How the League Adjusts to Hitters Over Time

Mets first baseman Ike Davis has seen the number of fastballs thrown to him drop significantly since his rookie season in 2010. In that year, 57% of the pitches thrown to Davis were some type of fastball. So far in 2012? Only 51%. There have been only 30 seasons between 2007 and 2011 where a hitter with more than 100 plate appearances saw a lower percentage of fastballs in a season than Ike this year — and only five where a player accumulated more than 500 plate appearances.

Clearly pitchers are adjusting to Davis, altering their approach based upon Davis’ perceived offensive strengths and weaknesses. This got me thinking about the extent to which major league pitchers adjust to hitters from year to year. Was this change significant, or more common based on the normal adjustments hitters can expect to see from year to year.

As a first cut, I decided to look at changes in the pitch types that batters faced in consecutive years. Throwing hitters a different mix of pitches (i.e. fastballs, curveballs, sliders, etc.) is just one way the league can adjust. Pitchers can alter location, sequence and speed. However, the data was more readily available for pitch types, so the choice was made to focus there first. I decided to use the pitch-type distributions that are based off of PITCHf/x data. This allowed me to collect data on hitters with 100 at least plate appearances each year from 2007 through 2011. For reference, here are pitches as classified by PITCHf/x:

Abbreviation Description
FA Four-seam Fastball
FT Two-seam Fastball
FC Cutter
FS/SI/SF Sinker, Split-fingered Fastball
SL Slider
CH Changeup
CU Curveball
KC Knuckle-curve

Now, there are obviously some coding issues that come into play. PITCHf/x and the classification algorithms that it uses aren’t perfect and have changed a bit over the years. For example, it isn’t uncommon for sliders and cutters to be mistaken for each other. This is a problem, but not a fatal one as long as we acknowledge it up front. To get a first look I decided to group all fastballs together as this might alleviate some of the potential coding issues within the fastball category. The table below shows the results for all hitters with at least year-two data and for those with year-two and year-three data:

Sample N Fastball% YR1 to YR2 Fastball% YR1 to YR3
All hitters with only two years of data 142 .620** NA
All hitters with at least three years of data 367 .586** .519**
**Sig at the .01 level

For hitters with only two years worth of data, we see a highly significant correlation between the percent of fastballs seen (.620). The correlation drops a bit, to .586, for hitters with three years of data. Additionally — for hitters with three years of data — we see that the relationship decreases further once we get to that third year. This likely reflects the fact that hitters who only managed to accumulate at least 100 plate appearances in two consecutive years were lesser hitters and, therefore, required less adjustment from the league. These hitters eventually drop out of the sample and we’re left with hitters who perform well enough to require additional approach changes from opposing pitchers. The sample above includes all hitters, but is there a difference between established hitters who have major league track records and rookies who haven’t accumulated a significant number of plate appearances? My initial hypothesis was, yes, we should see a greater adjustment by the league in terms of rookies versus players who have already established their habits and tendencies. To test this I calculated separate correlations for hitters whose first season in the sample was their first with at least 100 plate appearances. Here are the results for hitters with two years of data, compared to those with three years:

Sample N Fastball% YR1 to YR2 Fastball% YR1 to YR3
Non-rookies with only two years of data 88 .569** NA
Rookies with at least two years of data 54 .710** NA
Non-rookies with at least three years of data 291 .575** .527**
Rookies with at least three years of data 76 .633** .508**
**Sig at the .01 level

Contrary to what I expected, the league feeds rookies a similar percentage of fastballs between year one and year two than for non-rookies, regardless of which sample we look at. But the adjustment from the first year to the third year is much larger for those who were rookie hitters. My guess here is that pitchers need more than the first year’s worth of plate appearances to update their approaches. The average number of plate appearances in players’ rookie years was 359, compared to 418 in the second season and 450 in the third. For non-rookies, the average plate appearances were 404, 405 and 438. And if we assume that year one in the data set for non-rookies is at least their second year in the league, we can assume that the league as on average 763 plate appearances to refer to for non-rookies going into year two, 112% more than for rookies. (Remember, year one in my data set isn’t the first season in the league for non-rookies, just the first season in the data set. That means pitchers have had a longer look at those hitters than rookies.)

