Tag Archives: Sample size

Forget Pitching, Hitting, and Fielding! The New York Mets’ Most Glaring Area Of Weakness May Be Statistics

It’s difficult to believe that Major League Baseball’s New York Mets won 11 of their first 12 games this season. Earlier today, the Chicago Cubs completed a four game sweep of the club, continuing a stretch in which the Mets have lost 29 of 45 games.

During this woeful period, fans have witnessed displays of poor pitching, hitting, and fielding skills. And to make matters worse, earlier this week, they witnessed a managerial display of poor statistical skills.

At a critical moment in a game against the Milwaukee Brewers, Mets Manager Mickey Callaway removed an effective pitcher and replaced him with an ineffective one. The change enabled the Brewers to score four runs and convert a two run New York lead into a two run Milwaukee surplus.

So why did Callaway bring in Jerry Blevins to replace Robert Gsellman? Given that Blevins has struggled all season, while Gsellman has delivered periods of clutch pitching? Callaway explained:

The seven times [the Brewers batter] faced Gsellman he got three hits. He’s never gotten a hit (0-for-2 with a walk) off Blevins. The overall numbers suggest Blevins has a much greater chance to get the hitter out and you have to go with those. It is part of managing the game today.

At first glance, it does seem reasonable to bring in a pitcher who has experienced success against a batter. But “0 for 2 with a walk” means that the pitcher had only faced the batter three times in his entire career!

That isn’t even close to a statistically meaningful number of past attempts. Callaway himself acknowledged the “small sample size” that he relied upon to make his decision.

In all fairness, there is no single numerical minimum of observations that must be considered when making a statistically valid decision. The minimum number varies by one’s willingness, in any particular situation, to tolerate risk and uncertainty.

And yet no mathematician would agree that a counter-intuitive baseball decision could be made on the basis of three prior outcomes. Thus, the Mets can now add Statistical Analysis to their List of Necessary Improvements.