You Are Here > Baseball-Reference.com > Blog >

SITE NEWS: We are moving all of our site and company news into a single blog for Sports-Reference.com. We'll tag all B-R content, so you can quickly and easily find the content you want.

Also, our existing B-R blog rss feed will be redirected to the new site's feed.

Baseball-Reference.com ยป Sports Reference

For more from Andy and the gang, check out their new site High Heat Stats.

Great 1-2-3 Rotation Combos In a Best-of-7 Series

Posted by Neil Paine on October 13, 2010

Last week I posted about how often the team with the best Game 1 starter by WAR wins the Division Series (only 55%, as it turns out), and with the League Championship Series coming up on us quickly, I wanted to do a variation of that study for best-of-seven series.

In the wake of the Phillies' sweep of Cincinnati, many are saying their top 3 pitchers (Roy Halladay, Roy Oswalt, Cole Hamels) are all they really need to overwhelm an opponent, even in a 7-game series. How does that triumvirate stack up to other Top 3s since the strike? Combined, Halladay (6.9 WAR), Oswalt (5.1), & Hamels (4.7) posted 16.7 Wins Above Replacement during the 2010 season, a total that ranks 6th among LCS participants since 1995:

1. 2001 Arizona Diamondbacks (18.5 WAR):

2. 2005 Houston Astros (18.3 WAR):

3. 1996 Atlanta Braves (18.0 WAR):

4. 1995 Atlanta Braves (17.9 WAR in 144 games):

  • Maddux (8.8)
  • Glavine (4.7)
  • Smoltz (4.4)

5. 1997 Atlanta Braves (16.8 WAR)

  • Maddux (7.3)
  • Glavine (5.0)
  • Smoltz (4.5)

6. 2010 Philadelphia Phillies (16.7 WAR)

  • Halladay (6.9)
  • Oswalt (5.1)
  • Hamels (4.7)

7. 2003 Chicago Cubs (16.1 WAR)

8. 2007 Cleveland Indians (15.8 WAR)

9. 1998 San Diego Padres (15.7 WAR)

10. 2001 New York Yankees (15.2 WAR)

Now, those are some impressive rotations, but does having the better top 3 necessarily mean you're a lock to win an LCS or World Series matchup? Based on the past 15 years of data, not quite...

Year Round Winner Top3WAR Wins Loser Top3WAR Losses
1995 ALCS CLE 10.8 4 SEA 11.7 2
1995 NLCS ATL 17.9 4 CIN 11.4 0
1995 WS ATL 17.9 4 CLE 10.8 2
1996 ALCS NYY 12.2 4 BAL 9.7 1
1996 NLCS ATL 18.0 4 STL 6.9 3
1996 WS NYY 12.2 4 ATL 18.0 2
1997 ALCS CLE 7.1 4 BAL 13.1 2
1997 NLCS FLA 11.6 4 ATL 16.8 2
1997 WS FLA 9.1 4 CLE 7.1 3
1998 ALCS NYY 11.3 4 CLE 8.5 2
1998 NLCS SDP 15.7 4 ATL 14.9 2
1998 WS NYY 11.3 4 SDP 15.7 0
1999 ALCS NYY 11.4 4 BOS 12.5 1
1999 NLCS ATL 13.0 4 NYM 7.4 2
1999 WS NYY 11.4 4 ATL 13.0 0
2000 ALCS NYY 11.0 4 SEA 6.5 2
2000 NLCS NYM 11.7 4 STL 7.7 1
2000 WS NYY 11.0 4 NYM 11.7 1
2001 ALCS NYY 15.2 4 SEA 9.4 1
2001 NLCS ARI 18.5 4 ATL 12.0 1
2001 WS ARI 18.5 4 NYY 15.2 3
2002 ALCS ANA 8.4 4 MIN 3.7 1
2002 NLCS SFG 7.7 4 STL 6.7 1
2002 WS ANA 8.4 4 SFG 7.7 3
2003 ALCS NYY 14.9 4 BOS 13.3 3
2003 NLCS FLA 10.5 4 CHC 16.1 3
2003 WS FLA 9.3 4 NYY 14.9 2
2004 ALCS BOS 13.6 4 NYY 7.5 3
2004 NLCS STL 3.9 4 HOU 9.7 3
2004 WS BOS 12.0 4 STL 3.9 0
2005 ALCS CHW 14.4 4 LAA 12.1 1
2005 NLCS HOU 18.3 4 STL 7.8 2
2005 WS CHW 14.4 4 HOU 18.3 0
2006 ALCS DET 10.3 4 OAK 9.2 0
2006 NLCS STL 6.4 4 NYM 2.8 3
2006 WS STL 6.4 4 DET 10.3 1
2007 ALCS BOS 10.9 4 CLE 15.8 3
2007 NLCS COL 3.2 4 ARI 10.1 0
2007 WS BOS 10.9 4 COL 3.5 0
2008 ALCS TBR 10.3 4 BOS 14.0 3
2008 NLCS PHI 8.3 4 LAD 9.8 1
2008 WS PHI 8.3 4 TBR 10.3 1
2009 ALCS NYY 8.9 4 LAA 9.0 2
2009 NLCS PHI 9.6 4 LAD 8.5 1
2009 WS NYY 8.9 4 PHI 9.6 2

