LOGIN

One of the keys to a successful fantasy season is not drafting a pitcher who misses an extensive amount of time or performs much worse than the previous year.  Anyone who drafted Chris Carpenter or Dontrelle Willis in 2007 or Rich Hill or Aaron Harang in 2008 can attest to this.

There have been several theories posed in traditional publications (SI.com – Tom Verducci) and in the fantasy baseball blogosphere (FantasyPhenoms, Beyond the Box Score, RotoAuthority) that claim a correlation between a pitcher’s innings/pitches from the previous year and their performance the next year.

While these articles mentioned several positive examples, none of them really tested their theories to see if they are true predictors of either injury or decreased performance.  We decided to put them to the test along with several theories we’ve been kicking around:

  • High Pitch Volume – Does a high volume of pitches have a carryover effect the next year?  (posed by FantasyPhenoms and RotoAuthority)
  • Spike In Pitch Volume – What is the effect of a significant increase in pitch volume vs. previous year?  (posed by Tom Verducci)
  • New To The Workload – Is a pitcher who first reaches the 2,700 pitch threshold more likely to fall back the next year vs. a pitcher used to the workload?
  • High % of Breaking Pitches – Is a pitcher who throws breaking balls more susceptible to fall back vs. a fastball/changeup pitcher?

We also added a Common Sense test which predicted that weaker pitchers from year prior (4.00+ FIP) would be more likely to get hurt, moved to the bullpen, or demoted to the minors than stronger pitchers.  This ‘common sense’ test also serves as a proxy for a typical fantasy baseball drafter’s judgment as they would likely have more faith in pitchers who were successful in the previous year.

For the test, we focused on the past three years (2006-2008) and on pitchers who threw at least 2,700 pitches (~ 155-160 IP) in that year.  Pitchers who retired/semi-retired the next season for reasons other than injury were removed.  This left a total of 247 pitcher seasons.  We defined ‘miss significant time’ as someone who threw 2,000 or less pitches (~ 20 GS) in MLB the next year and ‘measurable decrease in performance’ as a 0.50 or more increase in their FIP (this is an ERA variation that stands for Fielding Independent Pitching based on only things within a pitcher’s control – K, BB, HR – and is more stable year-to-year than ERA).  If a pitcher threw less than 2,000 pitches the next year, we ignored their FIP so as not to add insult to injury (ha!).

It’s worth noting that throwing 2,000 or less pitches in MLB does not necessarily mean someone was injured – it could also mean their performance decreased to a level that was no longer MLB-worthy and they were demoted.  Since these amount to the same thing for a fantasy team owner (bupkus), we did not differentiate.  We also did not credit for pitches thrown in the minor leagues prior to arriving in the MLB or pitches thrown in the postseason.  This is partly out of convenience (our data source FanGraphs does not include this in their ‘Leaders’ section) and partly out of intent (a pitch in the minor leagues is not as stressful as a pitch in the major leagues).

In total, 59 (24%) of the 247 qualified pitcher seasons were followed up by seasons of less than 2,000 pitches and 53 (21%) were followed up with FIP increases above 0.50.  That means that about 45% of starting pitchers either pitched significantly less or had a measurable decrease in performance in the following year.  Ouch!

Below are the test results:

Test Group (2006-2008, 2700+ Pitches)
Pitching Performance vs. Previous Year
Test # Previous Year Pitch Volume Total <2000 Pitches FIP Up 0.50+ Combined Index
# % # % # %
1

30+% Sliders/Curve Balls

54 18 33% 16 30% 34 63% 137
2 27+% Sliders/Curve Balls 80 27 34% 20 25% 47 59% 128
3 +700 Pitches vs. Previous Year 56 19 34% 14 25% 33 59% 128
4 1st Year at 2700+ Pitches 49 17 35% 9 18% 26 53% 115
5 FIP 4.00+ 153 44 29% 25 16% 69 45% 98
6 6500+ Pitches Previous 2 Years Combined 76 10 13% 21 28% 31 41% 89
7 9000+ Pitches Previous 3 Years Combined (2007-2008 only) 73 15 21% 14 19% 29 40% 86
8 3500+ Pitches 23 0 0% 9 39% 9 39% 85
9 3400+ Pitches 43 2 5% 14 33% 16 37% 81

The best theory proved to be the High % of Breaking Pitches Theory as both the 27+% and 30+% thresholds finished in the top two for index (137 and 128 respectively – the 137 means that a pitcher who threw 30+% Sliders/Curve Balls in the previous year is 37% more likely than the average pitcher to fall below 2,000 pitchers or see their FIP increase 0.50+.  The Spike in Pitch Volume Theory (go Verducci!) and New to the Workload Theory also performed above average.

The ‘Common Sense’ test (FIP 4.00+) finished  at about a 100 index meaning picking a random pitcher would’ve been as effective as choosing one with this criteria.  This average performance may be misleading since it did a better than average job at predicting significant pitch decreases (29% to 24%) and its below average performance at predicting FIP increases may be because some of the FIPs were so high to begin with in the first place.

