Sabermetrics Is The Revival Of An Old Discussion (Abolishing Fielding Percentage, 1877-1886)

[Draft Post, October 1, 2018]

The Newsletter of the Official Scoring Committee
Society for American Baseball Research (SABR)
Volume 2, Number 2, June 2017

Nineteenth Century Sabermetrics: Range Factor by Richard Hershberger

The National Association, the only professional league apart from the National League, in its convention of February 1879 voted to abolish the error column from the official scores of its games. What could possibly have motivated such a bizarre action? It turns out that this was a fairly mainstream idea at the time.

Baseball statistics underwent, as is well known, a revolution in the late 20th century, with the effects still being worked out. One common theme in the controversy (now mostly, and blessedly, past) was that the traditional baseball statistics—batting average, earned run average, and so forth—are straightforward, or even obvious metrics. Many on both sides in the debate agreed on this, while disagreeing whether this was a virtue or a failing.

The claim never really stood up to scrutiny. Batting average, the ne plus ultra of traditional stats, appears to be simplicity itself: Hits divided by At Bats. In reality a lot of complexity is hidden in those two terms. Hitting the ball and getting on base does not necessarily mean the batter got a Hit, nor does a turn at bat necessarily constitute an At Bat.

This observation merely scratches the surface. When we look closely at the history of the traditional statistics, they turn out to be the product of decades of discussion and experimentation by trial and error. They may seem obvious to us today, whether through long familiarity or simple hindsight. Either way, they were anything but obvious to even the closest observers of the day. The modern sabermetrics revolution is, it turns out, not a new phenomenon after all. It only seems that way because the discussion and experimentation had died down through much of the 20th century. Sabermetrics is not a new discussion. It is a revival of an old one.

Stew Thornley, our committee chair, has graciously invited me to write on the abstruse topic of scoring in the 19th century. My aim is to show the discussions and experiments that eventually led to the traditional statistics. My hope is that this will be a series, however irregular, and dependent upon your patience and tolerance. I begin in 1879 in media res to show a path considered, but not taken: the elimination of the error from scoring, and a surprising end to that path: the invention in 1886 of Range Factor.

The error was already an old stat by 1879. It went back to the 1850s, as a disapproving cluck of the tongue at the errant fielder. The 1860s saw the rise of the base hit as a stat. The error came into prominence with it, taking on a new and important role of ensuring that the batter did not get any undeserved credit. The newly prominent error was then in the early 1870s reapplied to fielding in a more systematic manner, resulting in the Fielding Average.

It was with Fielding Average that problems arose. Fielding Average was early recognized as an imperfect tool. Here is a discussion that nicely states the problem:

The sharp bounder between first and second base, that Gerhardt or Dunlap would field in a majority of cases, would be a safe hit were some other player on second base. The question then arises whether it is justice to Gerhardt or Dunlap to charge them with an error when they fail to stop such balls, while a lazy or indifferent second baseman allowed them to be scored as base hits by making no effort to stop them. The same is true of every other in-field position. A hard hit grounder past third base may, by the exercise of great agility, be stopped and thrown to first base in time to retire the batsman. The fielder gets credit for an assist only, no matter though he make the brilliant play a half dozen consecutive times. The seventh time he fails and is charged with an error, while a less agile baseman would fail to make an error, even, and the seven batsmen would score base hits. This is a manifest injustice. Base hits should depend upon the merits of the batsman, not upon the demerits of the fielder. If the league managers can frame any rule to rectify this error, they should do so. (Detroit Free Press, October 18, 1881)

The writer's challenge for a rule to rectify the problem went unmet until a century later Bill James invented Range Factor. There also was an early recognition of the subjectivity of scoring errors, and its susceptibility to homerism:

A pitcher would be charged with earned runs and base-hits against him by one scorer, while another would charge the field with the errors, thereby relieving the pitcher. In fact, this error business is ... ill defined in its rules ... (New York Clipper, February 8, 1879)

Some action should be taken in regard to official scorers. They are appointed by the club managers, and are generally, no doubt, moral young men, who want to secure a dead-head ticket to the games; but ... [i]n plain words, official scorers are liable to stretch their elastic consciences in favor of their home club, and will continue to do so until there are some fixed and definite rules for their guidance. (Detroit Free Press, October 18, 1881)

For all the failings of Fielding Average, and the error tabulation underlying it, it was the best fielding metric they had. It was, absent anything better, generally considered the best way to assess a fielder, and negotiate pay accordingly. People respond to incentives. Some players adopted the simple stratagem of only fielding balls they were sure they could handle. These were known as "record players" and widely condemned, even as the incentives to record playing remained in place. The solution is to change the incentives—to create statistics that better reflect team play. Many innovations, such as scoring sacrifice hits, had improving incentives as the underlying goal. This was the background to the proposal to eliminate errors. Here is an early proposal for the elimination of the error:

The abolition of the error columns. Bold, daring fielding on the part of every fielder would liven up the game twenty per cent. Base ball patrons will remember how, at times, a remarkable play by a fielder in taking great chances has enthused the spectators and given vim to the sport. But, with the error column staring them in the face, alas, most players take but few wide chances. Nothing is so disgusting to a crowd of lookers-on as to see a player shirk a difficult play when it is patent to all that he feared there were too many chances for an error against him to induce him to attempt the play. With no error record to go against him no chance would be slighted by a player, for then he would have every thing to gain if he made the play and nothing to lose if he failed. Give him the benefit of his assists and put-outs as usual, but demolish that demoralizing factor, the error column. (Cincinnati Enquirer, December 4, 1877)

The idea was a regular topic of discussion among scoring aficionados. In the end nothing came of it. The National Association was well into its tailspin into oblivion and already was irrelevant. Newspaper reports of its games often followed the traditional practice and included errors.

The National League seems never to have seriously considered the idea. The idea popped up from time to time through the 1880s, but had acquired the status of "old chestnut"—a theoretical notion to be chewed on in the winter months, but not a practical proposal.

I wrote earlier that Fielding Average was the best measure available until Bill James invented Range Factor. This is not quite true. The flamboyant and contentious sportswriter O. P. Caylor was the leading advocate of abolishing the error. Here he runs the idea up the flagpole in 1886. His proposal doesn’t stop with eliminating the error. He has a positive proposal for its replacement:

Do you ask what I would have instead of the error column? This: I would give every fielder credit for all he did—every assist and every put-out—without recording his failures. Then every fielder would be interested in taking every chance, however desperate, without fear of loss by doing so. I would then make out the players' averages by the number of assists and put-outs he had, divided by the number of games he played, and compare every man's record only with the record of the other fellows of his position. (The Sporting Life, February 3, 1886)

Put-outs plus Assists, divided by Games played: this is Range Factor, invented by O. P. Caylor in 1886. Caylor in 1877 was the baseball writer for the Cincinnati Enquirer. It is entirely likely that he wrote the proposal to eliminate errors previous quoted. His thinking had progressed over the ensuing decade. In 1886 he had a flash of genius. Sadly, the idea was both after its time and ahead of it. Had he worked out Range Factor in 1877, when the idea of eliminating errors was a viable proposal, Range Factor might have been adopted and grown beloved over the years, finding its way onto the back of baseball cards. As it was, the idea disappeared almost as soon as it appeared, not to be seen again for nearly a century.

Here we have a road not taken. While a missed opportunity, it shows the scope and the sophistication of the discussions. In later installments I will look at the roads that were taken and why.

Richard Hershberger is the author of Strike Four: The Evolution of Baseball (Introduction by John Thorn).