Teams who had the better combined Top 3 starters only won the best-of-7 series 55.6% of the time, and even teams with a 4+ WAR advantage only won 59% of their series. It isn't until you get to instances of a 6+ WAR advantage (80% of series won) that you see the team with the huge pitching edge win as frequently as you may expect. In fact, setting up a simple logistic regression between top 3 starter WAR advantage and series winning %, we get the following equation:

Series WPct ~ 1 / (1 + exp(-0.08 * WAR advantage))

All else being equal, this regression implies the following expected series winning %s for each level of WAR advantage:

WARadv WPct
+8 65%
+6 62%
+4 58%
+2 54%
0 50%
-2 46%
-4 42%
-6 38%
-8 35%

Going against the Giants in the NLCS, the Phillies' top 3 starters will likely have an advantage of +5.9 WAR over San Francisco's 10.8-WAR trio of Matt Cain (3.9 WAR), Tim Lincecum (3.5), & Jonathan Sanchez (3.4). Based on the regression results, that dominant rotation edge gives them a 62% chance of advancing to the World Series, less than you might expect from such a big starting pitching advantage in Games 1-3. Three ace pitchers don't hurt, of course, but in the end there will be many other factors beyond Halladay-Oswalt-Hamels that go into determining the winner of the series.

This entry was posted on Wednesday, October 13th, 2010 at 7:51 am and is filed under Postseason, Sabermetrics, WAR. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.

25 Responses to “Great 1-2-3 Rotation Combos In a Best-of-7 Series”

  1. It would be interesting to see if the team with the better top 3 fared better if we went by Fangraphs WAR instead.

    Just to cherry pick, in 07, the Red Sox top 3 are 5 WAR behind the Indians top 3 according to bWAR. According to fWAR, they are roughly equal (13.3 to 13.6). The Red Sox won the series in 7 games.

  2. Thanks for this post. Very interesting.

    I think the WAR for the Giants may be a bit on the low side, as its skewed by Lincecum's late-season struggles, which appear to be resolved. I certainly view him as dangerous a pitcher as any of the Phillies top three, even if his WAR is a few points lower.

    I'd also be curious as to whether the third starter even matters that much here when you've got two REALLY top-notch guys. I mean, would you rather have WAR figures of 8.5, 8.5 and 0, or 6.0, 5.5, 5.5? I think I'd take my chances with the first rotation.

  3. Someone I know a few years back did some work that was essentially a shortcut to this, suggesting that the team with the best number 3 starter was usually at a great advantage. The rationale was the obvious one--once you get to the CS or WS, both teams are likely to have a superior ace and an awfully strong number two, but all too often get as far as they do on the backs of those two stars. Of course, nothing is infallible, but it also seemed to work a great deal of the time. I dug through my archived mailboxes to find the analysis, but came up empty. So, on to Hamels v. Sanchez!

