BP Comment Quick Links


August 28, 2009 Checking the NumbersWhiffing the Pitcher, Part One
Cliff Lee has performed at an incredibly high level since joining the Phillies rotation last month, compiling a minuscule 0.68 ERA in 40 frames, effectively making Phans phorget that they were ever engaged in the Roy Halladay sweepstakes. Lee's junior circuit strikeout rates were generally of the good but not great variety, somewhere in the range of 6.1 and 6.8 per nine innings, yet his 39 whiffs while wearing red pinstripes has resulted in a fantastic rate. Switching to the National League tends to improve a pitcher's performance for a few reasons, but the most obvious of them involves the chance to face pitchers (Ed. Note: Phlailing?) as opposed to additional hitters with With that in mind, is it possible we inflate or deflate the worth of a pitcher based on his strikeout rate if said pitcher drastically delivers Ks above or below the average rate of punching out pitchers? Or for that matter if pitcher whiffs comprise a very high percentage of his overall strikeouts? A slight investigative issue occurred right from the getgo, as I don't necessarily have a database table column devoted to classifying whether or not the batter in a plate appearance was the opposing pitcher. To circumvent this problem as well as the misclassification of, say, Aaron Miles or Mark Grace as pitchers, anyone with five or more innings pitched in a season was pooled together and all of their plate appearances from 19792008 were evaluated. Throughout that span, pitchers fanned their hurling colleagues in 31.4 percent of their total trips to the dish; sac bunts comprised much of the remaining percentage points. When the minimum number of Pitcher PAsthe number of plate appearances for pitchers against pitchers hittingis adjusted to 20, the 31.4 percent mark remains unchanged. In determining the percentage of overall strikeouts that are pitcher whiffs, it may seem wise to restrict the span to NL pitchers only, but by instituting the minimum of 20 Pitcher PAs we can avoid issues wherein AL pitchers skew the averages by barely facing pitchers from the opposite league. Pitchers from the American League are unlikely to meet the "20 Pitcher PAs" criteria, meaning that they would be excluded and therefore unavailable to distort the data; if they did however meet the criteria, then their results should qualify for inclusion. This really just pertains to the percentage of pitcher whiffs out of overall strikeouts, since clearly the bulk of Ks for AL pitchers will come in their own league. Over the last 30 years, pitchers have struck out opposing pitchers almost onethird of the time, and the resulting punchouts have accounted for 14.3 percent of their raw strikeout total. As with any metric, players can end up well above or below the average, so here are the top and bottom five seasons in terms of the percentage of pitcher plate appearances resulting in a whiffPitcher Whiff Percentage, or PW%and the percentage of pitcher strikeouts as they relate to overall strikeoutsPitcher WhiffOverall %, or PWO%. (For the record, I'm more than willing to alter these names/acronyms with suggestions. Year High Pitcher P/PA P/K PW% Year Low Pitcher P/PA P/K PW% 1995 Sid Fernandez 20 16 80.0 1994 Andy Hawkins 22 1 4.6 1996 Sid Fernandez 21 15 71.4 2008 Jared Fernandez 20 1 5.0 2002 Mark Prior 35 23 65.7 1985 Greg Minton 20 1 5.0 1985 Carlos Diaz 20 13 65.0 2002 Greg Reynolds 20 1 5.0 1983 Cecilio Guante 22 14 63.6 1982 Kevin Ritz 30 2 6.7 PW%: Percentage of pitcher strikeouts out of pitcher plate appearances Year High Pitcher P/K K PWO% Year Low Pitcher P/K K PWO% 1979 Adrian Devine 9 22 40.9 1999 Kenny Rogers 3 126 2.4 1984 Jeff Lahti 18 45 40.0 1990 Mike Hartley 2 76 2.6 1984 Joe Sambito 10 26 38.5 1985 Greg Minton 1 37 2.7 1988 Danny Cox 18 47 38.3 2005 Chan Ho Park 3 113 2.7 1985 Frank Pastore 11 29 37.9 2002 Jared Fernandez 1 36 2.8 PWO%: Percentage of pitcher strikeouts out of overall strikeouts One thing has become abundantly clear: Sid Fernandez really knew how to whiff the opposing pitcher. In fact, 10 of his seasons met the aforementioned criteria198490, '9293, and '96and all but 1984, a year in which he struck out 31 percent of the opposing pitchers, ended up well above the 31.4 percent average. His data alone invites the question of whether or not either of these statistics are stable. It's fun just to ask, but can striking out the pitcher be considered a repeatable skill? And, building off of that, since strikeout rates themselves are generally very stable, if pitcher whiffs are also stable, then the percentage of overall strikeouts comprised of opposing pitcher strikeouts would have to be stable as well, right? Before moving forward, the reason to use both of these numbers deals with sampling. For instance, Aaron Harang fanned 218 hitters in 2007, with a 41.7 PW% as he struck pitchers out in 25 of the 60 plate appearances. Those 25 strikeouts, however, amounted to a very small portion of his total strikeouts, so while he ended up above average in terms of whiffing the pitcher, his overall strikeout total still consisted of approximately 90 percent "regular" hitters, and he fell below average in the PWO% department. Despite "below average" being the more accurate term, Harang's numbers portend sustainable success in that his strikeouts really were not padded much in spite of his aboveaverage rate of fanning the opposing pitcher. Inversely, those with above average PW% marks and PWO% results in excess of 14.3 percent should draw red flags, as their strikeout rates may be in line for some regression. To answer the first question revolving around stability, I will once again call upon the AR(1) Intra Class Correlation, a statistical test becoming so popular in this space that it is very close to achieving sitcom freezeframe applause status when mentioned. This particular ICC works similarly to standard, yeartoyear correlations, instead applying its methodology over a larger number of years. A pitcher need not post identical PW% marks in successive seasons to score well and be considered stable either as this test looks for data points hovering close to the mean as well as following a trend; jumping from 18 to 20 to 24 percent would be considered stable, while 18 to 20 to four percent would not be, but the latter would still