Arizona Diamondbacks Atlanta Braves Baltimore Orioles Boston Red Sox Chicago Cubs Chicago White Sox Cincinnati Reds Cleveland Indians Colorado Rockies Detroit Tigers Houston Astros Kansas City Royals Los Angeles Angels Los Angeles Dodgers Miami Marlins Milwaukee Brewers Minnesota Twins New York Mets New York Yankees Oakland Athletics Philadelphia Phillies Pittsburgh Pirates San Diego Padres San Francisco Giants Seattle Mariners St. Louis Cardinals Tampa Bay Rays Texas Rangers Toronto Blue Jays Washington Nationals
 No Previous Article << Previous Column Aim For The Head: Good... (05/09) Next Column >> Aim For The Head: Barr... (05/23) No Next Article

May 16, 2001

## Expected vs. Actual Wins

This week's question comes from Chuck Hildebrandt, who writes:

Being a lifelong Detroit Tiger fan, I was studying the 1984 season in Total Baseball when I was startled by something I saw in the National League. I noticed that the New York Mets, managed by Davey Johnson, were outscored by their competition, 652-676, yet finished the season with a 90-72 record.

This is a stunning display of overachievement, a team being outscored yet finishing with a .556 winning percentage. Being too lazy to pursue the issue myself, I ask the question of you: what are the greatest differentials, both positive and negative, between expected and actual winning percentage for any single team?

Bonus philosophical question: What do such differentials really say about a manager's influence on a team, versus dumb luck?

Thanks for the question, Chuck.

First, let's limit ourselves to teams that played at least 100 games in a season, as some teams in the early days of professional baseball played incomplete schedules. In order to find a team's expected number of wins, we'll use the Pythagenport formula--Clay Davenport's refinement of Bill James's original Pythagorean formula. The formula is:

```Win% ~=  R^E/(R^E + RA^E) , where E = 1.5*Log((R+RA)/G) + 0.45
```

Given an estimated expected winning percentage, we can compute the difference between a team's actual and expected records either based on winning percentage or difference in wins.

(To answer this question, I'm using the free downloadable statistics database at www.baseball1.com. The database is currently unavailable due to server problems, but is expected to be back online soon.)

Chuck, you have a keen eye. The 1984 Mets turned out to be the second-biggest overachievers ever, at 11.7 wins above expectation, beaten only by the 1905 Tigers, who went 79-74 while scoring 512 runs and allowing 602. They "should" have gone about 66-88, but instead managed to be five games over .500, and 12.8 wins over expectation.

On the underachieving side, there are two teams that lost 13 or more games beyond expectation. The worst underachievers were the 1993 Mets, whose 672 R/744 RA differential should have been good for a 73-89 record. They instead posted a gruesome 59-103 record, or 14.3 wins below expectation. The other unlucky team was the 1986 Pirates, who scored 663 runs and allowed 700, which should have earned them a 77-85 record, but instead they went 64-98, a 13-game differential. The worst underachievers who actually outscored their opponents were the 1907 Reds, who scored 526 runs and allowed 519. They projected to a 79-77 record, but actually went 66-87.

Here's a list of all teams with differentials of 10 or more games:

```G500 = Games over .500
Pyth = Pythagenport expected winning percentage
P_W  = Pythagenport expected wins
P_L  = Pythagenport expected losses
DIF% = Difference between actual and expected winning percentage
DIFW = Difference between actual and expected wins
```

