps are reflected in his BFW. Which means the Cubs, too, may have been telling at least some of the truth.
SOSA AND GILES MANO A MANO
If we can identify those gaps in such an obviously talented player, they may shed light on the less readily obvious yet important aspects of talent as it contributes to what baseball is all about: winning games.
What in the heck did Brian Giles do in 1999 to help the Pittsburgh Pirates win games that Sammy Sosa didn't do for the Chicago Cubs?
First, he got on base more ... and it was a lot more. Savvy fans today recognize that if on-base average isn't the most important single statistic, it's in the top two. Giles collected 164 base hits in 521 official at bats for a batting average of .315. Sosa had 180 hits--16 more than Giles--but he burned up 104 additional at bats in getting them; that's 88 more outs laid at Sosa's feet than at Giles's. At three outs to an inning, it's fair to say that Sammy Sosa killed a lot more innings than Brian Giles did.
Giles then compounded his on-base advantage by drawing 17 more walks than Sosa, and by doing it in that same 104 fewer plate appearances. The impact of a walk on potential run production, which goes unrecognized by many sluggers, is effectively doubled by this reality: Batters who draw a lot of walks do so by not swinging at pitches that batters who draw few walks are being retired on. Should it be viewed as coincidence that in 2004 Barry Bonds batted .370 while at the same time drawing more than 100 unintentional walks? Not in the slightest; they are in fact very closely related. What do you suppose is the normal batting average on pitches thrown out of the strike zone that are put into play? We've no way to truly tell, but a pretty good guess would be about 150--after all, that's why they make pitchers throw the ball over the plate. The consequence of patience thus is thatnot only is a player more likely to reach base, he is at the same time less likely to be put out. Without even swinging, the on-base average and batting average both look better, the one because of the walk, the other because of the out he didn't make.
In the cases of Giles and Sosa, we can apply the time-tested Linear Weights formula to parse the data. Sosa's 63 home runs produced 88.2 runs; Giles's 39 produced 54.6. Sammy also out-singled Giles, 91-89. At .46 of a run per single, that's an additional run in Sammy's favor, moving his edge to 34.6. But it's about the only edge Sosa has.
Giles hit 33 doubles (good for 26.4 runs) while Sosa hit 24 (good for 19.2 runs). That reduces Sosa's edge to 27.2 runs. Giles had three triples (3.06 runs), Sosa had two (2.04), costing another run.
The differences in walks and outs erode most of the rest of Sammy's advantage. At a cost of a quarter run per out, Sammy's additional 88 outs erase 22 runs. Sammy's edge over Giles in value falls to 4.2 runs. Giles's 95 walks amount to 31.4 runs; the 78 walks Sosa drew only count for about 25.7. Suddenly Giles measures as the better batsman by a factor of about a run and a half.
In days past, Sosa could have counted on his baserunning speed to boost his rating. But not lately. In 1999, he only tried to steal 15 bases, and only succeeded 7 times. That performance is a net cost to the Cubs of 2.7 runs. Giles only stole 6, but he only tried 8 times, meaning that he created 6 fewer Pirate outs on the bases. His baserunning net: just over a half a run. Add them all up, and offensively Giles actually produced about 5 runs more for Pittsburgh than Sosa did for Chicago.
Their performances in the field continue the effective standoff. Sosa (2.41) showed better range in right than Giles (2.38) did in center; that's a clear plus for Sammy. On the other hand, Giles committed only two errors, while Sosa made eight. Giles also rang up six assists, one more than Sosa. Considering that Sosa played 300 more innings in the field, that can't be overlooked.
When the Cubs filed their brief against Sosa with the ticketbuying public, they were largely taking issue with his defensive performance. Part of what bugged Cubs management can be measured, andits impact on run-prevention quantified. The quantifiable elements include put-outs, assists, errors, and range. Playing mostly in right in 1999 (he also played 195 innings in center), Sosa registered 322 put-outs, translating to that rather high range factor. It was in fact fifth best in the majors (behind Michael Tucker, Jermaine Dye, Mark Kotsay, and Tim Salmon). That also represented a significant upgrade from Sammy's far more average 2.17 range factor of 1998. In that very measurable and meaningful sense, Sammy had become a more complete "go-getter" type player.
