By Alan Schwarz
July 9, 2004

---

Chat Transcript: Alan Schwarz took your questions

---

Most fans, players and even team executives assume that baseball's infatuation with statistics is simply a byproduct of the information age, a phenomenon that blossomed only after the arrival of Bill James and computers in the 1980s. They couldn't be more wrong.

Baseball America's own Alan Schwarz will correct that misperception with his new book, "The Numbers Game," to be published by St. Martin's Press in early July. Schwarz provides the first-ever history of baseball statistics, showing how baseball and its numbers have been inseparable ever since the pastime's birth in 1845, and telling the story of this obsession through the lives of the people who felt it most: Henry Chadwick, the 19th-century writer who invented the first box score and harped endlessly about which statistics mattered and which did not; Earnshaw Cook, a scientist and Manhattan Project veteran who retired to pursue inventing the perfect baseball statistic; John Dewan, a former Strat-O-Matic maven who built STATS Inc. into a multimillion-dollar powerhouse for statistics over the Internet; and dozens more.

In this exclusive excerpt, Schwarz travels back to the 1950s to profile the work of George Lindsey, a Canadian military strategist whose work in statistics was a full generation ahead of its time. So you think baseball stats are just a modern fan fixation? Read on . . .

---

The Georgian Bay off Lake Huron lay peaceful and still, tall trees standing sentry over the scene's verdant tranquility. A solitary rowboat bobbed out in the distance. In it was a fisherman, out for bass, but one whose pursuit of a bite or two was surely impeded by the crackling dissonance coming from the heavy, dusty box that sat in the boat with him. He didn't mind. To this fellow, fish were but a pleasant distraction. The Toronto Maple Leafs were playing baseball on radio station CFRB, and Lt. Col. Charles Lindsey had statistics to catch.

This was the passion of Lindsey's retired life. Like many Toronto families in the 1950s, every summer weekend the Lindseys drove their old Model T two hours north and rented a lakefront cottage. Mothers would tend to the kids as fathers drank whiskey and told war stories into the night. Charles Lindsey owned more than a few. The strapping, decorated infantry officer had spent four years in the Canadian army during the Great War, then five years in World War II. His free time after retirement, though, begged for something to occupy him. So Lindsey, a lifelong baseball fan, decided to help his son George with that . . . well, that "project" of his, as the rest of the family dubiously referred to it.

George Lindsey, you see, had a day job: figuring out how to intercept Russian bombers before they reached Halifax. A pioneer in the new military field of Operations Research that grew out of the Cold War posturing of the 1950s, Lindsey modeled flight formations and tactical strikes for the Canadian Department of Defence, assessing risk and cost and strategy. Yet something nagged at him. Data was murky and incomplete. Combat was too fluid to pinpoint the value of decisions each side made. Where could he learn from a system whose patterns lent themselves to such analysis? Which had built-in stoppages of time, clear strategic decisions, with distinct, calculable states that could be expressed in manipulable numbers?

"Doggone it," Lindsey said to himself, "you can do that with baseball."

And so he did. Lindsey spent the next 15 years, whether stationed in Quebec, Colorado, Washington, or under a distinctly disapproving Norwegian commander in Italy, devoting his days to military modeling and his nights to the study of baseball games. Was the stolen base a good strategy? The sacrifice bunt? How many runs scored after these decisions, and did they wind up worth the effort? For the first half of the twentieth century, baseball fans had been content to debate these matters with little to no statistical proof. George Lindsey would dig deeper than anyone had before.

"Baseball is the highest form of human activity—it should be analyzed," recalled Lindsey, now 84, grinning as mischievously as he surely had then. When he was through, the nuclear warheads his department had feared remained in their silos. But George R. Lindsey, along with his father and his rowboat, had become the first to use formal statistics to split the atom of baseball strategy.

Analyzing statistics had intrigued, even consumed, people before. F.C. Lane of Baseball Magazine, way back in 1917, wrote a series of articles that was one of the first published attacks on batting averages and suggestions of a better system. Since baseball's earliest days, the batting average—the percentage of hits per opportunity to get one—was virtually the sole accepted means of quantifying batting talent. A .330 hitter was excellent, a .300 hitter good, a .250 one mediocre and so on, with little attempt, let alone widespread desire, to do better. Lane asked, "Would a system that placed nickels, dimes, quarters and fifty-cent pieces on the same basis be much of a system whereby to compute a man's financial resources? And yet it is precisely such a loose, inaccurate system which obtains in baseball . . . Pretty poor system, isn't it, to govern the most popular department of the most popular of games?" In order to assess the relative value of singles, doubles, triples and home runs, Lane and some colleagues kept strict account of 62 games in order to identify exactly one thousand hits—and from that group investigated just how many bases were advanced, and ultimately runs scored, by each type of hit. The results valued singles as worth 45.7 percent of a run, doubles 78.6, triples 115.0 and home runs 155.1. Clearly, allowing batting average to treat them all the same was horribly inaccurate. But with Babe Ruth still a pitcher for the Red Sox and slugging still frowned upon as the shameless display of brute force, Lane's appeals for change went unanswered.

