I. Adjustments to the Speed Score Methodology
II. Changes in Speed in Baseball Over Time
Since I did the work that was reported here on Monday (The Fastest Player in Baseball) I have revised the Speed Score formulas. The Speed Score system is intended to be centered at 5.0. …an average player is at five. But I invented the formula in the 1980s, with the very limited computer resources and with the very limited data sets that were available to us in the 1980s, and I based the system on the norms prevailing in the 1980s. In retrospect, and with the ability to work with much more data, I see the problem: the 1980s were not typical in this respect. There was more speed in the game in the 1980s than at any other time since before Babe Ruth. What was typical then is not typical now, and has not been typical through most of baseball history. The result was that the system—intended to be centered at 5.0—was in fact rating the average player at about 4.45.
Well, OK, how do we fix that? Since I was working with five different speed indicators, I looked at the averages generated by each of the five. Some of them were a little over 5.0, some under, but the first indicator—what I call S1—was way off. S1 has to do with Stolen Base Percentages. S1 was delivering an average output of 3.0, thus dragging the system off-center. I changed that formula, then, from this:
{[(SB + 3) / (SB + CS + 7)] –. 40} X 20
To this:
{[(SB + 4) / (SB + CS + 7)] –. 35} X 20
Not to exceed ten, not to be less than zero (as was the rule before.)
Suppose that a player was 5-for-8 stealing bases. The formula before treated him as if he were 8-for-15, 53%; subtract .40, you’ve got .13, multiply by 20, you’ve got 2.6—a very low score. Now we treat him as he were 9-for-15, 60%, subtract .35, you’ve got .25, multiply by 20, you’ve got 5.0. If a player goes 5-for-8 stealing bases, that indicates average speed.
The second factor, what I call S2, was also a little off-center. That formula has to do with the frequency of stolen base attempts. The formula for that was.. ..maybe I should write this out in English.
1) Divide the player’s Stolen Base Attempts (Stolen Bases and Caught Stealing) by the number of times he was on first base (Singles + Walks + Hit By Pitch).
2) Take the Square Root of that.
3) Divide by .07, but with the result not to exceed 10 or to be less than zero.
Let’s say a player is on first base 80 times and attempts to steal 10 times. That’s .125. The square root of .125 is .3536, call it .35; that number divided by .07 is five. Easy enough; it just looks complicated when you write it in a formula.
I changed the process to:
1) Divide the player’s Stolen Base Attempts plus one (Stolen Bases and Caught Stealing+1) by the number of times he was on first base plus nine (Singles + Walks + Hit By Pitch+9).
2) Take the Square Root of that.
3) Divide by .05.
4) Subtract 1.3, but with the result not to exceed ten or to be less than zero.
Let’s say that a player is on first base 80 times and attempts to steal 10 times. Divide 11 by 89; that’s .1236. The Square Root of .1236 is .3516; divide that by .05 and you have 7.0. Subtract 1.3 and you have 5.7.
The addition of the one "phantom stolen base attempt" in nine phantom times on base makes the system center near 5.0 for players with very little playing time. As the system was before, if a player was on base once and did not attempt to steal, he would get a Score of 0.0 by this method—an indication that he was extremely slow. A better read on that data is "no information" or "no reliable information"—thus a score around 5.0. The other changes move the scale up a little bit, centering it near five.
The third element of the Speed Score method (S3) has to do with the frequency of triples. I made a similar change there, changing this:
(3B)/ (AB – HR – SO)
To this:
(3B + .1) / (AB – HR – SO + 10).
Adding one-tenth of a phantom triple and ten phantom at bats. This, again, creates a score near 5.0 if a player has very few at bats, rather than giving him a score of zero in the category if he had two at bats and didn’t hit a triple.
Making these adjustments changes the average speed score in a season to something much closer to 5.0, as well as making the system work a little bit better for players with limited at bats, although I would still strongly recommend not using it for players with very few at bats.
Reviewing how these adjustments change what I told you yesterday. . .these adjustments, for reasons I don’t quite understand and don’t much care about, change the designation of the slowest player ever from Gus Triandos to Ernie Lombardi. They change the score for Willie Wilson in 1980 from 9.5 to 9.9 (9.88), which is fine. …if it’s a zero to ten scale, somebody should be close to ten. They make some other little adjustments to the lists. Bengie Molina replaces Toby Hall as the slowest player of 2005; Wilson nudges ahead of Omar Moreno as the fastest player of 1981. .. .that sort of thing. More than 90% of the fastest and slowest players are the same, and the others are just little changes that don’t need to be recited, except that I will note that Adam Dunn replaces Chris Snyder as the slowest player of 2011.
OK, that out of the way, we can proceed to the next question, which had to do with changes to speed in the game, over time.
We could, of course, normalize everything so that the average speed score was the same in 1987 as it was in 1958; it’s always 5.0. That’s easy enough to do, but it’s wrong. We "normalize" or adjust for illusions in the stats. The differences in speed between players in 1987 and 2005 are not an illusion of the stats; they’re merely the stats reflecting real changes in the game. You don’t normalize out of existence that which really happened.