 YEAR TEA LG G W L WIN% G500 R RA PYTH P_W P_L DIF% DIFW 1993 NYN NL 162 59 103 .364 -44 672 744 .453 73.3 88.7 -.089 -14.3 1986 PIT NL 162 64 98 .395 -34 663 700 .475 77.0 85.0 -.080 -13.0 1907 CIN NL 156 66 87 .431 -21 526 519 .506 78.9 77.1 -.074 -12.9 1905 DET AL 154 79 74 .516 5 512 602 .430 66.2 87.8 .086 12.8 1911 PIT NL 155 85 69 .552 16 744 557 .630 97.6 57.4 -.078 -12.6 1905 SLA AL 156 54 99 .353 -45 511 608 .425 66.3 89.7 -.072 -12.3 1905 CHN NL 155 92 61 .601 31 667 442 .671 104.0 51.0 -.070 -12.0 1975 HOU NL 162 64 97 .398 -33 664 711 .469 75.9 86.1 -.071 -11.9 1984 PIT NL 162 75 87 .463 -12 615 567 .535 86.7 75.3 -.072 -11.7 1984 NYN NL 162 90 72 .556 18 652 676 .484 78.3 83.7 .072 11.7 1967 BAL AL 161 76 85 .472 -9 654 592 .544 87.6 73.4 -.072 -11.6 1946 PHA AL 155 49 105 .318 -56 529 680 .390 60.4 94.6 -.071 -11.4 1955 KC1 AL 155 63 91 .409 -28 638 911 .333 51.6 103.4 .076 11.4 1954 BRO NL 154 92 62 .597 30 778 740 .524 80.7 73.3 .073 11.3 1937 CIN NL 155 56 98 .364 -42 612 707 .434 67.2 87.8 -.070 -11.2 1917 PIT NL 157 51 103 .331 -52 464 595 .396 62.2 94.8 -.065 -11.2 1970 CIN NL 162 102 60 .630 42 775 681 .560 90.8 71.2 .069 11.2 1935 BSN NL 153 38 115 .248 -77 575 852 .321 49.1 103.9 -.073 -11.1 1972 NYN NL 156 83 73 .532 10 528 578 .461 71.9 84.1 .071 11.1 1924 SLN NL 154 65 89 .422 -24 740 750 .494 76.0 78.0 -.071 -11.0 1890 CL6 AA 140 79 55 .590 24 831 617 .643 90.0 50.0 -.053 -11.0 1906 CLE AL 157 89 64 .582 25 663 482 .636 99.8 57.2 -.054 -10.8 1919 WS1 AL 142 56 84 .400 -28 533 570 .470 66.8 75.2 -.070 -10.8 1924 BRO NL 154 92 62 .597 30 717 675 .528 81.4 72.6 .069 10.6 1999 KCA AL 161 64 97 .398 -33 856 921 .463 74.6 86.4 -.066 -10.6 1904 CLE AL 154 86 65 .570 21 647 482 .626 96.4 57.6 -.056 -10.4 1993 SDN NL 162 61 101 .377 -40 679 772 .440 71.3 90.7 -.063 -10.3 1911 CHA AL 154 77 74 .510 3 719 624 .566 87.1 66.9 -.056 -10.1 1970 CHN NL 162 84 78 .519 6 806 679 .580 94.0 68.0 -.062 -10.0 1932 PIT NL 154 86 68 .558 18 701 711 .493 76.0 78.0 .065 10.0 1961 CIN NL 154 93 61 .604 32 710 653 .539 83.0 71.0 .065 10.0

Of course, teams play 162 games today, versus 154 or fewer in years past, so it's a little easier to run up a larger differential over more games. If we look just at differences in winning percentage, there are nine teams that were 75 points or more off expectation. The '93 Mets still top the list, at 89 points below expectation, but a new team, the 1981 Reds, turns out to be the biggest overachiever, exceeding its expected winning percentage by 87 points (going 66-42, .611 versus a projection of 56.6-51.4, .524). Other teams with 75+ point differentials but not a 10-game overall difference include the 1884 Chicago White Stockings (later known as the Cubs) as 78-point underachievers, and the 1894 New York Giants as 76-point underachievers.

Let's consider Chuck's second question: "What do such differentials really say about a manager's influence on a team, versus dumb luck?" Strategic blunders by the manager can certainly influence a team's record, but the magnitude of this effect over the course of a season is hard to estimate. A manager probably has a more important influence on his team in playing the right lineup, managing the pitching staff, keeping his bench fresh, and so on, than in specific game tactics.