    The other series is a bit more of a head-scratcher in this regard, particularly if C. J. Wilson brings his A game for the Rangers and A. J. Burnett shows up for the Yankees (perhaps it comes down to the best pitcher with a middle name that starts with J?). Ultimately, I'm not sure Phil Hughes is better than Colby Lewis or Tommy Hunter or David Clyde or whoever the Rangers decide to throw, but he's been death to Texas over his short career.

  4. While Andy makes an excellent point about having the better pitching WAR not necessarily giving a huge advantage, it is notable, I think, that of the top nine teams (excepting the current Phils), six went to the World Series (with three of them winning it all-- although that might not be fair, since the 2001 Yankees had to play the 2001 Diamondbacks, so ONE of them had to lose). Of the remaining three, two lost their LCS in 7 games. Only the 1997 Braves "underachieved," losing the NLCS to the Marlins in 6. At least anecdotally, then, we can say that Philly's rotation LOOKS like a World Series winner-- but that's hardly new information, now is it?

  5. Great questions in #2 & #3 regarding the distribution of WAR among your top 3... In some cases the #3 is kind of tacked on behind 2 mega-aces (Batista, Hitchcock, Westbrook, etc), and in some cases he's not that much worse than the #1 (Glavine '96). I'd have to look at which configuration is better if you're going to have a great aggregate WAR from your top 3 guys -- two historically-great starters + just another guy, or three really good starters?

  6. Neil, very interesting and timely stuff.

    I'm wondering, though -- are 45 series really enough to draw meaningful conclusions from this study, to the point of expected series W% for 9 different grades of WAR differential? I don't know enough about probability to gauge this, but it just feels a little sketchy.

  7. I admit the sample size is pretty small. I would like to re-run with all the playoff series data we have, but I don't have that on hand right now.

  8. @5: ...and in some cases it's the #3 guy who picks up the slack when your supposed aces fall apart.
    2007 ALCS
    CC Sabathia: 2 GS/10.1 IP/12 ER
    Fausto Carmona: 2 GS/6 IP/11 ER
    Jake Westbrook: 2 GS/12.2 IP/5 ER

  9. Those 2003 Cubbies were 22, 26 & 22 years old respectively. What could have been...

  10. I'd consider a 62% chance pretty darn big, actually. Baseball isn't like football. The best teams don't win 80+% of their games.

  11. There's also the fourth guy in the rotation, since it is rare these days for a team use only three starters in a seven-game series.
    For example, the Diamondbacks started Brian Anderson, who a had a negative WAR, in the 2001 World Series. The Yankees' fourth starter was Orlando Hernandez with 1.4 WAR.

  12. In his New Historical Baseball Abstract, Bill James always took to using the form (#1*1)+(#2*2)+(#3*3) when figuring which was best in these sorts of things, thereby weighting the weakest link most heavily. Here are the sums I come up with when doing that with these rotations:

    1. 1996 Braves - 35.3
    2. 2005 Astros - 34.7
    t3. 1995 Braves - 31.4
    t3. 2001 Diamondbacks - 31.4
    5. 2010 Phillies - 31.2
    6. 1997 Braves - 30.8
    7. 2003 Cubs - 30.6
    t8. 2007 Indians - 27.2
    t8. 2001 Yankees - 27.2
    10. 1998 Padres - 25.3

    This probably doesn't mean anything, but I think it's interesting. I also think it's a little more pleasing aesthetically-- I mean, as great as Johnson and Schilling were in 2001, Miguel Batista does not deserve to be mentioned as part of the best postseason rotation of all-time.

  13. We also must realize that WAR is not necessarily predictive. WAR is great at telling us what DID happen, but not necessarily will happen. Regular season WAR might include early season struggles that have since been worked through or periods where a player was injured but still played. If there are legitimate reasons to believe that a player's regular season WAR is not necessarily indicative of his quality of play in the postseason, it is natural that there will be a limited correlation to team success.

