It’s important to own when you’re wrong. Here’s one, no two, such instances when I was wrong. In each of these articles I attributed a result to an innovation without first checking the data. From the first piece:
Consider something as subtle as the webbed glove. Its introduction in the early 1920s coincides with a rise in double play rates and the resultant swapping of second and third basemen on the defensive spectrum. It also helped cut down error rates, driving down unearned runs. One might argue that it allowed more athletic players to get at more balls because they no longer required two hands to corral the ball. Error avoidance becomes less important when webbing provides extra insurance against fumbles. Cardinals pitcher Bill Doak is widely credited with introducing the idea of the webbed glove, and we’ll be considering him.
From the second:
Even at the equipment level, the first webbed gloves were patented [in 1920], ushering in a dramatic change in how teams aligned their defenses. Second basemen and third basemen began a process of switching places on the defensive spectrum as webbed gloves supported more double plays.
The lesson here is to avoid making sweeping statements before looking at the data. Once I did, however, I realized that my conjectures, which I’d phrased not as claims but as facts, did not match what the record tells us. Which, by the way, doesn’t mean I’m wrong. It means I can’t support a seemingly simple statement with the known data. But if we want to avoid parsing, I’m just wrong. And that’s OK, because I learn best from making mistakes, not from getting things right. So let’s take a look at why I’m making this mea culpa today.
Last week, Miller and I met up to see Bill James give a talk on the Cape. Naturally, we also talked a great deal about pioneers and execs since we’re in the midst of electing them. Bill Doak came up again and again as a pioneer for his webbed glove. We needed more information to see how viable that made him. If, as I’d believed, the webbed glove caused a spike in DP rates and a cliff-dive in error rates, then that’s a big freakin’ deal.
The data, however, didn’t cooperate.
Actually, at first it did. I looked at the percentage of double plays per all balls in play that didn’t result in an extra base hit. Here’s the two decades surrounding Doak’s invention through that lens:
***Webbed glove begins production***
You can see this in the red line on my first chart below.
The decade of the 1920s shows a nice little spike. There’s an obvious uptick of nearly a full percentage point between 1910 and 1930, or what would represent about 35 double plays annually per team. But can we attribute that just to the glove? In a word, no. See the problem is that after 1913, steals declined precipitously. In 1913, the league stole 3257 bases and was caught an additional 2593 times. That’s 5850 attempts, and that was down a couple hundred from previous years. By 1916, MLB teams attempted “only” 4838 steals (1300 more than today’s 30-team leagues try for). In 1920, with offensive totals skyrocketing, the league wisely responded by cutting back on attempts, with a resulting plunge to 3256. By decade’s end, the majors had whittled that number to 1882 attempts, roughly the same rate as today’s teams. The steal is the enemy of the double play, which means that MLB as a whole created more DP opportunities simply by changing its strategy to encourage more big innings. (Sacrifices rates didn’t budge, oddly enough.) So I needed to look closely at the interaction of small ball and the double play.
Next I looked at DPs per opportunity, net of various means of advancement: steal attempts (including pickoffs and balks), sacrifice attempts and sac flies, wild pitches, and passed balls. That is how often the league turned DPs when they had a runner at first, didn’t give up an extra base hit, and nothing happened from an advancement perspective to eliminate the DP opportunity. This bulleted list should tell the story:
Percentage of DPs turned per net opportunities:
***Webbed glove begins production***
It is certainly possible to squint and see that DPs were turning at a higher rate in the mid and late 1920s. But it’s equally possible to see that the rate had begun rising dramatically in 1916, had dipped during World War I, and recovered by 1921, whereupon it began to climb once again. DP rates measured this way stabilized in the 1930s between about 8.7% and 9.3%. In the 1940s they gradually edged up to a range hovering around 9.5%. Then in 1950, they shot up over 10% where they’ve stayed since then, never topping 11% and only dipping below 10% three times in the late 1950s and early 1960s (each time at 9.9%).
If Doak’s glove did support the increase that led rates to jump from the mid 8% to 9.3% range, then we’re talking about another 15 twinkills per team. That’s not insignificant, of course. But it’s hard to give Doak’s innovation a lot of credit for increasing DP rates when the rate of acceleration simply followed the long-term trajectory in DP rates. But that’s not the only reason DP rates could have increased. The elimination of the spitball and new instructions to umpires to replace dirtied baseballs more often meant that fewer balls could be described as soft, lopsided, or soggy. It is possible, therefore, that the exit velocity of a ball off the bat may have increased. We can’t say for sure since we don’t have that information from the 1920s, but the increase in extra base hits during the early 1920s may attest to that being the case. If true, it could also mean that the ball reached fielders more quickly, which enabled them to the turn the deuce more often. In other words, there’s no smoking Bill Doak gun here. You can see this in the blue line in the table I’ve presented with this article.
[Note: We are missing SB, CS, SH, and pickoff data for many early seasons and SF data through 1953. Rates prior to 1890 aren’t very reliable for this reason. In 1890 we get SB and CS. In 1894 we get SH as well.]
What about errors? Did the Doak glove improve fielding percentages? The answer here is probably not. In the second chart, you’ll see the reduction in error rates per balls in play over time. This time I’m defining balls in play in the usual way (AB – HR – K + SH + SF).
Looking at 1894 on the chart (the first year that we have league-wide SH numbers), we see that error rates dropped from about 9% of all BIP to about 6% in 1910. By 1919, that number had reached 4.6%. In 1920, it dropped to 4.4% and dipped under 4% in 1923. From there it bounced a little but slowly edged down to 3.8% by 1930. During the 1930s it gently bobbed around 3.7%, and after the war drifted down toward 3%. By the late 1970s it reached under 3%, and today it’s under 2.4% annually. If you look at the chart, you’ll see that the curve on error rates is essentially a reverse J-curve or learning curve (thanks, Hermann Ebbinghaus!). The league is learning to field better and better, though the degree of that curve is diminished over time. But it’s very difficult to look at that chart and ascribe any major significance to the Doak glove.
In Win Shares, Bill James questions why the defensive spectrum jumped, causing second and third basemen to switch spots. One of the biggest statistical data points associated with the jump is the rate of DPs to errors. Until the 1930s, the league made more errors than it turned double plays. But once that ratio flipped, second basemen became more important than third basemen defensively. If the Doak glove had something to do with that, then maybe Bill could yet emerge as a key candidate for a pioneer plaque. As the green line on the first chart shows, however, the data don’t prove out. The big jumps in DP/ERR rate are in the late teens and late forties, not the 1920s.
So, I feel bad for leading everyone down the garden path regarding Bill Doak. But I feel worse yet for Bill, himself. The webbed glove is clearly a cool invention and must have had a salutary effect on defense, but there’s just no way that I’ve got to show it. The only other avenue I could think up would be to see if error rates or DP rates on batted ball types changed immediately after 1919, but much of that data is yet obscure. If you’ve got ideas, or better yet your own research, let us know in the comments.