Double Plays are more heavily contextual than any other statistic that we have to deal with here—MUCH more heavily contextual than home runs, for example.   The number of double plays that each team turns is usually, in the majority of cases, more or less just an expected outcome of (a) the number of opposition runners that the team has to deal, and (b) the number of ground balls by the pitching staff.   In 2019, for example, the Cleveland Indians had an expectation, based on their runners on base and ground balls, of turning 113 double plays, and actually turned 111.   The Houston Astros had an expectation of turning 98, and turned 97.  The Kansas City Royals had an expectation of turning 157 and actually turned 152.  The Twins had an expectation of turning 128, and actually turned 129.   The Oakland A’s had an expectation of turning 122, and actually turned 123.  The Arizona Diamondbacks had an expectation of turning 141, and actually turned 138; the Cincinnati Reds, 125 and 125; the New York Mets, 124 and 129.

            Seven major league teams missed their expectation by 15 or more.  The Red Sox, playing a rookie third baseman at second base much of the season, had an expectation of 138, but turned only 115; their -23 performance was the second-worst in the majors.  The Padres missed by 24, 125 to 101; I don’t know why.   The Cardinals, with Gold Glove winner Kolten Wong, were the best in the majors,  turning 170 double plays against an expectation of 143.   The only other team that was +20 was the White Sox, 170 to 150. 

            Because the double play data is essentially created not by the ability of the fielders as much as by the situation in which they play, we can’t approach Double Plays the way we have approached everything else.  We have to use a more circuitous approach.

            Here’s what we’re going to do.  First, we need to calculate the expected double plays for each team.

            Second, we compare each team’s actual to expected double plays, by simple subtraction. 

            Third, we re-state the double plays relative to 100; in other words, if you have an expectation of 130 double plays and you turn 125, we state that as 95.  We’ll call that CNDP, for Context-Neutralized Double Plays. 

            Fourth, we look at the standard deviations for each era, and convert the SDP into a Standard Deviation number, as we did before, for the other categories. 

            Along this march, there are all kinds of different teams that might wind up as the #1 team, and at this time I really don’t know where I’ll wind up, when I get to the part of the study where I am actually using the data to try to figure Runs Saved.   The Philadelphia A’s of 1949 still hold the record for double plays in a season, with 217.  The 1966 Pirates had 215.

            Compared to expectations, however, the 1-2 teams of all time are the 1941-1942 Yankees, Joe McCarthy teams.  The 1941 Yankees turned 190 double plays against an expectation of 133.4, which makes a context-neutralized score of 156.6.  The 1942 Yankees are second, at 149.9. 

            There are two more issues, however.  First, there is the issue of adjusting this by the standard deviation of the era.  For the 1940s, the standard deviation of CNSD numbers is 18.04, which is significantly higher than the Standard Deviation for any other decade.  Assuming we use that figure, the 1941 Yankees would be 3.1 standard deviations better than the period norm. 

            It is not clear, however, whether there is actual and meaningful variance in the standard deviation over time.   We have taken out the normal up-and-down cycles in the category by getting rid of the raw totals, centering everything at 100.   For most of history, while there is very significant fluctuation in raw DP totals, there is little change, decade to decade, in the standard deviation of the CNDP numbers.   The standard deviation of the numbers, centered at 100, is 12.5 in the 1900s, 14.0 in the 1910s, 15.9 in the 1920s, 16.0 in the 1930s, 18.0 in the 1940s, 15.6 in the 1950s, 14.0 in the 1960s, 13.9 in the 1970s, 14.5 in the 1980s, 12.6 in the 1990s, 13.4 in the 2000s, and 13.9 in the last decade. 

            There is not a LOT of difference in the standard deviation over time, point (a), and, point (b), it is not logically clear why there would be a larger difference in the 1940s.  The peak in the 1940s may have occurred largely because the 1940s Yankees were so outstanding at turning the double play that they stretch out the chart, creating a higher standard deviation by their own actions.   I might do something weird like use 30-year standard deviations, rather than 10-year standard deviations. 

