Do Pitchers Try Harder for Their 20th Win?

1
Do Pitchers Try Harder for Their 20th Win?

• Phil Birnbaum
• www.philbirnbaum.com
• July 31, 2009

2
Do pitchers try harder for their 20th win?

• Bill James, "The Targeting Phenomenon"
• Historically, there are more players who finished
  with one win than with two
• More with 2 than with 3. More than 3 than with 4
  
• Every number is harder to hit than the previous,
  except
• There are more pitchers with 20 wins than 19.

3
Do pitchers try harder for their 20th win?

• Why?
• Bill James "players WANT to wind up the season
  hitting .250, rather than in the .240s. They
  tend to make it happen."
• But HOW do they make it happen?

4
Do pitchers try harder for their 20th win?

• Clutch pitching?
• Do players actually try harder when they can win
  their 20th?
• That would mean they wouldn't be trying their
  hardest in other games
• Not a nice theory

5
Season win totals, 1940-2007
6
Maybe it should look more like this
7
Why?

• Five possible factors I could think of
• There could be more
• Take them one at a time

8
Factor 1 extra starts

• Maybe pitchers are given an extra start late in
  the season to try to get to 20
• In that case, those pitchers would have a larger
  proportion of starts in September

9
Percentage of starts in September for pitchers
who eventually finish with

• 16 wins 17.53
• 17 wins 17.77
• 18 wins 18.36
• 19 wins 18.49
• 20 wins 18.47
• 21 wins 18.15
• 22 wins 18.18
• -- A little bulge at 19-20, maybe 0.25

10
Percentage of starts in September

• The 19-20 win pitchers had 9229 starts
• An extra 0.25 means an extra 23 starts
• Maybe 10 wins
• By this analysis
• 19-win pitchers, circumstances created 10 fewer
  (-10) of them
• 20-win pitchers, 10 extra (10)

11
Factor 2 relief appearances

• Maybe 19-win pitchers got a relief appearance to
  try to reach 20
• Check the historical record
• Big metaphorical wet kiss to Retrosheet

12
Factor 2 relief appearances

• Eventual 20-game winners with relief wins
• 1951 Early Wynn wins 18th
• 1951 Mike Garcia wins 19th
• 1956 Billy Hoeft wins 20th
• 1957 Jim Bunning wins 20th
• 1966 Chris Short wins 20th
• 1991 John Smiley gets 19th
• 1997 Randy Johnson gets 20th
• 7 extra wins

13
Factor 2 relief appearances

• Eventual 19-game winners with relief wins 3
• Eventual 20-game winners with relief wins 7
  (previous slide)
• Eventual 21-game winners with relief wins 6
• Take these numbers at face value, since they're
  exact historically

14
Factor 3 clutch pitching

• The obvious question did they just pitch better
  with 19 wins, with the 20th on the line?
• How did the pitchers actually perform with
  various numbers of wins?

15
Team RA and win pct. for pitchers

• 17 wins 3.72 RA, .658
• 18 wins 3.54 RA, .652
• 19 wins 3.54 RA, .655
• 20 wins 3.62 RA, .615
• 21 wins 3.53 RA, .676
• 22 wins 3.34 RA, .774
• (Note RA is for team, not just the starter wins
  are at the time of the start, not at end of
  season)
• Only a tiny bit of difference maybe .08 RA?
• I feel better pitchers are still team players

16
Team RA and win pct. for pitchers

• Move the 20-win group from 3.62 to 3.54
• Difference 39 runs over 490 starts 4 wins
• Four 20-win pitchers "should have" moved to 21
  wins
• So four extra 20s, four fewer 21s

17
But wait!

• The 20-win group's RA was only a bit higher than
  you'd expect
• But their winning percentage was much, much too
  low!
• 18 wins 3.54 RA, .652
• 19 wins 3.54 RA, .655
• 20 wins 3.62 RA, .615
• How come? Run support.

18
Factor 4 run support

• 19 wins 3.54 RA, .655, 4.45 RS
• 20 wins 3.54 RA, .615, 4.05 RS
• 21 wins 3.62 RA, .676, 4.46 RS
• Holy crap!
• After achieving their 20th win, pitchers' batters
  let them down in their tries for 21
• A huge 0.4 run per game shortfall!

19
Factor 4 run support

• There were 490 starts by pitchers with exactly 20
  wins
• 0.4 runs per game is 196 runs
• That's 20 wins! Maybe 15 of those wins would have
  gone to the starter
• So 15 pitchers got "stuck" at 20 wins instead of
  moving to 21
• That's an extra 15 twenty-game winners, and 15
  fewer twenty-one-game winners.

20
Factor 5 more decisions

• Maybe when a pitcher is going for 20, the manager
  will leave him in longer
• Looks like it! Wins per start, 17 to 22 wins
• .484, .496, .521, .463, .505, .554
• Suppose they should have been .505. That's 11
  extra wins.
• So we have an extra 11 twenty-game winners, and
  11 fewer nineteen-game winners.

21
Totalling it up
Starts Relief Clutch Support Decisions Total
19 10 3 11 18
20 10 7 4 15 11 47
21 6 4 15 13
22
One last adjustment

• However some of these pluses and minuses need to
  "move up" a category
• Example a manager gives his starter an extra
  relief appearance he wins his 20th. But five
  days later, he wins his 21st.
• We think that should be a move between 19 and 20,
  but it really wound up as a move between 20 and
  21
• I'm arbitrarily going to adjust
• 19s from 18 to 16
• 20s from 47 to 37
• 21s from 13 to -7

23
Final score

• Final total
• 19 game winners 16
• 20 game winners 37
• 21 game winners 7
• If we back all these effects out of the original
  data, there should be no more bulge at 20

24
Not perfect, but not bad
25
Summary

• By this estimate, there were 37 "extra" 20-game
  winners compared to expected.
• 10 because of extra starts in September
• 7 because of relief appearances in September
• 3 because 20-game pitchers didn't pitch well
  enough to get to 21
• 11 because 20-game pitchers got such bad run
  support that they couldn't get to 21
• 6 because managers left the pitchers in longer in
  hopes they'd get their 20th that day.

26
Summary

• So is it because pitchers want it to happen?
• No it's mostly because managers want it to
  happen.
• Broken down
• 23 manager decisions
• 3 pitchers pitching worse when already at 20
• 11 run support luck.

27
Reference

• Bill James, "The Targeting Phenomenon," The Bill
  James Gold Mine 2008, p. 67