That improvement was not reflected, however, in two other signposts of defensive acumen, assists, and errors. Sosa committed eight errors in 1999, a total exceeded among right fielders only by Vladimir Guerrero. His .976 fielding average was below the .982 average that was par for regular Major League right fielders in 1999. In fairness to Sosa, it is dangerous to indict him too harshly--as some might interpret the Cubs as trying to do--for the subpar fielding average. Playing a position where few errors are made by either the best or worst fielders, small events carry great statistical weight. If, for instance, Sosa had committed just two fewer errors (six rather than eight) his fielding average would have precisely matched the Major League average. It would be hard to argue that the occurrence of two errors over a season is enough to differentiate between an average fielder and a poor one.
His assist total is another matter. Sosa posted five assists in right field in 1999, three fewer than the Major League average for his position. But he significantly trailed many of his contemporaries. Mark Kotsay led the majors with nineteen, nearly two and a half times as many as Sosa. Jermaine Dye and Albert Belle also had more than twice as many as Sosa, while Guerrero, Magglio Ordonez, Paul O'Neill, Larry Walker, and Matt Stairs all reached double figures. Only the previous season Sosa had run up a much more representative total of fourteen. (The trend has continued since 1999, by the way. In 2003, Sammy recorded just two assists. Among regulars in right field, nobody got fewer than that. By 2004, Sosa's liabilities were viewed by many as offsetting his strengths.)
LESSONS FROM LITTLE LEAGUE
Why fuss over a few assists? If that's indeed what got under the skin of Baylor or the rest of Cubs' management, it might well justify their pique with their star outfielder. For starters, let's keep in mind what an outfielder's assist does. It is more than merely an out. In virtually every case, it both posts an out and erases a base runner, often at the plate, and it often follows on the heels of another out--a caught fly ball. An outfielder's assist is the very definition of a rally killer.
When an outfielder's assists decline, it's almost certainly for one of a handful of reasons, the majority of them bad. Granted, it is plausible that an outfielder's assists might decline for the same reason Paul LoDuca threw out nearly three times as many base stealers in 2003 as Ivan Rodriguez: Everybody thought they could run on Paul LoDuca, but hardly anybody even tried to run on Ivan Rodriguez.
But as it pertains to Sammy Sosa, a question inevitably follows: Do you think base runners are afraid to run on him? Do you think base runners view him as the new Roberto Clemente? Because unless you do, there has to be a more onerous answer.
That's the judgment Cubs' management presumably came to. Its indictment: That Sosa became a classic "scatter-arm"; he too frequently overthrew the cutoff man, allowing trailing runners to take extra bases when he had no chance to retire lead runners. That he threw wildly up or down the line, especially on throws to home plate. That he gave runners too big a head start by the simple act of taking too long to throw.
This list of grievances is interesting to baseball statisticians for one commonality. All three of the on-field "crimes" mentioned in the paragraph above are recognized to occur, yet not a single one of them can be reduced to a number. Unless they are formally scored as an error--and they routinely are not--no scoresheet, data service, or record book will maintain them under the heading of "throws off-line," "runners unretired," or "poor execution."
That means we have no way of closely approximating the legitimacy of the Cubs' complaint as it pertains to the impact of Sosa's defensivedeficiencies on Cub fortunes. Yet every one of his presumed flaws leads to runs being scored. Since these results often occur on plays at the plate, the damage can in fact be instantaneous.
In the absence of hard data, let's play a few "what-if" games. What impact might Sammy Sosa have on run production if he were a more polished player in these impossible-to-quantify ways? One of his deficiencies--it's a common one at the Major League level--is simply that he catches the ball with one hand. Since their introduction to Little League, fielders are taught to "use two hands," but that's generally on the rationale that they'll drop it if they do otherwise. The admonition falls by the wayside somewhere around the American Legion level, and by the big leagues few managers would have the nerve to order outfielders to use two hands to catch the ball.
Yet there's a perfectly valid reason, one which has nothing at all to do with the likelihood that a fly ball would otherwise be dropped, for even the most reliable of Major League outfielders to use two hands as a matter of routine.
