Introducing StatCast/xStats Week(s) at RotoGraphs
From now until some point before the season starts, the focus of most RotoGraph articles will be on getting the most useful information out of the StatCast information. It’ll include as many or as few of articles to create a good understanding of the data. With tons of available information much has been written about the subject especially here at RotoGraphs. The key with this series is to cut through all the different data points (e.g. launch angle, spin rate, sprint speed) and how they measured (e.g. max, average, weighted) to find which values to focus on and which ones to ignore.
The point of this article is to provide an initial forum, via the comments, to collect and try to answer the question with any previous research. There is no need to completely recreate everything.
Starting off, here are the very best articles I could find on various StatCast subjects. I’m 100% sure I missed better articles on making StatCast information usable for fantasy owners. Let me know in the comments of any articles I should add (they don’t have to be from a FanGraphs website, it just happens Andrew Perpetua writes here and run the website xStats.org).
Overall hitting article
Three articles on converting StatCast data to fantasy stats.
- Introducing xFantasy: Translating Hitters’ xStats to Fantasy by Ryan Brock
- xFantasy, Part II: Triple Slash Converter by Ryan Brock
- xFantasy, Part III: Can xStats Beat the Projections? by Ryan Brock
Sprint speed
Pitch Spin Rate
As a group, we’ll be head deep in the subject so it’ll a perfect time to ask questions and hopefully, they can get answered. One article, I’m for sure writing is finding which stats on the Baseball Savant dashboard are useful and which ones aren’t. Many more are planned. Please ask away now. Thanks.
Jeff, one of the authors of the fantasy baseball guide,The Process, writes for RotoGraphs, The Hardball Times, Rotowire, Baseball America, and BaseballHQ. He has been nominated for two SABR Analytics Research Award for Contemporary Analysis and won it in 2013 in tandem with Bill Petti. He has won four FSWA Awards including on for his Mining the News series. He's won Tout Wars three times, LABR twice, and got his first NFBC Main Event win in 2021. Follow him on Twitter @jeffwzimmerman.
When have you scheduled Wander Javier Week for? I would like to take that week off of work.
Are you guys at all concerned that relying on xStats might not be all that productive in light of Baseball Prospectus’s analysis suggesting that (at least for pitchers) xStats aren’t actually all that predictive? (See https://www.baseballprospectus.com/news/article/40026/prospectus-feature-siren-song-statcasts-expected-metrics/)
I agree with BP’s assessment. On the hard hit numbers I couldn’t find anything. The launch angle (GB%) is useful.
Throughout 2018 I’d analyze xStats data to identify outstanding performances or changes in performance and make waiver wire pickups based on xStats. Retrospectively, I agree with sentiment above, they are more descriptive not predictive. Matt Chapman is a great example, when he was on early in the year and then again in the 2nd half, his xStats were great. But then he had some spells that were less than stellar. From a predictive use, my hunch is they can provide ceiling and floor levels of a player, but not necessarily forecasting value
xStats are not meant to be predictive, they are meant to be descriptive.
Of Hard%, EV (avg), or EV (max), which ones are the most predictive of future hitting output (is wOBA the proxy)? Or is it some frankenstein such as EV of non-grounders etc.?
and is EV variance (distribution of EV’s) predictive? Is it better to have a tight EV distribution of 97-99mph or something like avg of 96 but sd of 5 mph?
There is no “tight” with so many foul balls. While I don’t agree with the parameters of the barrel metric, I believe it’s more important to focus on % of good contact but will dive into this topic
Perpetua has done a ton on this and brings it up in his first article above (not the Hard%). He takes into account more factors than you listed.
How soon (eg plate appearances) can a prospect’s 20-80 game/raw power be validated once he’s in MLB with StatCast EV?
can some sort of log5 comparison be done where the hitter’s EV is compared with the pitcher’s historical EV? (Similar to how in NFL strength-of-schedule is used to better understand team’s talent level). e.g. hitting 101 mph off a pitcher who normally suppresses contact vs htting 101 mph off of a pitcher who routinely serves up meatballs could/should intuitively mean different things.
I see this is asked and answered in the Baseball Prospectus article of adjusted EV
High spin rates generally make for higher swstrk% (because batter’s expect the four seamer to drop more than it does…in broad terms). It stands to reason that as pitchers focus on increasing their spin rate (or teams focus on using pitchers with high spin rates) that the batter’s “expectation” of flight path would change (as they are exposed to more and more ‘high spin rate’ pitches). Therefore, would we expect that low spin rate pitchers would then induce more whiffs? ie is it cyclical? In other words, is a pitcher’s higher whiffs on higher spin rate because it is anomalous to have high spin rates? is ‘z score’ of spin rate the more important statistic?
Thanks for the shoutout on those ’16 xFantasy pieces I wrote. I’ve been meaning to revisit them, though honestly I’ve also been waiting for the dust to clear on the xStats data. Some initial followup I did on the ’16 numbers showed they weren’t terribly useful for predicting ’17, but now that we have another year under our belts, it will be worth another look.
A shooting-from-the-hip hypothesis: xBABIP and xHR would seem to be the two biggest factors that could improve hitter/pitcher projections over what Steamer (for example) puts out. Though there are known issues for very fast/slow players that consistently over/under-perform xBABIP, so some effort would have to be made to remove that source of error.