In the late 1940s, Branch Rickey hired a Canadian named Allan Roth as his personal statistician, using him to rate players by what we now call on-base percentage. Roth, baseball's first full-time statistician, also devised new ways to break down the performance of Dodger players in different situations: hitting versus lefthanded and righthanded pitchers, at home and away, during various months and more. Ahead of their time by a good generation, these statistics captured detail that influenced trades and other Dodger personnel decisions.

Up in Canada one decade later, George Lindsey had different goals in mind. As a military strategist, he was more interested in how various maneuvers (singles, steals, walks, and so on) fit together as if advances on a battlefield. Lindsey's original passion as a boy growing up in Toronto was hockey—he received some professional offers as a goalie—but that sport was too fluid, with too few stops in the action and therefore measurable states of combat, to lend itself to such analysis. Baseball was perfect, though finding proper data was impossible. The annual Spalding guides Lindsey bought since childhood had only the players' final season statistics, which told nothing of how the hits and outs interacted within innings. Even daily box scores, detailed as they were, recorded the events of each game only by summing up the feats of individual players. Lindsey wanted the inning's feats—the players to him were not Mantle, Mays, and Musial, but home run, stolen base, and single. How did they combine to score runs? What mixtures were most successful? No one, at least publicly, had quantified such data. Lindsey would have to tabulate it himself by keeping score of live games off the radio. He did just that, night after night after night.

This was not necessarily conduct becoming a Lindsey. Besides being a military officer, George was a direct descendent of William Lyon Mackenzie, the first mayor of Toronto who later advocated open rebellion to break from the British throne; he also was a cousin of William Lyon Mackenzie King, Canada's prime minister during World War II. Nor did Lindsey's household provide much encouragement for his baseball endeavors. "I developed a great loathing for baseball," his wife, June, harrumphed to me during a visit to their Ottawa home. June Lindsey, a Yorkshire-raised woman of striking intellect, holds a Ph.D. from Cambridge University in X-ray crystallography. At Oxford she worked on the molecular structure of vitamin B12 with Dorothy Hodgkin (later winner of the Nobel Prize) and modeled the structures of adenine and guanine that two doctors across the hall, named Crick and Watson, incorporated into their celebrated model of DNA. So she can be forgiven her distaste for a sport that stole much of her husband's attention for most of their marriage's first 15 years. "If you turned the television to baseball," she said, "I would leave the room." She did that often.

George Lindsey, his wife aghast, watched and listened to roughly 400 games during the mid-1950s, scoring them with exacting precision so that he could investigate several questions that always had dogged him. Specifically, to what extent did platooning—sending up a righthanded hitter to face a lefthanded pitcher, and vice versa—improve batting averages? If a team loads the bases, how many runs typically score? And how much does a player's batting average fluctuate due to luck? During his breaks at the Air Defence Command center in St. Hubert, Quebec, Lindsey examined his scoresheets, tabulated the totals, and wrote up his findings in a paper called "Statistical Data Useful for the Operation of a Baseball Team." It was accepted and published in the March/April 1959 issue of the journal Operations Research.

"It may occur to the reader," Lindsey wrote, "that the amassing of statistical data on baseball has already exceeded all reasonable bounds . . . But in one application, the interpretation of data on past performance in order to influence decisions regarding future situations, there may be some thoughts still worth developing." Using tables, equations, and distribution graphs, Lindsey detailed his findings. When hitters faced pitchers of the opposite handedness, batting averages did indeed go up, by 32 points. When a team loaded the bases, 2.3 runs usually scored with none out, 1.6 with one out, and 0.9 with two. As for the luck issue, Lindsey also showed how a manager who reacted to a hitter's slump—say, a .180 batting average over one week—by pinch-hitting for him was probably panicking; binomial probability theory showed that a true .300 hitter would often bat .180 over seven games through the vagaries of randomness.