I should also note, though, that the differences observed rookies and non-rookies were not clearly significant. If you compare the correlations for each category, the closest we come is between rookies and non-rookies with only two years of data (p-value of .087). For rookies and non-rookies with three years worth of data the p-value was .24. And what about off-speed offerings? Obviously, if the fastball percentages are fluctuating yearly, we should see changes in off-speed percentages. For that, I decided to use the specific pitch-type categories — rather than a composite. The table below shows correlations for sliders, curveballs and changeups for all hitters with data for only years one and two, as well as those with data for all three years:

N YR1 to YR2 YR1 to YR3
SL% (pfx) – two years only 142 .499** NA
SL% (pfx) – three years 367 .673** .663**
CU% (pfx) – two years only 142 .335** NA
CU% (pfx) – three years 366 .363** .240**
CH% (pfx) – two years only 142 .650** NA
CH% (pfx) – three years 367 .595** .541**
**Sig at the .01 level

At first glance, sliders and changeups pop as the most consistent off-speed offerings that hitters face each year. Sliders in years one and two have a .673 correlation, and the correlation between years one and three is only slightly different (.663). Changeups also show a fairly strong consistency, even three years out. Curveballs, however, start with a low correlation (.363) and get even less consistent by year three (.240). As with fastballs, I wondered whether there would be a difference between rookies and non-rookies. Here are the results:

N YR1 to YR2 YR1 to YR3
SL% (pfx) – Rookies 76 .726** .650**
SL% (pfx) – Non-rookies 291 .660** .669**
CU% (pfx) – Rookies 76 .216 .343**
CU% (pfx) – Non-rookies 290 .405** .215**
CH% (pfx) – Rookies 76 .577** .600**
CH% (pfx) – Non-rookies 291 .599** .531**
**Sig at the .01 level

Outside of sliders, non-rookies showed a higher correlation between the off-speed pitches they faced between years one and two. Sliders were highly correlated for rookies between years one and two (.726); the relationship between year one sliders and year three sliders for non-rookies was nearly identical (.669 vs. .660). Curveballs were inconsistent regardless of hitter type. Rookies also saw their year-three correlations increase for both curves and changeups, versus year-two. Both were higher than non-rookies.

As with fastballs, I wanted to see if the difference between the rookie and non-rookie correlations was significant. It was but only in the case of curveballs between years one and two (p-vale of .05).

There are a factors that likely impact the data and patterns we see. One of those factors is age. The percent of fastballs faced increases in a fairly linear fashion as batters age. This increase corresponds somewhat to hitters’ run-creation ability as they age. Take a look at the graph below:

The percent of fastballs faced increases slightly until about age 26. Pitchers then throw hitters fewer fastballs until about age 30. The rate then increases dramatically as hitters age. If we think about a hitter’s productivity through time then this makes sense. Hitters who last until age 30 are likely better-than-average and will have hit their offensive peak during their age-25 through age-29 seasons. As these hitters begin to age, they’re likely losing bat speed and making pitchers throw them more fastballs. There are obviously exceptions, but the pattern passes the initial sniff test.

As an example, here is Justin Upton‘s fastball percentage versus his wRC+ since 2007:

The pattern resembles the story above. As Justin Upton entered his peak offensive year, he saw a decrease in the percentage of fastballs thrown to him. So far this season, he’s seen fewer fastballs — and while his wRC+ is currently below average, my guess is he’ll finish his age-27 season north of 100.

So back to my original question about Ike Davis: Is the change in the percentage of fastballs he’s seeing that uncommon? The answer appears to be no.

While the low rate of fastballs is rather unique, the way in which the league adjusted to him between years one, two and three is in line with what we see in general. Rookies see a higher correlation between their year-one and year-two fastballs and less of a correlation between years one and three. Davis’ percent of fastballs seen has been 56.8%, 56.7% and 51.4%. Also, the 5%+ decline in fastballs seen is far from an outlier, as there have been 69 seasons where hitters saw the same or greater decline in fastball percentage since 2007, not including Davis.

Obviously, there is more to adjustments than pitch-type distributions. I’ll eventually get a chance to look at location and velocity, but if PITCHf/x and database-savvy readers want to tackle this, please jump in.




Print This Post



Bill works as a consultant by day. In his free time, he writes for The Hardball Times, speaks about baseball research and analytics, consults for a Major League Baseball team and appears on MLB Network's Clubhouse Confidential. Along with Jeff Zimmerman, he won the 2013 SABR Analytics Research Award for Contemporary Analysis. Follow him on Tumblr or Twitter @BillPetti.

