Notice: Trying to get property 'display_name' of non-object in /var/www/html/wp-content/plugins/wordpress-seo/src/generators/schema/article.php on line 52
keyboard_arrow_uptop

The National League may lay claim to two of the last three World Champions, but little doubt exists in most observers’ minds that the American League has become the superior league in recent years. Not only do they have 13-year undefeated streak in the All-Star Game (including the infamous 2002 tie), but they’ve gotten the upper hand on the senior circuit in interleague play, winning at a .522 clip since it was instituted in 1997. They’ve held an even bigger advantage in each of the past five years, posting a .566 winning percentage over that time span.

Whether that’s due to the designated hitter, the evolutionary pressure of trying to keep up with the Yankees and Red Sox, the abject fate of the Expos/Nationals and the thoroughly mismanaged Pirates, or some other reasons is a conversation I’ll leave for another day. Having confronted this problem in the context of strength of schedule issues and in Baseball Prospectus Hit List power rankings, I’m simply trying to get a feel for the magnitude of the gap.

Here’s the year-by-year breakdown of interleague results, not only the actual ones but also the Pythagenpat projected marks, with the winning percentages as seen from the AL’s vantage:


Year      AL     NL    Win%  AL Runs  NL Runs  Pyth%
1997      97    117    .453     994    1007    .494
1998     114    110    .509    1089    1078    .505
1999     116    135    .462    1246    1374    .452
2000     136    115    .542    1278    1233    .517
2001     132    120    .524    1175    1175    .500
2002     123    129    .488    1127    1097    .513
2003     115    137    .456    1239    1267    .489
2004     127    125    .504    1252    1210    .516
2005     136    116    .540    1230    1056    .571
2006     154     98    .611    1336    1116    .585
2007     137    115    .544    1352    1172    .568
2008     149    103    .591    1249    1014    .596
2009     137    114    .546    1201    1061    .558
Tot     1673   1534    .522   15768   14860    .528
05-09    713    546    .566    6368    5419    .576

That 56-ish percent figure holds up pretty well across a few different samples. The AL has won at a .560 clip in interleague play from 2007-09, a figure that winds up being identical using a typical 5/4/3 weighting system, with the most recent result weighted the most heavily. But winning at a .560 clip doesn’t mean that we can simply say that the AL is a .560 league and the NL a .440 circuit, with a 120-point gap between the two leagues.

To figure out what the strength of the two “teams” are that could produce a result where one won at a .560 clip, we turn to what Bill James called the Log5 method, one I’ve referenced in my articles on schedule strength, and one that Clay Davenport uses literally millions of times a day to generate the daily Playoff Odds reports at Baseball Prospectus. The formula boils down to Win% = .500 + A – B, where Win% is the observed outcome percentage (.560) and A and B are the two teams. Since we also know that in this case, the winning percentages are complementary (A + B = 1.000), it’s simple algebra to determine that a .530 team playing a .470 team would produce that observed .560 winning percentage.

When I did the strength-of-schedule pieces, I used 2006-2008 results, which produced an AL winning percentage of .582, to apply a 40-point bonus to the AL team in each head-to-head matchup and a 40-point tax to the NL one. (I stuck with round numbers for the purposes of demonstration.) Now that the 2009 results are almost entirely in the books, that estimate winds up on the high side, as the 2006 results were the most lopsided to date.

One revealing aspect about the AL’s advantage over the NL is that even the lousier junior circuit teams are beating the senior circuit’s weak sisters consistently. Sticking with the last five years of data (including this unfinished season) and splitting each league into upper and lower halves in terms of interleague records-the 35 best (or worst) team-seasons in each half in the AL, 40 in the NL-we find that the AL’s better half, which won at a .561 clip in those intraleague games, boosted their winning percentage to .610 in interleague games. The lower half, which produced a measly .438 winning percentage in intraleague play, kicked NL tail at a .523 clip. The NL’s better half posted a .551 winning percentage in intraleague play but just a .447 mark in interleague play, while the lower half dipped from .450 to .421.

This tendency persists if we break the teams into smaller groups. Here it is in quintiles:


Group   Intra   Inter
AL1     .594    .595
AL2     .550    .627
AL3     .504    .587
AL4     .458    .466
AL5     .392    .556
NL1     .580    .439
NL2     .538    .449
NL3     .503    .461
NL4     .460    .408
NL5     .418    .414

Granted, we’re not talking about huge sample sizes here (14 seasons apiece in the AL groups, 16 in the NL groups), but… wow. Every NL grouping, from the best 20 percent to the worst, won significantly less than 50 percent of its games against the AL. The top three AL groupings dominated interleague play, and while the fourth AL group won less than half its games, the bottom grouping won at a robust .556 clip, thanks to a couple recent Orioles teams going 11-7, a couple of Royals teams posted winning records (including 13-5 in 2008), and just two of the 14 teams in the group finishing below .500 in interleague play.

Via linear regression (and the much-appreciated assistance of my colleague, Eric Seidman), we can express the predicted relationship between intraleague and interleague winning percentage over this five-year period as follows:

AL_Inter = 0.353 + 0.427*AL_Intra
NL_Inter = 0.324 + 0.221*NL_Intra

So, an AL team that goes .450 in intraleague play would be expected to post a .545 winning percentage in interleague play; for a .500 team, it’s .567, and for a .550 team, it’s .588. For a .450 intraleague team in the NL, the expectation is for a .423 winning percentage in interleague games; for a .500 team, it’s .434, for a .550 team, it’s .446. By this reckoning, the Dodgers (.605 in intraleague) would produce a .458 winning percentage against AL competition, about what the Blue Jays (.464 in intraleague) have done, and the Jays would produce a .551 winning percentage in the NL, good enough to contend for the Wild Card.