The High Pitch Volume Theory proved to be a relative failure.  We tried four variations – one-year (3400+ and 3500+), two year (6500+) and three year (9000+) and all of them finished at a below average rate.  Both the 3400+ and 3500+ theories did a good job at predicting 0.50+ FIP increases but that is tempered by the fact that they had a higher percentage of pitchers throw enough innings to qualify.  For instance, if you were to back out the 19 of 56 pitchers who fell below 2000 pitches in the Spike In Pitch Volume Theory, you would have 38% (14 of 37) that had 0.50+ FIP increases.  So there may be a slight correlation that a heavily used pitcher does not perform as effectively the next year but they will perform.

I’d theorize the failure of High Pitch Volume Theory in predicting significant pitch declines is because totaling high pitch counts in a year is more of a skill than an abuse.  It is easier for someone who has run a marathon in the past to run another one versus someone who has not.  While this group is not immune to missing time, they have built the stamina to at least pitch close to a full season (2700+ pitches).  Below are the 16 members of 2007’s “3400+ Pitch Club” .

Name Pitch Count
2007 2008
Carlos Zambrano 3692 3018
Dan Haren 3635 3339
Jake Peavy 3610 2860
Scott Kazmir 3609 2749
Aaron Harang 3591 3055
CC Sabathia 3581 3814
Gil Meche 3579 3555
Daniel Cabrera 3565 3020
Dontrelle Willis 3491 523
Jeff Francis 3485 2385
Joe Blanton 3481 3250
Dice-K 3480 2904
Javier Vazquez 3465 3376
Brandon Webb 3437 3358
Bronson Arroyo 3432 3436
Barry Zito 3411 3206
Average 3534 2991

The collective average went from 3,534 pitches to 2,991 pitches which, though a decline,  still represents about 29 starts.  8 of the 16 pitchers on this list finished in the top 30 in 2008 total pitches.  The below 100 index in the test and the above chart indicate that a pitcher with 3400+ pitches in the previous year is a sign of durability not vulnerability (note: given the high injury rate with pitchers, though, it’s not that strong of a sign).

We then tested several combinations of the three above-average performing theories and threw a last bone to the High Pitch Volume theory (which tanked even after adding the High % of Breaking Balls requirement).

Test Group (2006-2008, 2700+ Pitches)
Pitching Performance vs. Previous Year
Test # Previous Year Pitch Volume Total <2000 Pitches FIP Up 0.50+ Combined Index
# % # % # %
1

30+% Sliders/Curve Balls + 700+ Pitches vs. Year Prior

10 6 60% 2 20% 8 80% 177
2 27+% Sliders/Curve Balls + 700+ Pitches vs. Year Prior + First Time Over 2700+ Pitches 9 6 67% 1 11% 7 78% 172
3 27+% Sliders/Curve Balls + 1st Year at 2700+ Pitches 15 8 53% 2 13% 10 67% 147
4 27+% Sliders/Curve Balls + 700+ Pitchers Vs.Previous Year 20 9 45% 4 20% 13 65% 143
5 30+% Sliders/Curve Balls OR 700+ Pitch Spike 100 31 31% 28 28% 59 59% 130
6 27+% Sliders/Curve Balls OR 700+ Pitch Spike 116 37 32% 30 26% 67 58% 128
7 700+ Pitches vs Previous Year and 1st Year Over 2700 Pitches 32 13 41% 5 16% 18 56% 124
8 27+% Sliders/Curve Balls + 3400+ Pitches 13 0 0% 4 31% 4 31% 68

While the top two combo theories are very good predictors, they are also very rare – about 1 in 25 pitching seasons.  The most impressive predictor of the bunch is test #6 – 27+% Sliders/Curve Balls OR 700+ Pitch Spike.   Why?  Because that 128 index applied to a huge sample – nearly half of the pitcher seasons in the study.  The impact is best seen by comparing this group against its mirror image – pitchers who threw < 27% Sliders/Curve Balls AND had < 700+ Pitch Spike.

Comparison of Pitching Seasons After A 2700+ Pitch Season
Sample < 2000 Pitches Next Year Index FIP Up 0.50+ Next Year Index <2000 Pitches Next Year OR FIP Up 0.50+ Next Year Index
All Pitchers 247 59 (24%) 100 53 (21%) 100 112 (45%) 100
27+% Sliders/Curve Balls OR 700+ Pitch Spike 116 37 (32%) 134 30 (26%) 121 67 (58%) 128
<27% Sliders/Curve Balls AND < 700+ Pitch Spike 131 22 (17%) 70 23 (18%) 82 45 (34%) 76

So a pitcher that qualifies in at least one of these categories is 28% more likely to throw less than 2000 pitches in the next year or to see an FIP increase of 0.50+ than the average pitcher.  Just as important, one that qualifies for neither category is 24% less likely than the average pitcher to meet these respective fates.

That’s it for now but we’ll be revisiting this topic in the near future.  In the next post, we’ll go over the top 20 starting pitchers risks for 2008, perhaps uncovering a couple pitchers that you’ll be surprised to see…