            But using the decade as the standard and using the practice described before, these would be the top Double Play performances of all time:

            There’s an implicit assumption there that the second baseman is the key player, which is not necessarily true; the shortstops here included Rizzuto and Francisco Lindor, who may have helped a little bit.   Anyway, let me try to illustrate the problem of using the 1940s standard deviation in this way.   Later in this process, I will be estimating how many double plays each team turned ABOVE THE LEVEL OF MINIMAL COMPETENCE.    Let us say that we set the level of minimal competence at five standard deviations below the norm.  If the norm is 100 and the standard deviation of relevance to the 1940s is 18, then the level of minimal competence would be 28, or four standard deviations below 100.   Rizzuto and Gordon, 1941, have a CNDP (context-neutral Double Plays) of 157, so they would be +129 double plays in 1942, the highest figure of all time.  That’s fine; I don’t object to their being #1 of all time.  But these would be the Top 6 of all time: 

            All from the 1940s.   Nine of the top 12 would be from the 1940s. Using the High Standard Deviation from that era—18—pushes the "minimum competence cutoff" down to 28, which increases the margins for the best teams.   Probably artificially, but I can decide that issue later. 

            So I have to explain to you how we determine the Expected Double Plays turned by each team.   We start by making an estimate of the number of runners on first base against each team.  The number of runners on first base is

            Hits allowed by the team

            Minus Home Runs,

            Multiplied by the league percentage of non-home run hits which are singles,

            Plus walks,

            Plus Hit Batsmen,

            Minus Sacrifice Hits allowed,

            Minus Wild Pitches,

            Minus Balks,

            Minus Passed Balls. 

            It’s just an estimate; it’s not precise.   Anyway, you figure the ERO 1b (Estimated Runners on First Base) for each team, and then add those up to get the league total.    Let us say that the Washington Nationals had 1,436 Estimated Runners on First Base, and that the National League total was 22,341.   1,436 divided by 22,341 is .06426, which is a little bit less than 1/15^th, since the Nationals’ had fewer opposing runners on first base than the average team. 

            In the National League there were 2,042 double plays turned.   .06426 times that number would be 131, so at this point we would be expecting the Nationals to turn 131 double plays.  

            But we have yet to adjust for Ground Balls.   We don’t know EXACTLY how many ground balls there were against each team, but (a) more than 85% of ground ball outs include an assist, and (b) more than 85% of assists result from ground balls.   Assists may be used as a stand-in for ground balls.  So then you have to figure the Assists/Batter’s faced for the Nationals:

            1373 /  6134  =  .223834

            And do the same thing for the league as a whole:

            22464 / 93448 = .240390

            The Nationals, because of Strasburg and Scherzer and company, had fewer ground ball outs than the typical team, so we have to adjust their expected double plays by that amount. 

            Ex DP =   Previous ex DP * Team Assist Rate /  League Assist Rate * .994.

            Which is

            131 * .22834 / .240390 * .994

            Which is 121.  (You have to reduce the raw estimate by .006 because high-assist teams tend to also be high-runners-on-first-teams, which causes the league total, if figured team-by-team, to be slightly higher than the team total if figured on the league level.)   The Nationals actually turned 112 double plays, so they were somewhat below average in turning double plays, not a lot below average, but a little.

            Very often, the number of DPs a team is expected to turn is within a small margin of the number they actually turn.   53% of teams turn a number of double plays which is within 10 of their expectation, but then, if you just predicted that every team would turn 135 double plays, that might still be true, I don’t know.   But here’s the thing:  if you take ALL of the teams which project to turn 120 double plays on the season, on average they will turn 120 double plays; some more, some less, but on average they’ll get there.  If you take all the teams which could be expected to turn 160 double plays, they will turn 160 double plays.   That’s how I know the system works.   Thanks for reading.