How to Use PITCHf/x in Your Fantasy League by Josh Weinstock February 6, 2012 Be a Scout with PITCHf/x Data We all wish we were scouts — after all, they can, after watching baseball for a few hours, evaluate a player’s ability and future. But scouting is almost shrouded in mystery; with their special reports and terms, it is only something we learn about through online prospect mavens. While this article won’t teach you to be a scout, it will teach you to be a PITCHf/x scout. Anyone can go to FanGraphs and stare at a player’s page, but not everyone can weave through the intricacies of PITCHf/x data. Having the tools to do so will give you an edge among your fantasy competitors. When PITCHf/x is more useful than a player’s stat line PITCHf/x data can be broken down into extreme granularities, so it can help us to pick up on subtle changes. It’s very useful when limited overall data exists; pitchers throw hundreds of pitches in only a few games, so we have much larger samples of PITCHf/x than we do of traditional statistics. Here are some scenarios where it’s useful: • Pitching prospect from the minors is getting his feet wet in the majors: How is his stuff holding up against Major League bats? • Pitcher has added a new pitch in spring training: How effective is it in the regular season? • Pitcher has changed his mechanics: What are the implications for his repertoire? • Pitcher is feeling discomfort in his elbow/shoulder: Can we see this in the data? These are just some of the applications of PITCHf/x data in fantasy baseball. But first you need to know where to find the data! Two great resources are the PITCHf/x tab on the player pages at FanGraphs and Trip Somers’ TexasLeaguers.com. These two websites will tell you about pitch repertoire, movement, velocity, and pitch results. A third tool is Joe Lefkowitz’s PITCHf/x Tool, which will allow you to easily download a player’s PITCHf/x data for an entire season. The last tool to consider is BrooksBaseball.net, which gives you nearly live-updated data for each game. How to evaluate the data The data is recorded with many fields, and you can get a detailed explanation about each field from physics professor Alan Nathan. Here are the most important fields in the data that you will usually be looking at: • start_speed: initial velocity (what you see on radar guns) • pfx_x: horizontal pitch movement in inches (compared to a pitch thrown without spin) • pfx_z: vertical pitch movement in inches (compared to a pitch thrown without spin) • x0: horizontal release point, in feet, measured at 50 feet from the plate. • z0: vertical release point, in feet, measured at 50 feet from the plate. • px: horizontal pitch location in feet • pz: vertical pitch location in feet In terms of pitch results, whiff rate (which you can find at Texas Leaguers or calculated from available data) is what you should usually focus on when evaluating the strength of a pitch. Whiff rate is the percentage of all pitches that result in a swing-and-miss. On a seasonal level, you find a pitcher’s total rate on FanGraphs as Swinging Strike Rate (SwStr%) under the plate discipline tab. The ability to get whiffs is extremely important; whiff rate has a very strong relationship with strikeout rate, which we know is integral to a pitcher’s success. The average whiff rate (whiffs over pitches) for all pitches is about nine percent. For each individual pitch type, here are the average whiff rates: pitch type whiff rate all 0.09 CH (changeup) 0.13 CU (curveball) 0.11 FA (fastball) 0.06 FC (cutter) 0.09 FF (four-seam) 0.06 FS (splitter) 0.15 FT (two-seam) 0.05 KC (knuckle-curve) 0.16 KN (knuckleball) 0.09 SI (sinker) 0.05 SL (slider) 0.14 IN (intentional) PO (pitchout) A note of caution: do not have full confidence in these pitch classifications. This will be elaborated upon later in the article. Focus on this chart when evaluating newly promoted prospects or pitchers that have changed their repertoire during Spring Training. Don’t worry about many common pitch-result metrics like foul rate and in-play rate –- these simply tell us much less than whiff rate does about a pitch’s overall effectiveness. Whiff rate is even more important than pitch type linear weights on both a game-by-game level and a seasonal level. Pitch type linear weights are subject to a great deal variation, largely because they are dependent on batting average on balls in play. And for the same reason we don’t place a great deal of importance on single-season batting average, we should also downplay the importance of single-season pitch type linear weights. Whiff rates, on the other hand, are much more stable. Ground-ball rate per pitch type can also be important for sinker-ballers. We have talked about evaluating pitch results, but what the movement and velocity of pitches? With fastballs, there is a continuum of movement. On one end, we have fastballs that don’t generate whiffs, but result in a lot of ground balls; think Derek Lowe and Trevor Cahill. On the other hand, we have pitches that result in more whiffs but also more fly balls; think Jered Weaver and Ted Lilly. We can tell a lot about what type a fastball will be on this continuum by looking at just two values; pfx_x (horizontal movement) and pfx_z (vertical movement). When looking at right-handers, the vast majority of fastballs will have a negative pfx_x and a positive pfx_z. On average, fastballs have a pfx_x of about -6 and a pfx_z of about 7. You can refer to this graph for a visualization of this continuum: Includes pitches of type FA, SI, FT, and FF. Using this graph, you can tell a lot about what a pitcher will be like with a limited set of data. Say a Minor League pitcher is known for having great downward life. Take a look at his PITCHf/x data and see how he compares to other MLB pitchers, and if we can expect high ground-ball rates in the majors. It’s also important to note that vertical movement (pfx_z) is significantly more important than pfx_x, based on statistical analyses. With breaking balls, pitch movement is less useful than one would expect. While more movement is generally good, the relationship is weak, and there are many breaking balls with limited movement that are very successful. It’s easy to get caught up in which breaking balls have the largest break, but it really doesn’t mean much. Much like with fastballs, there’s an interesting continuum with breaking balls. The hardest breaking balls –- usually in the mid to high-eighties –- generally have a high platoon split. Breaking balls that are around 80 MPH usually have a limited platoon split, and slower breaking balls often have a reverse platoon split. Using this information can help you estimate the platoon split that pitcher will have in a limited sample. Here is some more information about platoon splits and PITCHf/x data . Cautions While PITCHf/x data can give you an edge if you know how to use it, you will be led to false conclusions if you are not aware of the limitations. Pitch-type classifications are done automatically by an advanced classification algorithm. Some of the classifications are actually the same thing. Sinkers (SI) and two-seam fastballs (FT) are classified the same way, and we see the same issue with curveballs (CU) and knuckle-curves (KC). The fastball (FA) designation is also generally similar to a four-seam (FF). The classification system has a lot of trouble distinguishing fastball types and between splitters and changeups. Additionally, while the system is generally very accurate, occasionally PITCHf/x gets out of whack. This can cause inaccurate pitch classifications and data. Even small differences in calibration can cause significant differences in pitch classification; if a camera in one stadium has a pfx_z shift of -2, then a pitcher will be classified as throwing more two-seam fastballs than they actually are throwing. Being aware of classification and calibration error will help you make the most out of PITCHf/x data. An example: Roy Halladay is a good example of where PITCHf/x can help you identify when a pitcher has made a major change that will make him more valuable to your fantasy team. In spring training of 2010, there were stories of him changing the grip on his changeup from a standard grip to a splitter grip. Come 2010, we could easily see that the pitch was much more effective. In 2008-2009 he threw his changeup around five percent of the time with a poor whiff rate of about six percent. After his grip change through 2011, he has thrown his changeup about 13 percent of the time with a fantastic whiff rate of over 20 percent. Since altering his grip he has recorded a career high in strikeout rate, and the change is believable given the congruence of the PITCHf/x numbers and the surrounding details.