Most in his military community considered Lindsey's findings amusing but didn't pay them much attention. One journalist from Washington, D.C., came across the article, probably because of Operations Research's readership among military types, and phoned Lindsey not for a story but to recommend he contact the Washington Senators about consulting for the city's floundering ballclub. But the franchise moved to Minnesota the following year. "I got invited to give talks at mathematical associations and statistics groups," Lindsey recalled, "but they were mainly looking for something to keep people awake."

Lindsey pressed on. His next quest was to determine the relative value of different leads, i.e., how often each type of advantage (one run after the third inning, four runs entering the bottom of the eighth, etc.) resulted in a win or loss. He analyzed 782 games from the 1958 season and discovered, for example, that the visiting team won about 15 percent of the time if down by three runs after the top of the third; the home team, if ahead by two runs after the seventh, would go on to win 88 percent of the time. Operations Research saw too little of a strategic nature in this paper, though, deeming it merely statistical, so Lindsey got it published in September 1961 in the Journal of the American Statistical Association under the title, "The Progress of a Score During a Baseball Game." That article was merely the beginning of a larger study, though, that Lindsey knew would be his most significant.

Since the days of John McGraw and Connie Mack, managing "by the book" had meant employing fairly routine strategies that over time had grown universally accepted. Managers would defend the use of the sacrifice bunt by puffing, "I don't mind giving up an out to move a runner into scoring position." Purely out of reflex they would attempt to steal second if down by one, or intentionally walk a batter to set up the possibility of a double play. Okay, Lindsey thought, but by how much do such moves alter the chances of scoring? And, depending on the inning and score, how do they increase or decrease the probability of winning the game? Many fans might have harbored similar questions, but there were no Internet databases or computerized play-by-play data to consult. Only Lindsey, with his father's help, actually scored more than 1,200 games and spent hundreds of hours interpreting the sheets in order to finally find out. "G. H. Hardy once said, 'Here's a toast to mathematics—may it never be of any use to anyone,' " Lindsey said with a smile. "I wanted my mathematics to be of use to somebody, whether it's a baseball manager or a war general."

Lindsey's father gladly enlisted in this effort. Freshly retired from the service, Charles Lindsey had kept track of most of the 782 games for "Progress of a Score." He then kept in even greater detail the play-by-play of 373 more, comprising some 6,399 half-innings, on special green quadrille graph paper and in custom notation he and George had devised. In painstaking, loving penmanship, Charles scored every ballgame he could get his ears on—his wireless could pick up broadcasts from as far away as Wheeling, West Virginia—often two whole games a day, one in the afternoon and one at night. Charles delivered the scoresheets in bunches for George to sit down and decipher.

George wasn't even on the same continent when he pored over these sheets. By this time he, June, and their two young children had been transferred to the village of San Terenzo in Italy, living in a gorgeous hilltop villa. Not that George looked up from his baseball work much to notice. "I have to do everything around here!" June would complain. "When I need any help, you're doing baseball all night, working on these papers, these statistics!" George's military commanders at the antisubmarine research center run by NATO, many of them European, were even less impressed, especially his straight-laced Norwegian supervisor. "To study something called baseball, which he'd never even heard of, was unbelievable," Lindsey recalled. "He strongly disapproved of it. Then he thought, 'If this guy writes it and it's published, and it says that he was in the SACLANT ASW Research Center, I'll be the ridicule of Norway.' "

Lindsey kept a low profile. He carried bundles of his father's scoresheets, some still redolent of Lake Huron, back and forth from Italy and for two years went though every batter on every sheet to track scoring patterns. Each time a hitter came up with a man on second and one out, a tick was marked in the appropriate box on a master sheet; obviously, he had no computer to handle the computation, though back then, this heavy lifting was half the fun. The sheets filled up with so many hundreds of ticks, all in neat groups of five, that they resembled those of a prisoner recording days on his cell wall. But each one got George closer to the point where he could count everything up and report his results.

Much to the relief of his wife and young family, that time finally arrived in the summer of 1963 when Operations Research published Lindsey's seminal 24-page article, "An Investigation of Strategies in Baseball." Lindsey's study dove head-first into the following:

• Expected runs in any situation: exactly how many runs scored depending on the number of outs (none, one, or two) and any runners on the bases (empty, man on first, bases full, and so on). For example, with runners on second and third and one out, no runs were scored 27 percent of the time, one run 24 percent, two runs 28 percent, and more than two 21 percent. The average total runs scored after each of the 24 possible situations:

• The sacrifice bunt: The above chart clearly shows how moving a runner over by giving up an out can help or, as often is the case, hurt a team. For example, with none out and a man on first, teams score an average of .81 runs. A successful bunt would leave the runner on second with one out, which typically results in just .67. A sacrifice becomes worthwhile only when a club needs one and only one run late in a game. (This assumes an average hitter both at the plate and on deck, however, and real-life managers would surely take this into account. But the concept as Lindsey asserted it was both groundbreaking and irrefutable.)