earn credit for being stable in two of the three years. Running the numbers from 19792008, PW% produces an ICC of 0.32, close to moderate stability in a baseball dataset. The PWO% produces a higher 0.41 ICC, suggesting that pitchers tend to be more stable in terms of the percentage of pitcher whiffs out of overall strikeouts. The correlations may be of moderate strength, but are nowhere near the threshold at which it may make sense to avoid further research revolving around strikeout rates in subsequent seasons regardless and regarding pitcher whiffs. To start, I made two separate tables of data from 19792008, one featuring pitchers who switched leagues from one year to the next, and another with those remaining in the same league. From there, the weighted averages of everyone with 30 or more frames logged were calculated dealing specifically with the deltas in K/9 and K/PA. The table below shows the average changes, as well as the deltas in the control group designed to serve as the comparative basis to ensure that the leagueswitching deltas are actually out of the norm:
League K/9 Delta K/PA Delta
ALAL 0.056 0.002
NLNL 0.057 0.002
ALNL 0.250 0.009
NLAL 0.310 0.003
As these results indicate, we can expect pitchers, perhaps through natural aging curves, to decline ever so slightly from one year to the next, even if they stay in the same league. When they switch leagues, however, the data above confirms expectations, in that senior circuit pitchers on average experience a K/9 drop of 31 points, while AL pitchers are able to improve by a quarter of a whiff per nine innings in the senior circuit. Factoring in the control group's deltas, AL pitchers moving to the NL actually experience a 31point swing in the strikeout rate. The next step involves incorporating the pitcher whiff data, shifting the leagueswitch deltas into the control group area. This way, we can effectively test how pitchers with above or belowaverage pitcher whiff data fared during league and nonleague changes from one year to the next, relative to what the standard change in results looks like. The standard changes in the control group are going to look slightly different here given that the syncing up with pitcher whiff results facilitates the need for 20 pitcher PAs or more, a stipulation that was not present when calculating the weighted average deltas above; thus why the table two paragraphs up indicates a dropoff of 31 points for a switch from the NL to AL yet the number below is approximately 43 points, which is interesting in and of itself given that requiring the sampled pitchers to face a higher amount of pitchers hitting causes a greater decline. Here are some of the results for league change deltas in K/9 relative to PW%: League K/9 Delta PW% NLAL 0.72 > 0.31 NLAL 0.72 > 0.40 NLAL 0.62 > 0.50 Control 0.43 All And here are the results for those who remained in the senior circuit from one year to the next: League K/9 Delta PW% NLNL 0.19 > 0.31 NLNL 0.32 > 0.40 NLNL 0.59 > 0.50 Control 0.05 All The data above essentially feeds our intuition, which is great considering how frequently intuition can lead us astray. Regardless of whether an National League pitcher moves to the AL or stays in his usual surroundings, we can expect an automatic decline in strikeout rate in one form or another, because of aging, the improvement of talent in the same league, or the switching of leagues. The pitcher's whiff parameters exacerbate the decline in both situations, as NLonly pitchers can be expected to experience a significant decline the year after posting aboveaverage opposing pitcher whiff marks, even in spite of the moderately stable correlations. When switching to the American League, the already vast decline in K/9 grows, in theory making an AL team's acquisition of a NL pitcher who padded his strikeout totals and rates in this fashion a very poor idea indeed. The inverse is then true as well, making pitchers like Cliff Lee, with his average AL strikeout rate, all the more attractive as he can be expected to add at least another quarter of a whiff per nine with the switch. Evaluating the strikeout rate of pitchers based on their propensity to strike out their counterparts may seem basic, but it has the potential to help us better understand why certain pitchers experience declines in this area as opposed to following what appears to be an obvious trend. With that idea in hand, it should be treated similarly to walks and unintentional walks; while it makes more sense to remove intentional walks in evaluations given that the decision rests squarely on the shoulders of a manager and is not generally indicative of pitcher performance, looking at a pitcher's "true" strikeouts, or strikeouts of actual bigleague hitters, may be a more useful and predictive tool. (Sure, Carlos Zambrano, Mike Hampton, and Micah Owings may take issue with such an assertion, but pitchers who can hit are in the very small minority.) Next week we will delve into a few different facets of this line of study, involving how luck and peripherals can go hand in hand herepitchers with very low BABIPs are going to face pitchers with runners on base less often, affording fewer opportunities for sac bunts, and therefore increasing their chance of a adding a relatively easy whiff to their ledger. In other words, we'll be exploring how luck follows the peripherals, and vice versa. Additionally, we'll look into whether or not K/9 in Year One, with a hint of aging, is a better predictor of K/9 in Year Two than if the adjustment of pitcher whiffs to the league average works better. If the latter rings true, this would become another component to incorporate into projection engines. On top of that, we'll take a look at leaders in nonpitcher whiffs as this is much more likely to be an indicator of the hurlers with true strikeout skills. For now, however, when researching pitchers for fantasy teams or dollar valuations, be wary of the PW% and PWO%. Stay away from those who predominantly compose their strikeouts rates and totals of pitcher whiffs, as the averages portend a vast decline in the following season.
Eric Seidman is an author of Baseball Prospectus. 5 comments have been left for this article.

I actually like PW% and PWO%. They lend themselves rather naturally to the term "Pwned", which is an admirable and apt bit of cross pollination geekery.