That seems fairly extreme, and it’s worth remembering that the individual interleague samples are fairly small. Still, it’s quite clear that the AL’s advantage is a very real one which cuts even the best NL teams down to size. The four 2008 NL playoff teams went a combined 22-38 in interleague play last year, including 4-11 by the eventual World Champion Phillies, and this year’s three division leaders were a combined 24-27. That doesn’t mean the senior circuit can’t win the World Series, but via this way of looking at things the odds aren’t stacked in its favor.

A version of this story originally appeared on ESPN Insider Insider.

Thank you for reading

This is a free article. If you enjoyed it, consider subscribing to Baseball Prospectus. Subscriptions support ongoing public baseball research and analysis in an increasingly proprietary environment.

Subscribe now
You need to be logged in to comment. Login or Subscribe
sockeye
9/02
What's the interleague record for games played without the DH?

I'd be interested in this as supporting/contradictory evidence for the claim that the DH skews interleague records in the AL's favor?
jjaffe
9/03
In AL parks, the AL has a .573 winning percentage overall, .619 from 2005-2009, and .587 from 2007-2009.

In NL parks, the AL has just a .470 winning percentage overall, but it's .514 from 2005-2009, and .533 from 2007-2009. FWIW, those latter two numbers are much lower than the AL's Pythagenpat in those stretches (.548 and .566, respectively).

Either way, the AL has outperformed expectation given the home field advantage and put a serious dent in the notion that the DH rule is the primary reason for its advantage.
oneofthem
9/02
It would be interesting to look at a comparison of league strength in terms of individual player performance. To what extent does nl or al inflate individual stats. but I suppose the data available for such a comparison has too many noisy factors
jjaffe
9/02
That's a suggestion better directed Clay Davenport or anyone else who usually deals in handling such large batches of data.
sam19041
9/03
Jay, that was very interesting. Two follow-up questions:

1) Why is the difference so great between leagues?

2) How do revenues / media markets (i.e. DOLLARS) correlate to winning intraleague vs interleague? Are NL teams at a big disadvantage to AL teams? Clearly you have the Cubs, Mets, Dodgers, but after that?
jjaffe
9/03
1) is worth a whole article to answer. Offhand, I'm not sure why the advantage is so persistent.

2) An article Nate Silver did a few years ago ranked the market size of the 30 teams (http://www.baseballprospectus.com/article.php?articleid=6225). Taken among the 30 teams as a whole in a quick-and-dirty test, there's a positive correlation between intraleague winning percentage and market size (r = 0.6), but almost none between interleague winning percentage and market size (r=-.05). Similar correlations if we use attendance instead of market size.

Breaking it by leagues, the AL intraleague winning percentage corerlates well with market size and attendance (r ~ .7). The NL intraleague correlations for those are much lower (r ~ .45) because of the success of smaller market teams like St. Louis, Milwaukee and San Diego. Neither league produces correlations above .1 for interleague winning percentage and market size or attendance.
sam19041
9/03
Thanks very much, Jay. I appreciate the quick (yet thoughtful) response. I have to think there is some logical explanation for the discrepancy in performance between NL and AL (e.g. AL teams have higher payrolls, revenue, etc.). It just doesn't seem to be a level playing field right now.
sam19041
9/03
P.S. I hope we will see a full article on this. It's a very interesting topic.
Mountainhawk
9/03
I'm virtually certain that the intrinsic roster advantage of having a DH on the roster is the main reason it's so persistent. The difference between the average AL DH and the average NL first guy on the bench has to be massive compared to the difference between the average AL and NL pitcher in terms of offensive performance.
DrDave
9/03
This year, NL DH's are hitting .277/.357/.457 in interleague games, while AL DH's are hitting .254/.338/.448 overall, mostly against AL pitching. In other words, NL DH's hit AL pitching better that AL DH's do. That doesn't look like the answer to me.
Mountainhawk
9/03
You can't use NL DH statistics. Ryan Howard and Raul Ibanez were the Phillies DH for most games this year as I recall, but those aren't the bats added to the lineup. Instead, you have to look at how Greg Dobbs or Mayberry or Bruntlett performed at the regular's position.
jjaffe
9/03
Both of you make solid points, though I think it's worth noting that AL DHs aren't exactly towering over the rest of the league; AL non-DHs have hit .267/.334/.428, so we're talking 24 points of OPS in difference.

Meanwhile, it's not at all clear that the NL DHs are uniformly regulars getting a breather. Of the 21 guys with at least 10 PA in that role (PA = AB + BB in this quick and dirty instance), I count 11 of them as regulars and 10 as part-timers. The regulars hit .275/~.344/.509 as DHs, the part-timers .301/~.377/.468, which tells me that coming up with a decent extra bat couldn't have been all that daunting of a challenge for most teams. (data gathered from http://espn.go.com/mlb/stats/batting/_/split/86/league/nl/sort/atBats). Howard had all of 12 "PA" in that role, Ibañez nine; they hit .250/.286/.550 in those appearances.
BrewersTT
9/03
Strictly speaking, straight linear regression isn't appropriate for estimating percentages, because the errors can't be distributed normally (they can't go outside of +/- 1). But given the consistency of what you've observed, I'm not sure the overall sense of it would change. You would get different estimates from logistic regression, though. (end lecture)
nghunter
9/03
Just a nifty little article. Pleasure to read. Thanks, Jay.