How many times in the course of a ball game does a "bang-bang" play occur at a base? That's a play in which the ball and the base runner arrive at virtually the same time, and the umpire's call could be either safe or out. The result is determined by literal fractions of a second.
Here's the follow-up question: When Sammy Sosa, or any other outfielder, catches a fly ball with one hand, how many tenths of a second does he waste getting his ungloved hand up from his side and into position to remove the ball from his glove in preparation for a throw? Five tenths? Three? Let's say it takes Sosa an extra three-tenths of a second, once he catches the ball, to position his throwing hand, which until that moment has been resting languidly at some locale other than adjacent to his mitt. How many bang-bang plays do you suppose occur during the course of a season in which the result is decided by a matter of three-tenths of a second?
Well, even the slowest base runner can cover the ninety-foot distance between bases in about five seconds, a pace that translates to eighteen feet per second, or about five and a half feet in three-tenthsof a second. So the question really is: How many tag plays at a base are decided by a distance of five and a half feet or less? Virtually all of them; such a play probably comes up multiple times a week for every Major League team. So let's shorten the time: Let's say Sosa can shift his throwing arm from at rest into position to grip the ball in one-tenth of a second; that still translates to nearly a two-foot edge to the base runner, more if he has any speed at all. How frequently is a play at a base decided by two feet? Once again, while we don't have numbers, instinctively we'd say it happens on many plays involving a sliding runner.
That lapse in positioning his throwing arm to make a throw--not the prospect that he might need it near his glove to prevent dropping the ball--is part of the reason Cubs' management instinctively landed on Sosa for defensive deficiencies. (It's also the real reason why all outfielders--not just Little Leaguers--ought to make it a lifetime habit to catch the ball with two hands with runners on base.)
THE PERFECT SAMMY
Let's say we created the perfect Sammy Sosa ... what difference would it make? We've just described a circumstance related to arm positioning for a throw that probably comes up several times in Sosa's week, several chances to improve his prospects for retiring a runner on the bases. But there are many variables in throwing a ball, and even a well-positioned Sammy isn't going to make a perfect throw every time. So let's say--it seems a minimalist argument--that Sammy's revitalized arm positioning technique merely results in retiring one additional base runner at home plate every two weeks. What difference would that make to the Cubs' fortunes?
Over the course of a twenty-six-week season, it would add 13 assists to Sosa's total, and he would instantly soar to the very top of the Major League charts for right fielders. Let's say such a play occurred with a runner on third base and none out. According to Pete Palmer's base-out situation, the beginning run potential in that circumstance is 1.277; that is, the offensive team could expect to score on average1.277 runs. Once the ball is hit, there are three possible outcomes. Here is how the run potential is affected by each possible outcome:
1. The runner tags. Using his normal form, Sammy makes his usual throw and the runner scores from third. With one out and nobody on, the offensive team's run potential decreases to .249. However, since a run already has scored, the inning's true run potential is 1.249.
2. The runner tags. Using his revised and quickened form, Sammy pegs the runner out. Now instead of a runner at third and none out, there is nobody on and two out. Sammy's arm has reduced the inning's run potential from 1.277 to 0.095, less than a tenth of a run per inning. That's a swing of more than a full run.
3. The runner tags but, respecting Sammy's newly won reputation as a gun, doesn't try to score. The offense still has a runner at third, but this time with one out. Its run production potential declines from 1.277 to .897 runs. Without even making a throw, Sammy has crimped the other guy's offense by about four-tenths of a run.
In other words, if Sammy's revised arm positioning resulted in the erasure of thirteen runners from the bases over the course of a season, it could reduce opposition run production by a projected 15.4 runs. Those numbers change depending on the number of outs in the inning. If, for instance, Sosa makes his play on a fly ball with one out instead of none, a runner thrown out at home attempting to score reduces his team's expected run production for the inning to zero--because the play ends the inning. But as a general guidepost, a swing of 15 runs would not be out of line for such a change in performance. And even in these days of high-octane offense, a swing of 15 runs translates to an improvement of a game or two in the standings. That's a big reward for simply raising your hand when you're told to.
