NLWC and NLDS Starting Pitcher Predictions: Foul Ball Totals Accurate Predictors of How Long Starters will Last?

With the Wild Card and NLDS series’ completed, I thought I’d take this post to break down the wonder of foul ball predictions I’ve made this postseason thus far, and illustrate some interesting data regarding the first-of-its-kind study into the relevance of foul balls in baseball.

This is the postseason, so some of the commonly accepted “rules” of the game no longer apply. As a result of this fact, I am giving the predictions a modest +/- 1 inning range.

I based my numbers on the “100 pitches rule” used by most skippers in the modern era of baseball. This assumption became an issue when I crunched the numbers and realized that the NL starters weren’t averaging 100 pitches either. Their average in the NLWC and both NLD Series’ topped at 94.63 pitches on average, over three pitches fewer than the American League starters. Unlike, the AL there were no anomalies, although neither Gonzalez nor Wainwright go through the 5th inning.

FoulBallz  1So, how well has assessing how soon a pitcher will be removed from a postseason game based on the number of hitters who generated offense versus were out after their plate appearance with one or more foul balls gone? Given the stricter pitch counts and the +/-1 inning range I predict within, it’s going surprisingly well, maybe even modestly better than the guys on First Pitch and a few other SiriusXM MLB Network shows.



I was much more accurate with the NLWC between the San Francisco Giants and the Pittsburgh Pirates. While Bumgarner seems to have surprised everyone by going 9 innings, my prediction for Volquez was spot on. I had Bumgarner lasting 7 and Volquez leaving in 5. This means they combined for 14 innings; I predicted 12. Again, with the plus/minus my numbers are accurate.

NLDS 1: Giants v. Nationals

In Game 1 of this showdown, Peavy and Strasburg dueled it out for 5.2 and 5.0 innings respectively. Based on my predictions, I had them both most likely gone by the 6. I was off by 1 inning with Strasburg, but only .1 with Peavy. This 1.1 inning difference is nearly within my plus/minus range.

Game 2 between the Giants and Nationals brought us Hudson versus Zimmerman. I said Hudson would go a max of 7 innings; he went 7.1. Zimmerman, I predicted, would go up to 8 innings. He lasted nearly a complete game, 8.2 innings. This left me a full one inning off between the two.

The Game 3 face-off was between Doug Fister and Bumgarner was another successful prediction. I had Fister going 7 innings and Bumgarner lasting 6. Bumgarner ended up staying in for 7 innings.

In Game 4, Gonzalez and Vogelsong met. This was the game that had the most significant deviations in the foul ball predictions. Gonzalez went 4 and I predicted 6. Vogelsong lasted 5.2, 2/3rds over my prediction. With the plus/minus, the predictions are off by 1 inning here.


NLDS 2: Cardinals vs. Dodgers

During the first game of the St. Louis Cardinals and Dodger series, we saw Wainwright go against Kershaw. This was the NL game I did the worst with. I had Wainwright going 6 innings maximum, but he left at 4.1 innings. It ended up that Kershaw wouldn’t go nearly as long as I predicted either. I had him at 7.2. He was out in 6.2. This means that even with the +/-1 adjustment, I was still off .2 innings.

Game 2 is where I would have fared better if I had declared the both pitchers—Lynn and Grienke—would last 7 each. I would have been off by 1 on Lynn’s performance and exact on Grienke’s. However, I didn’t publicize this on time.

With the series tied, the Cardinals and Dodgers entered Game 3 and the match-up was Ryu versus Lackey. Ryu stayed in two innings longer that my prediction, but Lackey lasted exactly as predicted. For these starters, the foul ball data was off by one inning when we take into consideration the plus-minus.

The final game, Game 4, between the Cardinals and Dodgers saw Kershaw and Shelby Miller taking the mound for their respective teams. Kershaw ended up staying in six but the prediction I made was for seven, one inning off and within the parameters of the predictions. Miller went 5.2, a mere 1/3rd of an inning short of my prediction.



As I stated in the American League version, one thing I noticed while running the numbers up to this point is the Wild Card averages for pitches hurled by starters came to 97.75 (99.5 for ALWC and 95.5 for NLWC). That rate dropped in the NLDS series’ too, but not as significantly as in the ALDS’s. The averages in these series’ dropped to 94.63. It’s this increased pitch count—or higher count—that made the foul ball data significantly more accurate for the NL series’ than those in the AL.

As I noted in the American League version of this post, the percentages of foul balls vs. offensive play also illustrated the importance of foul balls in predicting not only the inning a pitcher would be pulled within +/-1 inning, but also allowed me to choose with significant accuracy which hurler would last the longest in the game. Of the nine NL games (1 NLWC and 8 NLDS games) I predicted using foul ball data only, I missed three starters by 2 innings without figuring in the plus-minus. When I adjust for the plus/minus, I was about 3 innings off (AL predictions were 4 innings off). With the National League series’ a few things need to be remembered: There were two more games played on the NL side (thus meaning more chances of being wrong), and the pitch counts were closer to the magic “100 Pitches Rule”. Thus, the improved accuracy of these predictions are much more accurate.

With this type of data, clubs can better assess their needs against starters. Of course, this means the teams need to change their approach to baseball, going with smaller ball. If clubs begin to think in terms of foul balls, they can get to the bullpen sooner. If a team is forced to go to the pen earlier, the arms in the pen will begin to decrease in effectiveness, another benefit of increasing the foul ball rate against a team.