Looking for Clutch Performance in One-Run Games
|
|
|
|
Looking for Clutch Performance in One-Run Games |
|
|
|
You used to hear a lot (and perhaps still do) about a particular team's
performance in one-run games and when you did, it was often used to
either praise or damn their ability to perform in clutch situations.
So I decided to look at which teams had the greatest differences
between their winning percentage in games both decided and not decided
by one run.
Here are the teams that improved the most in one-run games:
--- one-run games -- ------ others ------
Year Team G W L T Pct G W L T Pct Diff
1974 SD N 47 31 16 0 .660 115 29 86 6 .252 +.407
1955 KC A 45 30 15 0 .667 109 33 76 7 .303 +.364
1921 PHI N 45 25 20 0 .556 109 26 83 8 .239 +.317
1939 PHI N 50 25 25 0 .500 101 20 81 9 .198 +.302
1994 FLA N 36 23 13 0 .639 79 28 51 5 .354 +.284
And declined the most:
--- one-run games -- ------ others ------
Year Team G W L T Pct G W L T Pct Diff
1935 NY A 44 15 29 0 .341 105 74 31 2 .705 -.364
1948 CLE A 30 10 20 0 .333 125 87 38 2 .696 -.363
1963 MIN A 39 13 26 0 .333 122 78 44 3 .639 -.306
1947 NY A 51 22 29 0 .431 103 75 28 2 .728 -.297
1935 DET A 46 19 27 0 .413 105 74 31 2 .705 -.292
So what does this mean? Is this really an indication of clutch
performance? Not at all. Notice that the teams doing much better
in close games are very bad teams while the ones doing worse are
very good teams. My feeling is that upsets tend to be close games,
and that most very good teams do their worst in one-run games,
slightly better in two-run games, and so on. So rather than comparing
a team's performance in one-run games to their performance in other
games, perhaps we should look at how they do in those situations
compared to other teams of similar ability. To do this, I divided all
teams in groups by their overall winning percentage and looked at
median winning percentages in games decided by 1, 2, 3, 4 and 5 runs
or more. Here's what I found:
Wpct Teams 1-Run 2-Run 3-Run 4-Run 5+Run
< .375 149 .400 .360 .333 .316 .245
.375 - .425 212 .442 .414 .400 .400 .340
.425 - .475 338 .467 .458 .455 .444 .421
.475 - .525 386 .500 .500 .500 .500 .500
.525 - .575 417 .527 .538 .556 .556 .575
.575 - .625 255 .560 .586 .593 .611 .641
> .625 109 .600 .621 .667 .667 .733
Which, surprisingly enough, is exactly what I expected to see.
So dividing the games into only two groups (one-run games and all
others), we find the following median differences:
Wpct Teams 1-Run Other Diff
< .375 149 .400 .316 +.084
.375 - .425 210 .442 .382 +.060
.425 - .475 341 .467 .442 +.025
.475 - .525 383 .500 .500 +.000
.525 - .575 418 .527 .555 -.028
.575 - .625 256 .560 .614 -.054
> .625 109 .600 .674 -.074
After adjusting for the type of teams we're dealing with (subtracting
.084 if the team has a winning percentage of less than .375, and so
on), which ones did the best and worst in one-run games? Here's the
updated list of the best teams:
one-run games ---- others ---
Year Team W L T Pct W L T Pct Diff Adj
1974 SD N 31 16 0 .660 29 86 0 .252 +.407 +.323
1955 KC A 30 15 0 .667 33 76 1 .303 +.364 +.304
1981 BAL A 21 7 0 .750 38 39 0 .494 +.256 +.284
1994 FLA N 23 13 0 .639 28 51 0 .354 +.284 +.259
1972 NY N 33 15 0 .688 50 58 0 .463 +.225 +.253
And the worst:
one-run games ---- others ---
Year Team W L T Pct W L T Pct Diff Adj
1966 NY A 15 38 0 .283 55 51 1 .519 -.236 -.261
1929 NY N 15 28 0 .349 69 39 1 .639 -.290 -.262
1963 MIN A 13 26 0 .333 78 44 0 .639 -.306 -.278
1948 CLE A 10 20 0 .333 87 38 1 .696 -.363 -.289
1935 NY A 15 29 0 .341 74 31 0 .705 -.364 -.310
While these lists are still biased toward very bad teams (no team
that plays .705 in their other games can be expected to do much better
in one-run games), at least now there are two winning teams on the
"best" list and a losing team on the "worst".
Still, this adjustment doesn't help to explain the variations we often
see among teams of similar ability. For example, here are the best and
worst one-run teams since 1901 in each of our groups:
Overall one-run games ---- others ---
Wpct Year Team W L T Pct W L T Pct Diff
< .375 1974 SD N 31 16 0 .660 29 86 0 .252 .407
1981 SD N 12 30 0 .286 29 39 0 .426 -.141
.375 - .425 1955 KC A 30 15 0 .667 33 76 1 .303 .364
1919 WAS A 14 36 0 .280 42 48 2 .467 -.187
.425 - .475 1994 FLA N 23 13 0 .639 28 51 0 .354 .284
1966 NY A 15 38 0 .283 55 51 1 .519 -.236
.475 - .525 1959 PIT N 36 19 0 .655 42 57 1 .424 .230
1973 MIN A 12 27 0 .308 69 54 0 .561 -.253
.525 - .575 1981 BAL A 21 7 0 .750 38 39 0 .494 .256
1963 MIN A 13 26 0 .333 78 44 0 .639 -.306
.575 - .625 1913 WAS A 32 13 0 .711 58 51 1 .532 .179
1935 NY A 15 29 0 .341 74 31 0 .705 -.364
> .625 1908 PIT N 33 12 0 .733 65 44 1 .596 .137
1948 CLE A 10 20 0 .333 87 38 1 .696 -.363
What accounts for these variations within groups? Is it luck or
an ability (or lack thereof) to score and prevent runs when the
game is on the line? I usually groan when people start talking about
clutch performance (much like the dreaded "intangibles" which,
by the way, can now be accurately measured using a method I recently
developed called, simply, the "Rafael-Belliard-O-Meter"), but as bad as
my attitude might be on the subject, I would dearly love to be the
first one on my block to show that such a talent does exist.
