Projecting the Twins Statcast-style

[Patrick Wozniak]

Current projection systems such as ZiPS, Steamer, ATC, and the Bat use the results from a player’s previous seasons in an attempt to project their future results. That’s great, but wouldn’t it be cool if there was a system that incorporated Baseball Savant’s Statcast data?Jeff Zimmerman of FanGraphs has attempted to do just that. Instead of using past performance, Zimmerman used Statcast metrics such as Max Exit Velocity, Barrel %, Launch Angle, and Sprint Speed to create projections for the 2020 season. He also used O-and Z-swing and contact rates to account for plate discipline. Next came the normal weighting/age-related regression type stuff and then the conversion of the Statcast results into more standard baseball stats, and finally, out popped the projections. Or something like that.

 

The full results can be viewed in this spreadsheet, but a cursory glance will show the usual suspects at the top (Mike Trout, Yordan Alvarez, and Mookie Betts) and bottom (Dee Gordan, Billy Hamilton, and Ronald Torreyes) when ranked by OPS.

The model does not account for shifts (which are obviously a large detriment for heavy pull hitters), league differences, or ball parks so it’s basically projecting how hitters would perform in a vacuum. It also doesn’t attempt to account for playing time, so every hitter gets 600 plate appearances. The obvious advantage is that it’s easy to compare players with all things being equal, the disadvantage being that in reality things aren’t equal. So it goes.

 

With those caveats in mind, let’s check out how our Twins stack up. I’ve included each player’s MLB ranking (based on OPS) and in the last row the ZiPS projections, so we have a more traditional projection system to contrast with the Statcast one. Please take a moment to appreciate the aesthetically pleasing colored rows which represent the pinnacle of my newly-found technological knowhow.

 

Download attachment: Statcast Projections Chart pic.png

 

My takeaways:

• That truly is a beautiful table.
• I was surprised that it was ZiPS, rather than the Statcast-based projections, that favored the trio of Nelson Cruz, Miguel Sano, and Josh Donaldson as they were three of the leagues best at “barreling up.” Maybe they were hurt a bit by whiffs in regard to plate discipline?
• The Statcast method is much more favorable to LaMonte Wade Jr. than ZiPS, which I also wouldn’t really have expected. However, his Statcast numbers are better than I imagined (mandatory small sample size warning), leading to a .373 expected wOBA, and Wade does have a great eye.
• Luis Arraez’s projected OPS is exactly .773 in both systems (though he gets there with a bit more power in the Statcast projections vs. a higher batting average in ZiPS). Combined with his uninspiring defense, a significant dip in his offensive production would put a damper on Arraez’s overall value (I’m not saying this will happen, but it could).
• The projection systems disagree on Eddie Rosario as much as Twins fans do (Statcast Eddie might be out of a job).
• The stolen base projections certainly seem overly optimistic for a team that runs as infrequently as Minnesota (Buxton’s total of 28 would match the entire 2019 club, which includes 2019 Buxton). But the numbers are interesting because they are intended to show how often a runner will take off based on their sprint speed. Although Buxton’s the only burner, could someone like Polanco, Kepler, or Wade Jr. (if he sees enough playing time) take a step forward?

How useful is this new projection model? Time will tell, and revisions will probably be warranted, but it’s certainly unique and entertaining (and potentially beneficial for Fantasy purposes). And who can argue with having a little extra nerdy-baseball fun in our lives right now?

 

MORE FROM TWINS DAILY

— Latest Twins coverage from our writers

— Recent Twins discussion in our forums

— Follow Twins Daily via Twitter, Facebook or email

 

Click here to view the article

[DocBauer]

My biggest takeaway is 6 guys with an OPS of .800+. NICE!

[Patrick Wozniak]

 

My biggest takeaway is 6 guys with an OPS of .800+. NICE!

But ZiPS gives us 7 so it must be more accurate.

[adorduan]

I'll take the under on Sano stealing 6 bases.........

[Patrick Wozniak]

 

I'll take the under on Sano stealing 6 bases.........

Yeah, I don't think the Statcast stolen base projections are very useful for prediction, but more as a bar for what a player of similar speed would typically steal. The more traditional projections that consider past performance would better predict how well a player will perform in the "art" of base stealing by virtue of being based on actual results which are influenced by reading the pitcher, good jumps, team's willingness to run, etc.