Sammy Sosa commits other unmeasurable transgressions--so do all outfielders. The most common involves overthrowing the cutoff man. It's a macho thing, don't you know? "I don't need no stinkinghelp throwing this guy out." But even great outfielders are taught to hit the cutoff man for three reasons, all of them valid:
1. The cutoff man can redirect a throw that is in time, but flying off-line.
2. The cutoff man can intercept a throw that will be late, and do so in time to retire some other base runner or at the least prevent that base runner from attempting to take the extra base.
3. Even if the cutoff man doesn't touch the throw, the mere possibility of his cutting off the throw may deter adventurous trailing runners.
The value of the first reason is an extension of our earlier point. It translates to additional runners retired at home or third. Instead of five assists our real-life Sammy recorded in 1999, a perfect Sammy employing good positioning technique and hitting the cutoff man might (gunning out one additional runner every two weeks) throw out eighteen runners, or possibly more.
The second and third points translate to reducing the odds of other runners scoring. Let's say Sammy always hit the cutoff man. We have no way of measuring the impact such a perfect Sammy would have on defense, but let's say such a perfect Sammy erased from the bases one runner trying to advance from first to second every other week, and kept one additional runner per week from attempting to advance. If this always occurred in a one-out situation, and employing Palmer's base-out chart, the impact on run production would be as follows:
1. The run production potential of a runner at first with one out, the situational starting point, is .478. If one such runner were thrown out every other week (thirteen occasions over the course of a season), thus changing the situation to none on and two out (.095 run production potential), the reduction in offense would project to 5 runs.
2. If one such runner were prevented from advancing each week, the situation would be changed from runner on first, one out, to runner on first, two out. The run production potential of a runner on first, two out, is .209 runs. Occurring 26 times, such an event would decrease projected offense by about 7 additional runs.
In other words, allowing those two circumstances only, our perfect Sammy would decrease the other guy's run production potential by an additional 12 runs. Along with the projected 15 runs he's begun to erase at the plate, Sammy now crimps the other guy's offense by an additional 27 runs. That's a three-game improvement in the standings. Sammy's BFW, by the way, would climb from 4.4 to about 7.3, giving him a legitimate claim to the 1999 Most Valuable Player Award.
And that improvement would occur at the mere price of catching the ball with two hands and hitting the cutoff man. So when Don Baylor and the Cubs argued--against all Sammy's home runs and against all his RBIs--that he let in too many runs by sloppy defensive play, their case was not without merit.
Much of what has preceded in this chapter falls quite firmly within the realm of raw speculation. The fact is that while we can calculate the value of a strikeout and a walk, we cannot assert with finality how many would-have-been walks end up as strikeouts. Too, while we can measure the theoretical value of a poor throw, just as we can measure the value of a home run, we can't translate that theory into practice because we can count only the home runs, not the poor throws. Doing the latter would require making a judgment that a throw that did not retire a runner would have retired the runner had it been executed with greater propriety and precision. That isn't done. Keep in mind that we are not talking about errors here, at least not as baseball recognizes errors; we are, instead, talking of mistakes.
The limitation of much of statistical analysis--including much of what follows in this book--is that it is necessarily founded on that which can be measured. It must inevitably exclude from consideration that which cannot be measured but which exists notwithstanding. When Sammy Sosa argues that he is a great ballplayer, his defenseconsists largely of that which can be measured and quantified. When the Cubs argue against his case, their defense is more esoteric and less substantive. That is not to say it is either more or less true.
Statistics can aid us in sorting out selective truth-telling from the opposite. We ought to pursue that sorting process--and we will in the pages of this book. But our pursuit remains limited by the constraints of our knowledge. Our search for the perfect statistic, like the Cubs' search for the perfect Sammy, is not yet fulfilled.
1.1 A Diversion into Win Shares and Loss Shares
In the field of assessing player performance, the big recent development was the unveiling of Bill James's Win Shares formula. Bill got a full book out of it in 2002, and his seasonal Handbooks contain the Win Shares ratings for players based on subsequent seasons.
