Update 1/25/2014: This article lays the groundwork for the Team Similarity Scores, but more accurate results are available here.

Many moons ago, Bill James came up with a way of mathematically classifying how similar two players are to one another. But to my knowledge no one has applied this logic to teams. So when Tom Tango asked me why I used the 1974 Orioles to make predictions about future Orioles teams, I asked myself: what teams should I be using as a comparison point? I didn’t quite answer that question (yet) but I did end up with a project that was as fun as that, if not more so.

I used the Lahman database to adapt James’s algorithm to teams. Like James, I started each comparison with 1000 points and subtracted points for differences between them:

• Subtract 1 point for each difference of 10 runs scored.
• Subtract 1 point for each difference of 10 runs allowed.
• Subtract 1 point for each strikeout by a batter.
• Subtract 1 point for each walk by a batter.
• Subtract 1 point for each home run by a batter.
• Subtract 1 point for each strikeout achieved by a pitcher.
• Subtract 1 point for each walk allowed by a pitcher.
• Subtract 1 point for each home run allowed by a pitcher.

There are certainly other dimensions on which to compare teams, but I think these aspects are a good start. Note that outside of runs scored & allowed, which measure a team’s talent more accurately and precisely than wins and losses, the other stats I used are the components of FIP.

Some notes:

• Lahman’s DB doesn’t have strikeouts (by batters) for many of the early-era teams. I set these values to 0, which makes some of the comparisons to these teams useless.
• I converted all stats to 162-game ones simply by multiplying each stat by (162/gamesPlayed). Nothing fancy there.
• I used the 2012 version of the database. I’ll update these scores with the 2013 version when that’s out of beta.

For an example of how this works, let’s look at some teams!

2001 Seattle Mariners

The ’01 Mariners tied the major-league record for regular-season wins when they went 116-46 to capture the AL West crown. Although they beat the Indians 3-2 in the ALDS, they lost the ALCS in five games to the Yankees (who, I must point out, eventually lost to the Diamondbacks). On offense, the team was led by Bret Boone, who hit .331/.372/.578 and played stellar defense to rack up 7.8 fWAR. It was also the debut of a Japanese player who wore his first name on his jersey: Ichiro Suziki was a force to be reckoned with, hitting .350/.381/.457 and compiling 6 fWAR. Mike Cameron (5.5 fWAR), Edgar Martinez (4.7 fWAR), John Olerud (4.6 fWAR), and David Bell (2.6 fWAR) also contributed excellent seasons.

Pitching-wise, Freddy Garcia gave the team 34 excellent starts. He benefitted from an abnormally low .255 BABIP, but the fact is his FIP was 3.48 that year and he accrued 5.3 fWAR. Arthur Rhodes and Joel Piniero also excelled at run prevention, finishing the season with a 2.14 and 2.86 FIP, respectively.

Which team throughout history was most like this powerhouse?

Enter the 1967 Detroit Tigers, who went 91-71 and finished one game behind the “Impossible Dream” Red Sox in the American League. That year the Tigers received 7 fWAR from right fielder Al Kaline (.308/.411/.541) and 5.8 from catcher Bill Freehan. Dick McAuliffe chipped in for 4.8 fWAR and Norm Cash, at age 32, handled himself nicely with 3.8 fWAR. On the mound, Mickey Lolich had an excellent season (2.65 FIP), as did Joe Sparma and Earl Wilson.

It may not seem like it because of the large disparity in wins, but despite playing over 30 years apart the two teams are extremely similar to each other: 

• Runs scored: 927 (Seattle) to 678 (Detroit). The ’67 Tigers actually scored 683 runs, but since they did so in 163 games, I prorated their season to a 162-game one. Hence the drop in runs scored. Anyway, the difference is 249 runs, or 24 points subtracted from the similarity score.
• Runs allowed: 627 to 583. Difference of 44 runs, or 4 points.
• Strikeouts: 989 to 987. Difference of 2 points.
• Walks: 614 to 622. Difference of 8 points.
• Home runs: 169 to 151. Difference of 18 points.
• Strikeouts by pitchers: 1051 to 1031. Difference of 20 points.
• Walks allowed: 465 to 469. Difference of 4 points.
• Home runs allowed: 160 to 150. Difference of 10 points.

Using the algorithm described above, 1000 – (24 + 4 + 2 + 8 + 18 + 20 + 4 + 10) = 1000 – 90 = 910.

Thus the 1967 Detroit Tigers are 91% similar to the 2001 Seattle Mariners. What other teams are most similar to the ’01 Mariners?

• 2009 Twins (886); went 87-76
• 2002 Mariners (882); went 93-69
• 2006 Dodgers (867); went 88-74
• 2000 Braves (865); went 95-67
• 2003 Mariners (864); went 93-69
• 1998 Yankees (848); went 114-48
• 1987 Mets (826); went 92-70
• 1994 Astros (826); went 66-49 (a 92/93-win pace)

Those are pretty high similarity scores. It’s interesting to me how there is such a wide range of win-loss records here; from 114 to 87. Just goes to show how the relationship between wins and talent isn’t exact, I guess.

