Just Another Reason To Be Skeptical of Spring Stats

When FanGraphs was in Arizona earlier this spring, our merry band of nerds made our way to a Jarrod Parker vs Chris Sale afternoon tilt. The result on that 16th of March — an 11-5 win for the White Sox and bad performances from the two starting pitchers — was mostly unimportant to everyone in involved. But a few innings sitting behind the plate did provide some insights, including some reasons why those results were unimportant.

In some ways, the results were important. If you look in the box score, you’ll see that Hiroyuki Nakajima had an error. That means something, but only when combined with the observation that he was tentative in the field, and had his hand in two other plays that could have been scored errors as well. He dropped a ball at second that could have started a double play. He collaborated with an two other A’s to drop a short fly ball. He hesitated on throws. Combined with the thought that Patrick Newman had before he came over from Japan — that his defense probably didn’t make him an everyday MLB shortstop — there might be legit reasons to wonder about the A’s biggest free agent of the offseason.

The box score also says that Chris Sale walked two batters that day, same as Jarrod Parker, who added a wild pitch. And if you look at their stats last year, it was Sale that had the better control. But if you watched those two pitchers that sunny Saturday, you’d have a different impression of the two. Sale couldn’t find the plate in the second inning, with two four-pitch walks (Michael Choice and Daric Barton). But his poor control also helped contribute to the hits he gave up — seven in five-plus. He wasn’t getting around on the slider, and he was having some release point issues. And watching those elbows and knees and all the recoil in his delivery made his control problems seem even worse.

Quality of opposition is often cited as a reason to ignore spring stats, and there are even statistics that try to account for the uneven nature of a spring game. But if Sale was having trouble against an Oakland squad that was batting Chris Young third and Josh Donaldson fourth, wouldn’t that mean something worse than if it was a murderer’s row he was up against?

We know that the starters leave after three innings early in the spring, so spring stats are skewed because it. Jarrod Parker‘s afternoon gave us a different reason to be skeptical of spring stats.

Early in the outing, Parker’s curveball — his third or fourth best pitch depending on who you ask — was crisp. It had 12-6 break and some sizzle, and batters were missing the pitch. By the third inning, that changed. He kept throwing the pitch, but the sizzle was gone, and it flattened out. And yet he kept throwing it to batter after batter. Conor Gillaspie doubled off a curveball. Jordan Danks doubled off a curveball. Hector Gimenez tripled off a curveball. Most of these were curveballs inside, yanked down the line. And Parker kept throwing his curves up there.

In fact, in the the three innings I watched, I only saw two changeups. That would be rare for Parker, who averaged one for every five pitches thrown in 2012. Adam Dunn walked up to the plate, and with two strikes, Parker finally threw the first change I saw. Dunn barely held up and was spared a strikeout. Parker threw him another changeup, and this time Dunn couldn’t stop himself.

And here’s a strange conundrum — lesser players hit Parker around, but he wasn’t showing them his best. He was ‘working on something’ in the parlance of the season. There’s no specific mention of him working on his curveball in the game report, but he does mention that he was focusing on how he felt “the day after” and that he was hoping to work on his two-seamer in the next start out. And when that next start was a gem, Parker said that it was a night game, closer to the season, and that he “wanted to treat it a little more like the regular season.”

There are things to be learned in your average spring game. Hiroyuki Nakajima has some work to do on defense. Chris Sale may have slightly worse control than his walk rate last year suggests. Jarrod Parker is working on his curveball. But these things don’t show up in the box scores exactly like we think they might. Good thing we still watch the games.




Print This Post



Graphs: Baseball, Roto, Beer, brats (OK, no graphs for that...yet), repeat. Follow him on Twitter @enosarris.

21 Responses to “Just Another Reason To Be Skeptical of Spring Stats”

You can follow any responses to this entry through the RSS 2.0 feed.
  1. Victorious says:

    Great read Eno. Really reinforces the point that ST stats really doesn’t matter, and that you shouldn’t take the box score at face value.

    But really, Puig is the second coming. I kid! Maybe….

    Vote -1 Vote +1

  2. ML says:

    I dont agree with the blanker “they dont matter” statement; contextually there are certain things I believe that can be informative – I like to see how often guys are striking out, if expected speedsters are swiping bags, and to some degree homeruns. Would never base a draft decision solely on the data, but for example, I’m encouraged to see jennings has swiped 6 bags in the spring. doesnt change my projection or anything but gives me some degree of comfort

    Vote -1 Vote +1

    • Pirates Hurdles says:

      If it doesn’t change your projection then it does not matter. Statistically, we could care less if your are “comforted”.

      Most of the cases pointed too by the believers smack of confirmation bias.

      Vote -1 Vote +1

    • Baltar says:

      HR’s are probably the least significant of all spring stats.

      Vote -1 Vote +1

  3. One Mans Opinion says:

    “Chris Sale may have slightly worse control than his walk rate last year suggests” because of what you saw in 5 innings of a Spring Training game? Eh..