There is no need to go into the theory or substance of Win Shares in detail in this space; Bill has done so with his characteristic thoroughness and elan in Win Shares itself. Because it is one of the few efforts extant to try to synopsize the sum total of a player's contribution in a single number--a sort of Grand Unified Theory--Win Shares is worth noting. Because Bill James is the brain behind it, Win Shares is worth taking very seriously. But Win Shares should be analyzed warts and all. To date--particularly in the reviews that followed the theory's initial publication--there was far more paean than pan. Since we are on the prowl for ways to assess player performance, and since Win Shares is the vogue new tool, Win Shares deserves a more objective review.
That review starts with the underlying premise, which Bill himself states (on page 347 of his 2004 Handbook) as follows:
The Win Shares credited to the players on a team always total up to exactly three times the team's win total ... . If a team wins 80 games, the players on the team will be credited with 240 Win Shares, always and without exception. Nothing--not even a rounding error--is allowed to disrupt this relationship.
As a general rule of life, it's a good idea to beware of absolutes, and this is one of those moments. So the first critique of Win Shares as a tool for assessing overall player performance is this: It is based on a fixed and unyielding artificial stricture. For Bill, who has contributed so much to the statistical and mathematical contribution of baseball (and, by extension, beyond baseball), that is a shockingly nonmathematical constraint to insist upon.
There is no reason why one should be able to force any valid assessment of relative player performance out of such a stricture. If a team wins 80 games, why does it follow that 240 points of responsibility ought to be assessed, much less subdivided? Why not 238, 239, 243, or 300? If one team wins 80 games and another wins 79, why--in a mathematically valid system--should the first team's players get three more Win Shares than the second team's players? Bill's answer is the triumph of functionality over science: Because it works out, that's why.
Well, there's certainly something to be said for things working. You probably won't know the answer to this one, but just for fun take a guess: What's the standard deviation for wins by a team in a season after all the measurable skill factors are considered? In other words, how many games are decided essentially by luck? The answer is plus or minus 6.35 games. That means that if one team won 80 games, the only thing you can really, truly say about the skill levels of the players involved on that team--expressed mathematically--is that those skill levels should have resulted in between 73.65 and 86.35 victories.
So even if you buy Bill's premise of three Win Shares per team, the most you can say mathematically is that the team's players should receive somewhere between 220.95 and 259.05 Win Shares ... not precisely, eternally, and unbendingly 240.
There is a second basic flaw to the underlying logic of Win Shares, namely its presumption that all games are won. This is manifestly and obviously not so. At a statistical level, a team that wins 80 games also loses 82 games. Shouldn't we have Loss Shares? I suspect that Bill's initial response would be to note that the outcomes of those games are accounted for in some other team's Win Shares, as of coursethey are. But that is based on another instinctively bad presumption, namely that the outcomes of games are determined by the positive actions of the winning team as opposed to the negative actions of the losing team.
In the long ago of baseball, this was a commonly held notion. We can see it in the evolution of record-keeping. In baseball's Great Big Book of Everything, you can find records of stolen bases dating in an unbroken chain all the way back to 1886. Harry Stovey of Philadelphia's American Association team led the majors that year with 68 swipes. How many times did they nail old Harry back in the summer of '86? Excellent question; if you find out, let us know because although baseball started counting successful steals in 1886, it didn't start counting unsuccessful ones until 1914, twenty-eight years later. The point is, though, that it did eventually start because it dawned on somebody that counting successes only told half the story.
No one can say precisely how often game outcomes are determined by some incompetence, either of strategy, tactic, or execution. But instinctively (this may come from being a Cub fan), I'd say it happens fairly often, and maybe routinely. That's what "don't beat yourself" is all about. In Bill's Win Shares system, however, incompetence does not exist, because literally every player who is rated gets a positive rating. This takes place, in turn, out of some abhorrence Bill has developed for the concept of zero as an average. He says so in his book.
The fatal error in the method of measuring players as better or worse than .500 is that it forces one into the assumption that value consists in being better than average. That is NOT what constitutes value. What constitutes value in baseball is being good enough to play at this level. What constitutes value is being good enough to help the team win some games.
Bill goes on at some length on this subject--he's been known to do that--and when one extracts a single paragraph from such an epistle, the inevitable hazard is of quoting out of context. I hope that's not too greatly the case here, but in any event you would neither beharmed nor bored by reviewing Bill's full argument as outlined in Win Shares. When you do so, you will still have to confront the reality that--if Win Shares considers malperformance at all--it does so only in the most flawed way possible.