24 Responses to “How the League Adjusts to Hitters Over Time”

You can follow any responses to this entry through the RSS 2.0 feed.
  1. Eminor3rd says:

    SUPER interesting, Bill. Thanks for this.

    Vote -1 Vote +1

  2. Steve K says:

    Fastballs seem to be a main course for a batter, I wonder what attributes make a fastball the pitch of choice? Why do pitchers not just throw 50% curveballs, or sliders? What attributes of a pitch thrown make it preferable to others?

    I don’t know the answer, I am just curious about the decision making process used for which pitches get thrown.

    Vote -1 Vote +1

    • vivalajeter says:

      The obvious answers are that it’s easier on the arm, it’s easier to control location, and even though a bad fastball is easy to crush, there’s nothing worse than a hanging curveball.

      Vote -1 Vote +1

    • payroll says:

      Most pitchers are able to command their fastballs better than other pitches.

      Vote -1 Vote +1

    • Steve A says:

      Fastballs also make other pitches more effective. If a pitcher has a good fastball, then hitters have to get ready for it making them less prepared for offspeed pitches. I think of Craig Kimbrel because his slider has been so excellent according to pitch values. I think that is mostly because of his fastball, which I believe is the best in baseball in terms of speed and upward movement.

      Basically curveballs and sliders and such are probably better when isolated, but they would lose effectiveness if they were all a pitcher threw.

      Vote -1 Vote +1

  3. mcbrown says:

    This is an interesting line of inquiry. I’m not sure correlation on individual pitch frequency is the right metric, because low or even declining year-to-year correlation doesn’t exactly imply adjustment, necessarily. But I can’t think of a more appropriate one off the top of my head; I’ll give it some thought and maybe re-post later.

    If we do stick with correlation the analysis might benefit from some kind of control set. As a comparison you might look at the correlation from year N to N+1 and N to N+2 for ALL players (regardless of age) – intuitively you would probably think those correlations should be stable, but maybe they’re not, in which case there could be another variable at play here.

    Vote -1 Vote +1

  4. AJS says:

    “There have been only 30 seasons between 2007 and 2011 where a hitter with more than 100 plate appearances saw fewer fastballs in a season — and only five where a player accumulated more than 500 plate appearances.”

    Can you explain what this means? Fewer fastballs than what?

    Vote -1 Vote +1

  5. Rufio Magillicutty says:

    Great article, valiant effort

    Premature conclusions were drawn from the “Fastballs Seen vs Runs Created by Age” graph IMO. I would say its more random than driven by any trend, judging by the contradictory spikes between ages 33-35. The data has to reside somewhere amongst a sequence of numbers, doesn’t mean there is a trend to uncover. However your data on the year to year correlation of fastballs seen as fascinating, and contains significant information.

    Like you mentioned, one could use pitch velocity rather than pitch type to resolve the dataset. Certainly grand assumptions would be made for the league as a whole concerning batters facing a uniform distribution of pitcher styles, though this will counter the error in pitch type identification

    Vote -1 Vote +1

  6. Richie says:

    If Ike Davis is seeing 6% fewer fastballs, that means 1 out of every 17 pitches he sees has changed (in that way). Basically means he sees one less fastball per game than he did 2 years ago.

    Just because something can be shown as statistically significant, doesn’t mean it really is. The clear lesson derived here is that pitchers don’t adjust. Hardly, anyways. Either they’re happy with the scouting reports they’re getting re minor league hitters tendencies. More likely, they throw whatever it is that their good at throwing.

    Vote -1 Vote +1

    • James Lewis says:

      “Just because something can be shown as statistically significant, doesn’t mean it really is. The clear lesson derived here is that pitchers don’t adjust.”

      But see, that’s exactly the point of statistical significance – the odds that the observed difference is a result of random variation as opposed to an actual change in pitch selection is only 1 in 100. In this specific case the impact of the adjustment may be minor, but suggesting that this is somehow evidence that pitchers don’t adjust is categorically incorrect, and proven so by the author here.