2003 Detroit Tigers

Let’s flip things up a little bit and talk about the team that holds the modern record for most losses in a season. The ’03 Tigers went 43-119. Only the ’62 Mets and the 1899 Cleveland Spiders lost more games in a season.

The ’03 Tigers were ‘led’ by Dmitri Young. He posted an excellent 136 wRC+ (.297/.372/.537, 29 HR) but since he was primarily a DH and was awful in the outfield and at the infield corners, he notched only 2.0 fWAR. The other regulars on the team were execrable, barely above replacement level if they were even above it. Collectively the team hit .240/.300/.375 for a wRC+ of 80.

From a pitching standpoint, Nate Cornejo led the staff with 1.6 fWAR and Jeremy Bonderman followed close behind with 1.2. Their respective 4.70 and 4.89 FIPs didn’t portend much talent. Perhaps the brightest spot on the team was a young Fernando Rodney, who notched 0.5 fWAR in just 29.2 innings. His ERA was a ghastly 6.07 but his 3.50 FIP foretold some staying power.

With all that “talent” you can understand why the team scored just 591 runs and allowed 928. Who else in major league history is similar to this dreck?

No team is very similar, but the 1963 New York Mets (record: 51-111) come the closest with a similarity score of 781. Let’s run the numbers:

• Runs scored: 581 (Detroit) to 501 (New York)
• Runs allowed: 928 to 774
• Strikeouts: 1099 to 1078
• Walks: 443 to 457
• Home runs: 153 to 96
• Strikeouts by pitchers: 764 to 806
• Walks allowed: 557 to 529
• Home runs allowed: 195 to 162

Rounding out the top 10 teams similar to the 2003 Tigers:

• 2001 Pirates (768); went 62-100
• 1964 Senators (751); went 62-100
• 1966 Mets (749); went 66-95
• 2004 Royals (748); went 58-104
• 1962 Cubs (744); went 59-103
• 1969 Padres (735); went 52-110
• 1965 Mets (734); went 50-112
• 1960 Phillies (731); went 59-95
• 1963 Senators (724); went 56-106

Oh, those poor Mets fans in the ’60s. In fact it’s interesting to me how many of these poor teams come in the ’60s. I wonder if Tigers fans in ’03 knew they were watching a brand of baseball that hadn’t been played, with a couple of exceptions, in 40 years.

Now the kicker: how similar were the ’01 Mariners to the ’03 Tigers? As you can already guess, not very. The similarity score between the two teams is a measly 226.

But Wait … There’s More!

Here are a few more interesting teams:

1986 New York Mets:

• 2006 Dodgers (855)
• 1965 Tigers (832)
• 1983 Phillies (832)
• 1967 Tigers (820)
• 2005 Padres (820)
• 1991 Dodgers (818)
• 1972 Mets (816)
• 1989 Expos (812)
• 2002 Giants (807)
• 1972 Astros (806)

2008 Tampa Bay Rays:

• 2009 Rays (904)
• 2011 Rays (868)
• 2009 Rockies (858)
• 2010 Reds (857)
• 2006 Phillies (856)
• 2008 Brewers (838)
• 2011 Rockies (837)
• 2009 Diamondbacks (830)
• 2006 Brewers (822)
• 2011 Reds (810)

1975 Cincinnati Reds:

• 1976 Reds (811)
• 1989 Orioles (797)
• 1979 Reds (769)
• 1957 White Sox (755)
• 1984 Orioles (737)
• 1975 Orioles (725)
• 1958 Red Sox (720)
• 1976 Orioles (718)
• 1981 Angels (717)
• 1974 Brewers (710)

1994 Montreal Expos:

• 2002 Red Sox (920)
• 2011 Cardinals (859)
• 2004 Twins (826)
• 2008 Blue Jays (823)
• 2008 Yankees (810)
• 2010 Twins (810)
• 1987 Astros (799)
• 1999 Red Sox (799)
• 2006 Angels (793)
• 1967 Twins (792)

… the list goes on! You can see for yourself by downloading the set of top 10 scores here.

Future Work

As I said above, I realize there are many more dimensions on which teams can be compared. I just started off with direct measurements of simple fields. In the future I’d like to make the following changes:

• Compare teams on the average age of their roster
• Change SO and BB totals to K% and BB%
• Change HR to OBP and SLG (separately, not OPS)
• Compare IP by starters
• Split out pitching measurements by rotation & bullpen
• Base all measurements relative to the league’s average that year
• Account for park factors

I’m also certain that the “one point per strikeout/walk/HR” logic needs some more rigor applied, also.

We’ll see how far I get on this, but hey, it’s a start 🙂 In the meantime, what comparisons would you like to see made? And — are there any teams you’d like to know the similarity score for?