• The stolen base: At what success rate is it worthwhile to attempt to steal second, say, with one out? Using the chart, the benefit (1.19 expected runs versus .81, or in increase of .38) must be weighed against the risk of getting called out (and winding up with .24 expected runs, or a decrease of .57). Simple algebra puts the breakeven point at 59 percent. Once again, however, the state of the game is significant, and trailing by only one run can make stealing an advisable risk. (This no doubt could have provided some solace to Babe Ruth, who, with his New York Yankees down 3-2 to the St. Louis Cardinals in the bottom of the ninth inning, was caught stealing second to stunningly, and controversially, end the 1926 World Series.)

• The infield in: With the bases loaded and none out, the defensive team can choose to either play its infielders deep, in hopes of turning a double play (one run scores, man left on third with one out), or play them in and nail the lead runner at the plate (no runs score, but the bases are left loaded with one out). Teams almost always should go for the double play, Lindsey discovered. Only when a team is tied or ahead by one run late in the game, with none out, will playing the infield in pay off on average.

• The intentional walk: often used with runners on either second or second and third to set up a possible double play. This maneuver, Lindsey determined, more often than not increased the expected number of runs that would score in the inning. (Double plays, however tantalizing, don't happen often enough.) In real life, of course, a manager will walk a good hitter to get to a less dangerous one, but Lindsey's model could not account for different types of hitters. He foresaw a world when that could be possible, though: "The concept of an electronic computer in the dugout is a distasteful one," he wrote, "but, if progress demands it, this is the type of calculation for which it could be programmed."

"An Investigation of Strategies in Baseball" concerned mostly managerial maneuvers. But Lindsey ended his paper with a strikingly ahead-of-its-time application of his Expected Runs Matrix, that being to evaluate individual hitters. Batting average, he asserted—as F.C. Lane had 50 years earlier in Baseball Magazine—drastically undervalued extra-base hits, while slugging percentage, which arbitrarily weighted long hits by their number of bases (doubles were worth two, triples three, and so on), overvalued them. By manipulating the chart algebraically, Lindsey determined that a single was worth on average .41 runs, a double .82, a triple 1.06, and a home run 1.42. (These values were, in fact, not terribly different from Lane's more rudimentary calculations from way back in 1917.) Lindsey applied his findings to a recent controversial trade between the Tigers and Indians. The Tigers' Harvey Kuenn, a singles hitter who had just won the 1959 American League batting championship at .353, had been dealt straight-up for the Indians' Rocky Colavito, who had batted just .257 but tied for the league home run lead with 42. Who was the more valuable hitter, in terms of the runs generated by the weight values of each player's hits? Even though Kuenn had hit virtually 100 points higher in batting average—at that time still by far the most popular measure of batting ability—it was Colavito, 114.5 runs to 112.6.

This new analysis did not shake baseball; it caused nary a tremor. No teams contacted Lindsey. No journalists wrote a word. His Norwegian superior got over it, too, because soon after "Investigation of Strategies" was published, Lindsey and his family moved back to Ottawa, where George, his curiosity in baseball strategy fulfilled, returned his focus to more military matters. He remained with the Canadian Department of Defence, and later became head of the Operational Research and Analysis Establishment, before retiring in 1987. He continues to handle selected projects on strategic nuclear weapons and ballistic missile defense.

Lindsey's baseball work is all but forgotten (if it was ever known to begin with). His name is barely recognized, if at all, by even the most knowledgeable fans of baseball statistics. His papers appeared only in obscure journals, a full 20 years before Bill James's Baseball Abstract series popularized what we now call sabermetrics. Still, as we examine Lindsey's work today, it goes down as the first time anyone compiled reams of play-by-play data, applied sophisticated statistical analysis to weigh various strategies, and published the results.

Lindsey, for one, didn't realize how groundbreaking his work truly was. "It was really just out of curiosity," he admitted, sitting on the couch of his Ottawa home. He got up and went to his attic to fetch the old scoresheets, which he kept all these years. Flipping through them for the first time in decades, his enthusiasm bubbled anew. He looked through his reading glasses, settled far down his then 82-year-old nose, and remembered all those nights scoring games with his father and vexing his wife—all to use baseball statistics in a new and exciting way. He wore the blue blazer of his old Operations Research division, whose crest on the left breast pocket featured a Latin phrase, A Posse Ad Esse: "From the Possible to the Real."