  14. Although it's pre Strike 94/95, The '73 Mets had a combined 19.8 Top 3:
    Seaver-9.5 WAR
    Koosman-5.7 WAR
    Matlack-4.6 WAR

    That might explain how that team beat the Big RED Machine. Also, they were up 3-2 to the Oakland A's before Yogi Berra made the dumb move of pitching Tom Seaver on short rest in game 6. People still wonder why Berra didn't pitch well rested George Stone with his 2.80era/130 ERA+ in game 6 and Save Seaver for a possible game 7.

    Another odd twist is that both Koosman & Matlack had LOOSING records, Koos-14-15, Matlack-14-16, Another great example of the relative pointlessness of W/L record.

  15. #4, I did not write this!

  16. Defense could be a little bit of a confounding factor for something like this.

    While it makes sense to try and factor out defense when directly comparing pitchers, for the purpose of this type of study, it would make sense to look at how a rotation does including the defense played behind it.

  17. I do see a glaring flaw with this, and that's that while Jonathan Sanchez is a good pitcher, he's by no means the same pitcher as Tim Lincecum, as WAR would suggest.

    "The Freak" is much more on par with Roy Halladay (or at the very least Cole Hamels) than he is with Sanchez.

  18. Johnny Twisto Says:

    What's the glaring flaw? You hardly need WAR to show you that this season Sanchez was about as good as Lincecum.

  19. @Johnny: The flaw is that Lincecum's "true" level of talent (ie, what he could bring to any given game) isn't well represented by one season's worth or WAR data, and even moreso, his WAR was dragged down by two bad months this season. In fact, in the last three seasons, he's only had three months with an over 4 ERA.

    So while Sanchez might have been about as good as Lincecum this season on the whole, Lincecum is still the better bet to pitch one great game when you look at his whole body of work.

  20. JayT and JT-

    That is the point I was getting at in my post. For the season, Sanchez and Lincecum brought similar value. But looking at what they are likely to contribute at this point, differences emerge. Obviously, we can go too far with this and say, "Well, hey, I know Pedro Martinez didn't pitch this year, but I mean, you want to compare Phil Hughes to him???" But, we might want to find a way to get a better sense of where pitchers are at this point than just looking at them over the previous 6 months. I don't know that going back historically is necessarily the best bet, unless there is a specific reason to believe that this regular season was a major abberation but that the post-season won't be. Rather, perhaps we use some sort of weighted WAR, putting more stock into more recent performances? I don't necessarily know the best way to do this, but I'm sure there are both simple and complex ways that might correct for issues such as this.

    Of course, there is also the possibility that pitchers with lesser reputations are just as good, if not better, than those with greater reputations. So we shouldn't ALWAYS assume that regular season WAR tells us nothing about the postseason. We just need to look at each case individually and determine whether a guy's seasonal production is the best tool to predict his post-season performance.

  21. I think you can generalize regular season performance as indicative of post-season performance. Individual circumstances may change the individual odds, but I think that the overrate/underrate situations will even out over the long run and the generalization works for the most part.

    Predictions of this kind can't be exact, and they should be modified on a case-by-case basis. But to do that you need a baseline, and the WAR combo is a perfectly reasonable one to start with.

  22. DoubleDiamond Says:

    1997. I was trying to remember who the Braves faced in that year's World Series. And then I remembered playing around with a potential newspaper headline summarizing the NLCS:

    Atlanta gave them all their aces, but all they heard was "Go Fish!"

  23. Alex Remington Says:

    Good pitching beats good hitting. And vice versa.

  24. Mike Felber Says:

    I agree Zachary. Though the question of whether the #3 pitcher is disproportionately important, or taking the top 2 is more predictive remains. regardless, the conventional wisdom tends to way more towards good pitching being decisive Alex. I do not see why that should be true, EXCEPT in a short series when you can pitch your best guys more often than the regular season-you cannot do so with position players.

    To prove this true, we could take teams with nearly = overall strength, but reversed in offense & defensive value. Or just about = overall, but 1 team markedly better in pitching. The latter team should win somewhat more often.

  25. your blog is very good!!! registry first aid keythanks i love your !!!