There are scads of ways to illustrate the problem, one of which is to take two players at the same position on the same team. Let's take Jeff Cirillo and Willie Bloomquist of the 2003 Mariners. Cirillo was the third baseman at season's start, but he slumped out of the gate and never did get on track. Eventually, the Mariners, struggling to keep pace with the A's in the divisional race and with the Red Sox in the wild card race, benched Cirillo. Down the stretch they turned to Bloomquist, who got in 37 games at the position. Just looking at their lines, you'd say it was an improvement, although unfortunately for the Mariners not a decisive one in terms of the pennant race. Cirillo had batted .204, Bloomquist hit .250. In about 215 trips to the plate--65 fewer than Cirillo--he posted slugging and on-base averages that were both also significantly higher than the man he replaced. Cirillo was the better fielder, although his .978 percentage (very good) and 2.33 range factor (bad) merited a mixed critique.
In sum we have two players, one playing poorly and the other soso, the key difference being that the fellow who played poorly was on the field more. How should those players be rated?
Since he was worse over a longer period of time, there is no good reason why Cirillo should rate on a par with Bloomquist. Yet because it considers only positive accomplishments--of which an important one could essentially be described as hanging around--Win Shares rates Cirillo and Bloomquist dead even. Both get three Win Shares in the James system. Given his part-time status, it would indeed be surprising had Bloomquist gotten more than three Win Shares. Even a really good occasional player would not get the opportunity to move very far up the Win Shares hierarchy.
When I think of really good part-time players, I think of Manny Mota, the great Dodger pinch hitter from the 1970s. In 1977, Mota batted .395 for the Dodgers, almost entirely while coming off the bench in key pinch-hitting situations. He only played one game in thefield, got fewer than 50 plate appearances, but delivered 15 pinch hits and posted an OPS above 1.000. How many Win Shares did Manny Mota get in 1977? Two, that's how many. Do you know who else on the 1977 Dodgers got two Win Shares? Lance Rautzhen did. Rautzhen was a rookie fill-in pitcher who made 25 appearances out of the bullpen with a 4.29 ERA that was about a third of a point above the league average. In other words, relatively speaking he was pretty bad. Do you know who else got two Win Shares on that 1977 team? Glenn Burke did. A utility outfielder, he only hit .254, but he batted nearly four times as often as Mota so he had a lot more chances to pile up the Win Shares. Do you know who got more Win Shares than Manny Mota? John Hale did (he got three). Another utility outfielder, he only batted .241, but again he racked up substantially more plate appearances. Take a look at the stat lines for 1977 and tell me who was more valuable to the Dodgers: Manny Mota, Lance Rautzhen, Glenn Burke, or John Hale? Win Shares picks Hale, largely because Tommy Lasorda played him more.
Because they recognize the importance of malperformance in assessing player value, zero-based systems do not succumb to these flaws. Palmer's BFW/PW system, for example, determines that while Bloomquist (-1.3) was hardly an asset to the 2003 Mariners, he was significantly less of a liability than Cirillo (-2.7). BFW/PW, which gives no particular credit to hanging around, rates Mota's 1977 season at +0.5, a net contribution of a half game to the Dodger cause. Rautzhen, by comparison, receives a -0.2 PW. And for burning up far more at bats than Mota while producing far less, Hale and Burke both get what they deserve: a -2.0 BFW in Hale's case and -2.5 in Burke's.
So in Win Shares we have a system that uses as its foundation a sweeping and statistically unsupportable generalization about how to ration value, that argues against the significance of zero, and that recognizes Win Shares but not Loss Shares.
Which is why Palmer's system, possessing none of these flaws, is the one I prefer to rely upon in this book.
THE BOOK ON THE BOOK. Copyright © 2005 by Bill Felber. All rights reserved. No part of this book may be used or reproduced in any manner whatsoever without written permission except in the case of brief quotations embodied in critical articles or reviews. For information, address St. Martin's Press, 175 Fifth Avenue, New York, N.Y. 10010.