      Vote -1 Vote +1

      • Richie says:

        I understand full well what statistical significance is, having taken 3 stats classes. Yes, the statistical analysis here shows good reason to believe that pitchers as compared to 2 years ago have changed 1 out of every 17 pitches that they throw to Ike Davis. And such a change is an insignificant one. Especially considered over the length of two years.

        Vote -1 Vote +1

      • Anon21 says:

        Given that it’s the pitchers who choose when to throw him the off speed stuff, it might actually be quite significant. Even Beast Mode Bonds probably saw plenty of straight stuff when the Giants were down by 5 and the bases were empty. If Ike Davis is seeing predominantly off-speed stuff in his highest leverage at-bats (which, remember, may average out to fewer than one per game) where a year ago he’d have gotten mostly fastballs, that could seriously limit his production.

        Vote -1 Vote +1

      • Anon210 says:

        Given that it’s the pitchers who choose when to throw him the off speed stuff, it might actually be quite significant. Even Beast Mode Bonds probably saw plenty of straight stuff when the Giants were down by 5 and the bases were empty. If Ike Davis is seeing predominantly off-speed stuff in his highest leverage at-bats (which, remember, may average out to fewer than one per game) where a year ago he’d have gotten mostly fastballs, that could seriously limit his production.

        Vote -1 Vote +1

  7. Cozar says:

    I notice Ike Davis has a lower BB% and higher K% this year, so the data might reflect Ike Davis’ adjustment at the plate rather than pitchers’ adjustment to him.

    For example, in the past, Ike Davis would take a 1-2 curve for a ball, forcing a pitcher to throw him a 2-2 fastball. This year, Davis is swinging and striking out at that 1-2 curve, so the pitcher doesn’t throw that extra fastball.

    Vote -1 Vote +1

    • Cozar says:

      Upton doesn’t make sense either. If throwing more fastballs was successful in bringing down his wRC+, why would pitchers adjust by throwing less than next year? Looking at Upton’s numbers, in his 140 wRC+ year, he say .40 fewer pitchers per PA, a lower BB%, lower K%, lower BABIP, higher ISO and higher swing%. Again, at least part of the decrease in fastballs might be explained by Upton’s tendency to swing more often at early offspeed pitches, reducing the number of late count fastballs that he saw.

      Also, according to his player page on this site, Upton is only 24 (he turns 25 in August) so either your data is off or his birth year is.

      Vote -1 Vote +1

      • Bill Petti says:

        I don’t think I’d assume that more fastballs brought down his wRC+ two years ago. The larger point is that as hitters move into their prime they are presumably more dangerous against fastballs, so the league may adjust by decreasing the percentage they see. As hitters age, the percentage of fastballs will increase as bat speed and quickness decline.

        Your general point that hitter adjustments play into pitcher adjustments is a good one, as I think that’s another angle that isn’t covered by this initial pass.

        The age was an error. I’ve updated the chart. Thanks.

        Vote -1 Vote +1

    • vivalajeter says:

      Good point, Cozar. Ike looks completely lost at the plate, and seems to be falling behind in almost every at bat. Maybe their pitching strategy is the same as it always was, but he’s just in a lot more 0-1, 0-2, 1-2 counts, and that’s why he’s seeing less fastballs.

      One other potential issue: I seem to recall the Mets facing an absurd number of lefties earlier in the year. Lefty/Lefty matchups seem to get a high number of curveballs compared to Lefty/Righty. Is there pitch-type data based on whether it’s a Righty or Lefty pitcher?

      Vote -1 Vote +1

  8. payroll says:

    As a corollary, when a starter shifts his repertoire over time, how do the weighted values of his fastballs and off-speed pitches fluctuate?

    Vote -1 Vote +1

  9. fresheee says:

    What’s the reason with the dramatic change in the ranges of the Y-axes between the first and second line graph? Is it not more difficult to recognize the pattern when the scale has changed?

    It’s entirely possible my brain is simply incapable of dealing with two Y-axes.

    Vote -1 Vote +1

  10. Brian says:

    I don’t buy the result of this article in general or re: Davis either.

    In this article, by the same Mr. Petti, he pointed out that in one game, the Braves only threw one fastball to Ike out of eighteen pitches: http://www.fangraphs.com/blogs/index.php/braves-provide-preview-of-how-to-approach-ike-davis/

    This by itself takes such a large chunk out of the 5% difference discussed above to make it meaningless.

    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 *