    Vote -1 Vote +1

    • Eno Sarris says:

      Well, I cheated and looked at his 2011 and 2010 walk rates too. I’m well aware of the power of larger sample sizes, but I do think people can see things in games that are real.

      Vote -1 Vote +1

  4. Jaker says:

    “No data are meaningless to the statistician.”

    Can we please end this pointless debate? No data are meaningless.

    Even if ST stats lack inferential value, which hasn’t been sufficiently explored, there there are still qualitative takeaways and contextual factors to consider.

    ST stats are a quantitative reflection of how a player is performing. If a pitcher has 30 BB and 0 K in 30 IP do you really need to have watched every pitch to understand that the pitcher is struggling with command? Context then becomes important. Is he changing his mechanics? Trying a new pitch? Recovering from injury? Or is this a continuation of an issue from the year prior? These additional pieces of information help put the stats into context. We may not always not know the context but we often have a good idea.

    ST stats are information. The more information you have, the better. To simply espouse a blanket “ST stats are meaningless” statement is ignorant and frankly annoying. If you don’t find value in ST statistics then ignore them.

    Vote -1 Vote +1

    • Jaker says:

      I should say that this wasn’t directed at Eno! Just ranting. By all means be skeptical just not ignorant.

      Vote -1 Vote +1

    • Eno Sarris says:

      While you’re having fun with your straw man I’d point to the moments of nuance in my piece. I was providing context and nowhere did I say that spring stats were meaningless. I even cited some from the box score.

      I found the curveball thing the most interesting. That would be hard to quantify statistically and it’s noise in whatever you want to take away from spring numbers.

      Vote -1 Vote +1

      • Jaker says:

        You probably responded before my qualifying reply. My comment wasn’t at all in response to your piece or directed at you. It was some annoyance carried over from another forum.

        But I think you reinforced my point that context is important. The box score won’t tell you that Parker was working on his curveball but if you watched the game or follow the team closely you might have that additional information. Context gives ST stats more meaning at the individual level, especially at the individual game level, but if you start to look at the data as a whole it can be misleading and far too noisy to derive any meaning.

        Vote -1 Vote +1

        • Pirates Hurdles says:

          “Context gives ST stats more meaning at the individual level, especially at the individual game level”

          Prove it. We are all well aware of player X making adjustment Y leading to better ST number Z then when it counts nothing holds up. The opposite also often happens where player X is working on some thing Y and the results are poor in ST and it predicts nothing useful come April.

          Vote -1 Vote +1

        • Jaker says:

          I think you’re missing the point of what I’m saying.

          At an aggregate level, there’s too much noise in ST stats to take any meaning from them.

          At the individual level, given context, we can better understand their ST stats. So as Eno points out, if a pitcher like Parker is only throwing curve balls despite getting lit up, we understand not to read too much into that start. If a pitcher however struggled with command for the last 3 months of last season and walks almost a batter per inning in every one of his ST starts (e.g., Ricky Romero) it tells us that command might be a real issue.

          That’s all I’m saying.

          And unfortunately that’s not something you can quantitatively prove. It’s proven qualitatively all the time by bubble players who have poor springs getting relegated to the minors for example.

          Vote -1 Vote +1

    • Pirates Hurdles says:

      This is off course nonsense, since Jaker points out all the caveat variables that come along with these statistics. Show me any evidence of anything predictive on a large scale and then you can rant. Heck, show me something predictive on a small scale that holds up. Something like, the top 30 ST pitchers with highest walk rate and look at what happens.

      Vote -1 Vote +1

    • Baltar says:

      Your example uses the only 2 spring stats, pitchers K’s and BB’s, that may have some predictive value. More information is not always better if it is bad information.

      Vote -1 Vote +1

  5. MLB Rainmaker says:

    Its funny, this argument won’t even matter by the first week of April! The fantasy baseball season is 5+ months long!!!!

    Vote -1 Vote +1

  6. Spit Ball says:

    Has anyone looked at The Baltimore Orioles Spring Stats. They have five outfielders Lew Ford, Steve Pearce, Chris Dickerson, Jason Pridie and Trayvon Robinson who have 41 to 44 plate appearances. All of them are non roster invitees and absolutely raking. All of them are hitting 300 or better and have an OPS of 889 to 1340. Meanwhile Adam Jones, Nick Markakis, Nate McLouth and Xavier Avery are really struggling. Spring stats absolutely don’t matter, never have.

    Vote -1 Vote +1

  7. maqman says:

    The “all spring training stats are meaningless” references are ignorant. They ignore that they matter to the players trying to make a club, to FOs and fans and are to some degree informative as to expectations. They are not solid predictors of seasonal performances, which most people would agree with but to insist they are meaningless is simply hyperbole.

    Vote -1 Vote +1

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

Current day month ye@r *