A team that underachieves its projection as badly as the teams we're talking about probably lost more than its share of one-run games, which can be caused in part by a lousy bullpen. The 1999 Royals, who are in the table above, had a terrible bullpen, possibly one of the worst ever. When no one is getting the other guys out, it's hard to blame all of that on the manager.

Of course, the arguments above aren't very sabermetric. Let's ask a slightly different question: are teams who underachieve or overachieve likely to continue doing so the next season? This doesn't necessarily answer the question about the manager's impact, because a manager's job is somewhat more at risk following a season that didn't meet expectations, but it's a place to start.

I took the list of all teams with 100+ games and compared their DIFW in one season to the next (assuming the franchise still existed). I computed the correlation of the two differentials. If the correlation was close to 1.0, then teams were more likely to have the same kind of differential (above or below average) the following season. If the correlation was close to -1.0, the reverse is true, which would mean that teams that overachieve one year are more likely to underachieve the next. A value close to zero means that there's no relationship between the two, that nothing from the team's "luck" carries over to the next season. The actual correlation was +0.05, which is pretty close to zero, and suggests that there's no relationship.

We can refine the question a little bit more; since teams who overachieve are more likely to retain their manager, we can focus only on teams that were significant overachievers. I selected five or more games as a threshold. Plotting their win differentials in the following season, we get the following chart, which shows no real trend or pattern, furthering the theory that the manager has little consistent impact on whether a team over- or underachieves it's expected Pythagenport projection.

* * * * *

Regarding whether Wes Ferrell's Hall of Fame case is enhanced by his offensive production, Kevin Morse writes: "He's certainly more deserving than his Vet Committee-elected brother Rick."

Brian Simpson asks: "What about Orel Hershiser? I seem to remember him being a fairly good hitter before he got hurt." Indeed, Hershiser was pretty good with the stick for a pitcher. His best season was 1993, when he posted a 784 OPS (.356/.373/.411), but that was his only season with an OPS over 600 in 50 or more plate appearances.

Mike Ritzema writes: "I read your article and I was wondering where Darren Dreifort would place. I saw his two bombs against Chicago last year and it gives me hope that he'll hit for his money, too." Dreifort's two bombs helped him to just a 520 OPS last year (.210/.246/.274), his best year to date.

David (no last name given) inquires: "Interesting article about historical pitchers hitting performances. Can those numbers be converted to some familiar sabermetric figures--runs above average, games won v. average, etc. Basically, how does a good hitting pitcher affect a team's ability to win?"

Great question. The upper limit seems to be about 20 runs, for the very best hitting pitching seasons, as shown in the following chart (PMLV is the number of runs contributed on offense above what a league average pitcher would have hit, adjusted for park and league):

 YEAR NAME TEA LG PA AVG OBP SLG PMLV 1935 BOS A 171 .347 .427 .533 28.1 1955 BRO N 124 .359 .395 .632 25.6 1965 LA N 136 .300 .331 .508 25.6 1925 WAS A 101 .433 .455 .577 24.1 1923 CLE A 151 .361 .391 .472 22.4 1958 MIL N 117 .333 .385 .463 22.1 1930 CIN N 130 .336 .423 .442 22.1 1931 CLE A 126 .319 .373 .621 22.0 1930 NY A 106 .374 .415 .596 20.9 1923 NY N 92 .427 .446 .573 20.4 1959 CIN N 122 .305 .402 .410 20.2 1950 CLE A 150 .272 .340 .485 20.0 1943 PHI N 136 .300 .382 .458 20.0

In today's game, pitchers don't throw as many innings or complete games, and rarely would get a chance to bat often enough to clear 20 runs of value. Hershiser's 1993 was worth about 13.5 runs, and that's about the top end for the past couple of decades.

Keith Woolner is an author of Baseball Prospectus.
Click here to see Keith's other articles. You can contact Keith by clicking here