# Did stealing signs help the 1951 NY Giants?

JEL C10 * D00 * D80

In his recent book, The Echoing Green, Joshua Prager suggests the legendary 1951 pennant race between the New York Giants and the Brooklyn Dodgers might have been affected by the Giants' stealing signs from opposing catchers. Prager documents, beginning July 20, a Giants coach used a telescope to view the catcher's signs and a bell to relay the signs to players seated in the Giants' bullpen. A player in the bullpen then signaled information on the upcoming pitch to batters by sitting still to indicate a fastball or moving to indicate a curveball, for example. Consequently, Giants' batters would have information on forthcoming pitches and presumably have more successful plate appearances than they would without the sign-stealing.

To determine if the Giants' offense benefited from the sign-stealing, I estimated a straightforward regression model of runs scored (RS) by the Giants. A game is the level of analysis and there are 157 observations. Explanatory variables include HOME, a binary variable taking a value of unity for Giants' home games, POST, a binary variable taking a value of one for games played on or after July 20, and an interaction variable HOME x POST. Since signs were stolen only for home games on or after July 20, a positive coefficient on HOME x POST would indicate that the Giants benefited from stealing signs.

A crucial factor influencing the number of runs a team scores in a game is the quality of the opposing team's pitching. Hence, the regression model also includes OPPERA, the earned run average of the opposing team's starting pitcher (OPPERA data was obtained from http://www.baseball-reference.com). The expected sign of the coefficient on OPPERA is positive; the Giants should have scored more runs against pitchers with higher earned run averages. I also tried, both by themselves and in combination with each other and with OPPERA, using the opposing pitcher's ratio of walks and hits to innings pitched and dummy variables for opposing teams as alternative strategies for controlling for the quality of opposing pitching. The results were nearly identical to those reported below and are omitted for brevity.

The estimation results, with Newey-West corrected t-statistics in parentheses, are reported below:
```RS = 3.2-0.09POST + 0.54HOME - 0.61POST x HOME + 0.42OPPERA
(-0.15)    (0.78)     (-0.63)           (1.73)
```

The coefficient on POST x HOME indicates that the Giants scoring actually decreased by more than one-half run per game after they started stealing signs; however, this point estimate is not statistically significant. Hence, the estimation results indicate there is no evidence that the Giants benefited from stealing signs from opposing catchers. Although one cannot rule out the possibility of isolated benefits, e.g., Bobby Thomson's famous "shot heard round the world" to win the pennant, there is no basis for believing the Giants gained a systematic offensive advantage by stealing signs. Indeed, the negative point estimate raises the possibility, during the brief window between catchers giving the signs and pitchers delivering the ball, sign-stealing distracted Giants' hitters and reduced their productivity.

If Giants' batters didn't improve after July 20, what then was responsible for the team's astonishing improvement from a 47-41 record before July 20, to a 51-18 record over the last 2 months of the season? Pitching. Giants' pitchers allowed nearly 1.4 less runs per game (3.3 vs. 4.69) after July 20. Indeed, the coefficients from a linear probability model estimating the response of Giants' wins to the team's RS and runs allowed (RA) [WIN=0.68+0.080RS-0.111RA], indicating a 1.39 per game decrease in runs allowed accounts for three-fourths of the Giants' increase in winning percentage over the last 2 months of the season [-0.111 x -1.39=0.154, which is three-fourths of the 0.205 increase in winning percentage after July 20]. In terms of games won, the Giants expected winning percentage of 0.688 [the pre-July 20 winning percentage of 0.534 plus the 0.154 attributable to improved pitching post-July 20] accounts for 47.5 of the Giants 51 wins over the last 2 months of the season. Although the linear probability model estimate does not account for all Giants' wins post-July 20, it clearly supports the notion that improved pitching was the major change for the Giants in the last 2 months of the season.

Prager's book is a fascinating tale of baseball espionage, and there is no way to disprove his assertion that stolen signs determined the outcome of a pennant race so tight that it was not decided until the last game of the season. However, evidence suggests Giants' pitchers rather than Giants' batters were primarily responsible for the team catching the Dodgers in 1951.

Published online: 27 June 2008

E. Frank Stephenson

E. F. Stephenson ([mail])

Berry College, P.O. Box 5024, Mount Berry, GA 30149, USA

e-mail: efstephenson@berry.edu