In 1950 the average speed score for a major league regular (350 or more plate appearances) was 5.097, but this figure is a little bit hinky because there is no caught stealing data in the National League in 1950. The National League started counting caught stealing in 1951.
Anyway, speed was going out of the game in the 1950s; the average speed score was going down almost every year. In 1958 the average speed score of a major league regular was 4.58—the lowest on record. With Billy Bruton hurt, there were only four major league regulars in 1958 who could really be considered fast runners—Luis Aparicio, Willie Mays, Richie Ashburn and Don Blasingame. Other guys, like Bill Virdon and Johnny Temple, were faster than most other players of 1958, but they really weren’t fast. Roger Maris ranked as the 13th fastest regular of the 1958 season.
Beginning in 1959, speed started to crawl back into the game; sorry if that’s a mixed metaphor. Speed in the game increased gradually until 1987, when the average speed score of a major league regular reached a peak of 5.17. In 1987 there were 29 players in the major leagues who were certifiably fast (Speed Score 7.0); in 1988 and 1989 there were 30.
By 1987 the seeds of the steroid era had been planted; the Boys by the Bay had discovered Power Pills. From 1987 to 2005 the average speed score trended steadily downward, reaching a low of 4.67 in 2005. In 2005 there were still 18 regular players in the majors who were very fast, but there were 32 who were very slow.
Since 2005 the average speed score has recovered somewhat; speed has begun to crawl back into the game once again. In 2011 the average speed score was up to 4.84, and the number of fast players (25) almost balanced the number of players who were problematically slow (26).
That was the overview; let me repeat that with a little more detail. I considered any player with a speed score of 7.0 or higher to be problematically fast (so fast that it was a problem for the defense), and any player with a speed score of 3.0 or lower to be problematically slow. In 1955, with 112 players getting 350 or more plate appearances, only four of them were really fast (Bruton, Mantle, Mays and Gilliam), but only two of them were really slow (Walt Dropo and Dick Groat.) There were a lot of slow players around, but most of them were bench players and platoon catchers who didn’t get to 350 plate appearances.
The average speed score reached a nadir of 4.58 in 1958 and then started up, but it actually went straight up only until 1962, which was the year Maury Wills stole 104 bases. The expansions in 1961-62 and 1969 kept older players around and brought up marginal prospects. The average speed score went down after 1962, recovered slightly, and then went down again after the big expansion in 1969. By 1971 the average speed score was down to 4.65. The DH Rule in 1973 pushed it down again, down to 4.59. The average player of 1973 was only a tick faster than in 1958.
In the mid-1970s people began saying that artificial made the game faster, which may or may not be true. You can run faster on turf than you can on grass—the hard-surface turfs of the 1970s, at least—and almost every player who stole bases had a meaningfully higher stolen base percentage on artificial turf than he did on grass fields. Speed scores moved up rapidly in the mid-1970s, reaching five in 1976 and 5.14 in 1977. After that average speed scores were flat for about ten years, reaching a new peak of 5.16 in 1987, and stayed over 5.00 until 1995. In the heart of the steroid era the number was going down every year, reaching the low of 4.67 in 2005.
Since 2005 it has gone up every year but one, and we’re back now to 4.84. If this trend continues—which is anybody’s guess—we might see speed back at the levels of the 1970s in about eight more years.
In 1955, as I mentioned, there were only two regulars who were problematically slow. This number, which was a little bit of a fluke, went up to 8 in 1958, to 10 in 1965, and to 16 in 1971, in large part because of expansion. Then it went down; in 1980 there were only 9 major league regulars who were truly slow. That number grew steadily through the 1980s and through the steroid era, reaching a peak of 32 slow players in 2005. This number dropped to 22 by 2009, although it was up to 26 in 2011.
Fast players. . ..in 1959 there were only four. By 1963 there were 9 (Lou Brock, Maury Wills, Luis Aparicio, Jake Wood, Willie Davis, Vada Pinson, Julian Javier, Tony Taylor, Jose Tartabull and Curt Flood.) By 1967 there were 10, by 1972, 13, by 1975, 17, and by 1977, 26.
This reached a peak of 30 in 1988-1989, then declined. In 2003-2004 there were 16 fast players in the major leagues, which is still four times as many as there were in 1958-1959, but then there were also nearly twice as many teams. Speed has never entirely gone out of the game, and it is now on the way back.
The changes in speed in the game over time are less dramatic than some people would suggest. The speed of the average player probably really doesn’t change very much over time. When you think about it, it can’t. Speed could only change dramatically over time if speed was the controlling factor in whether you wanted a player on your team or not. Speed is some factor in whether you want a player on your team, but speed competes with hitting ability, throwing ability, position-specific defensive skills and locker room intangibles. No matter how fast Carl Everett might be, you’re going to get tired of having him around. No matter how fast Billy Hamilton might be, or Jarrod Dyson, he’s not going to play if he doesn’t hit .250. No matter how fast Larry Lintz might have been, there’s a limited role for a player who never hits a double. Since speed competes with these other virtues to make a complete player, the speed of the average player is, in a sense, locked in place by the competition.