If it isn't luck, you might expect that a penchant for winning close
games would stick around from year to year. Most people agree that
winning is not simply caused by good fortune, and (the Florida Marlins
notwithstanding) most teams with high winning percentages one year tend
to experience similar success the next. So my first attempt at proving
that clutch ability is to blame for teams excelling in one-run games
was to look at the variation in this area from year to year. In the
chart below, "Wpct" contains information on the delta from one year to
the next in a team's overall winning percentage, while "AdjD" contains
similar data on the adjusted difference between a team's performance in
one-run games and those decided by more than a single run. For example,
if our entire database consisted of the following three years:
-- Overall -- -- One-Run -- --- Others --
Year Team W L T WPct W L T WPct W L T WPct Diff AdjD*
1968 ATL N 81 81 1 .500 27 30 0 .474 54 51 1 .514 -.041 -.041
1969 ATL N 93 69 0 .574 28 17 0 .622 65 52 0 .556 .067 .095
1970 ATL N 76 86 0 .469 20 24 0 .455 56 62 0 .475 -.020 -.045
My chart would look like:
--- Wpct --- --- AdjD ---
Samples Avg StDev Avg StDev
2 .0895 .0155 .1380 .0020
Where:
.0895 = ( ( .574 - .500 ) + ( .574 - .469 ) ) / 2
.0155 = the standard deviation of .074 and .105
.1380 ( ( .095 - -.041 ) + ( .095 - -.045 ) ) / 2
.0020 = the standard deviation of .136 and .140
Here's what the chart looks like on all teams from 1901-1997:
--- Wpct --- --- AdjD ---
Samples Avg StDev Avg StDev
1838 .0594 .0455 .1003 .0770
Frankly, I was surprised that a team's overall performance varied as
much from year to year as it did, but that was nothing compared to the
variability in its success in one-run games. Still, this doesn't by
itself prove anything, since on one hand we were looking at a single
percentage involving (with some exceptions) between 154-162 games,
while on the other we were comparing two percentages involving around
50 and 100 games. Had I paid closer attention during my statistics
course in college, I probably could've done something more with these
results, but instead I decided to change the study slightly. Where
before I had compared each team's record to how it did the following
year, this time I compared its record to another team and year
selected at random. If I'm correct and success in one-run games is
merely a crap-shoot, the "Wpct" totals should jump quite a bit, but
the "AdjD" totals should stay about the same. The results:
--- Wpct --- --- AdjD ---
Samples Avg StDev Avg StDev
1838 .0952 .0711 .1034 .0776
In other words, how a team does one year in close games is absolutely
no use in predicting how it will do the next. Things like that are
usually called "the breaks of the game" or, more succinctly, luck.
After doing this study, I noticed an article on the same subject in
the 1997 Baseball Research Journal. The article by Bob Boynton, "Are
One-Run Games Special?" takes a very different route but arrives at the
same conclusion that I did.
Some loose ends:
As you might've guessed, 1935 was a strange year in the AL for one-run
decisions. Here are the standings, based on their one-run games,
compared to how they actually finished the season:
Team W L Pct GB Fin
BOS A 28 17 .622 - 4
CLE A 22 16 .579 2.5 3
STL A 21 16 .568 3 7
CHI A 24 19 .558 3 5
PHI A 19 16 .543 4 8
WAS A 20 28 .417 9.5 6
DET A 19 27 .413 9.5 1
NY A 15 29 .341 12.5 2
In the World Series that year, Detroit won all three of their one-run
games on route to a 4-2 victory over the Cubs. Go figure.
A team's record in one-run games does not also seem to be a good
indicator of the strength of the bullpen. For example: in 1997,
the Seattle Mariners were only slightly worse (.543 vs .560) in one-run
games, despite their historically awful bullpen, a smaller drop-off
than the Baltimore Orioles experienced (.560 vs .625) with an excellent
relief corps.
By the way, here are five teams that did exactly the same in both
situations:
--- one-run games -- ------ others ------
Year Team G W L T* Pct G W L T Pct Diff
1902 CIN N 38 19 19 0 .500 102 51 51 5 .500 .000
1992 NY N 54 24 30 0 .444 108 48 60 5 .444 .000
1993 LA N 54 27 27 0 .500 108 54 54 4 .500 .000
1957 BAL A 54 27 27 0 .500 98 49 49 7 .500 .000
1966 CLE A 60 30 30 0 .500 102 51 51 5 .500 .000
* You might have wondered why I included a "tie" category for one-run
games. After all, how can a team tie a one-run game? Well, it's
happened four times, in 1937, 1938, 1939 and 1940. The last time was
a 1-0 tie between the Yankees and White Sox on June 20th, 1940.
These were protested games in which the statistics were kept but the
loss and win were discarded.
Complete Data
Tom Ruane |
|
Comments or SuggestionsThanks for visiting the BasebaLL Think Factory. For information about contributing, to make a comment or suggestion, or to report a problem at my WWW site, please complete my Feedback Form. If your browser does not support forms, send e-mail to webguru@baseballstuff.com. |
|
|
Original material Copyright © 1996-1999
BasebaLL
Think
Factory
|
|
Back to the top of page | Baseball Scholars | Baseball Think Factory | BaseballStuff.com |