Foul Ball Predictions: NLCS and ALCS Prediction Results

The foul ball predictions for the Conference Series’ have been way off, even with the +/- 1 inning range I mentioned in an earlier posts. The issue with the Conference series predictions has come as a result of lower pitch counts, counts lower than the Wild Card and Division Series’ combined. Does that mean the ability to predict when a starter will be pulled from the game based on fail ball data is flawed? No. It simply means that the post season is, as we all know, a different beast than the regular season. As you will see, then the starter was left in to throw in the vicinity of 100 pitches, the predictions are remarkably accurate.

As has been the case, all stats are from the 2013 season for either a night game or a day game, depending on when it is being played, and at-bats that included 1 or more foul balls. The following data comes from our foul ball database that draws from Retrosheet.org.

As a reminder, all percentages represent a combination of batters who got on base themselves (1B, 2B, 3B, HR, Hit-by-pitch, BB), helped force an error, a wild pitch, a steal or otherwise generated offense.

 

ALCS: Kansas City Royals v Orioles

Within the plus/minus of +/-1 inning, the foul balls data from the 2013 season weren’t too far off for the ALCS. The main issue with the predictions was that the Kansas City Royals and Baltimore Orioles insisted on taking out their pitchers early. Very early. The starters in the American League Conference showdown averaged a meager 86 pitches per start. This became an issue with predicting more accurately. On average, the pitchers would have all been in for one more inning during the regular season.

That said, the foul ball data from 2013 for each starter translated “okay” to these four games. Game 1 between Shields and Tillman was off by 1.2 innings (neither starter got to the 6th inning). For Game 2 starters Ventura and Norris the numbers—again because of the low pitch counts—were off by an innings given the +/-1. Games 3 and 4 both fell within the accepted plus-minus range created for this research.

Whereas the numbers were much better in the Wild Card matches and the Division Series’ for these four games, then numbers proved off by nearly three innings.

 

NLCS: San Francisco Giants v. Cardinals

Through the five games of the NLCS, the numbers suffered a similar blow to accuracy as they did in the ALCS. Low starting pitcher pitch counts doomed the prognostications for the NLCS too. While five of the ten starting pitchers reached or came close to the century mark with their count, the overall average for the NLCS starters through the series was lower than the American League with an average of 87.2 pitches per start. Only Madison Bumgarner in the first game reached the 100 mark, reaching 112 tosses.

As a result of the lower counts, the predicted innings ended up being off nearly six innings over four games. That’s even when I figure in the plus-minus. Thus, I was at about a 50% prediction success rate for the National League Conference Series.

 

Observations

 

As it was in the ALDS report and the NLDS story, had starters been left in for even one more inning—as they would have been in the regular season—then the predictions would have been significantly more accurate. The foul ball data from the 2013 regular season is fairly accurate as a predictor of how quickly a starter will leave the game. The calculations are based on the number of at-bats that have a foul ball versus the number of at-bats without a foul ball. More specific data will be coming during the off season, but as a preliminary examination of foul ball rates, the starters came out due to pitch counts, driven up by the number of foul balls hit against them. Such a discovery can revolutionize the way managers stack their line-ups.

Many of the pitchers last season allowed 30% or more of the batters, regardless of which side of the plate the batter stood on, to generate an offensive play. That’s the equivalent of a .300 batting average, an average that is roughly .050 above the batting average of all of MLB in 2013.

The goal of baseball is to generate offense. It appears, from this preliminary research, that the best way to ensure offensive is created is to foul off as much as possible.

During the 2014-2015 off season, I’ll be bringing you the first of its kind detailed look at foul balls from 1988 to 2014. The various posts will include the most accurate foul ball counts ever done and continue to forward more theories on how we can learn about baseball from balls hit in to foul territory and out-of-play.

 

On to the World Series!

Come join the growing fans on Twitter too!