Baseball for the Thinking Fan

Login | Register | Feedback

You are here > Home > Baseball Newsstand > Baseball Primer Newsblog > Discussion
Baseball Primer Newsblog
— The Best News Links from the Baseball Newsstand

Thursday, May 01, 2014

BtBS: How much does a team’s early-season record inform their playoff odds?

This is the time of year when announcers, talk show hosts and web columnists start asking “How many games does a team need to play before we know what kind of team they really are?” So far this year, I’ve heard anything from 30 all the way to 60. I almost choked when I heard 60—at that point the season is almost 40% played, and absent drastic and unpredictable events (trades, injuries, phenoms like Yasiel Puig who totally outperform any reasonable expectation) the team has moved beyond the prediction phase to that of actual. [...]

Of the teams [since 1995, when the divisional playoff level was added] playing at less than .400 at the 60-game mark (a 23-37 record or worse), none made the playoffs. Conversely, almost 72 percent of teams with a record of 36-24 or better eventually made the playoffs. The demarcation point seems pretty clear—at the 60-game mark, teams need to be playing around .550 ball (33-27) in order to have reasonable chances of making the playoffs.

I don’t find these numbers surprising, but I’m more curious how many fewer games can be played before reasonable inferences can be made.

bobm Posted: May 01, 2014 at 12:35 AM | 18 comment(s) Login to Bookmark
  Tags: playoffs, sample size

Reader Comments and Retorts

Go to end of page

Statements posted here are those of our readers and do not represent the BaseballThinkFactory. Names are provided by the poster and are not verified. We ask that posters follow our submission policy. Please report any inappropriate comments.

   1. Pat Rapper's Delight (as quoted on MLB Network) Posted: May 01, 2014 at 10:04 AM (#4698130)
I’m more curious how many fewer games can be played before reasonable inferences can be made.

I'm reading a copy of BP's Extra Innings that I recently got at the local used book chain, and they gave three different answers to this question.

After 17 games, a team's record is more predictive of their final record than simply assuming they will finish .500.

Also after 17 games, a team's in-season record predicts the final record with the same accuracy as their pre-season expectation.

After 48 games, "a team's pre-season expectation and the portion of its in-season record that the preseason expectation couldn't predict held equal predictive value over the team's final record."

In short, anywhere from around 3-6 weeks, depending on what question you're really trying to answer.
   2. PASTE, Now with Extra Pitch and Extra Stamina Posted: May 01, 2014 at 10:21 AM (#4698140)
I know I'm being a no-fun dick by asking, but... isn't this article written right around this time every year?
   3. dr. scott Posted: May 01, 2014 at 03:25 PM (#4698356)
People have short memories... and journalism doesn't pay much. Recycling is the new market efficiency.

Well, not new...
   4. Moeball Posted: May 01, 2014 at 03:32 PM (#4698360)
I know I'm being a no-fun dick by asking, but... isn't this article written right around this time every year?

I'm even less fun. Because of the way the divisional system plays out, we get years where just about all the teams in one division may have winning records but some don't get to see postseason play (AL East in some years comes to mind), whereas in other divisions you can have a team with a crappy record still make the playoffs if they win the crappy division (2005 Padres? And it's a good thing for the AL West that 1994 had no postseason because some awful team in that division was going to win it with a record of double digits under .500).

Therefore, looking at how a team is doing after 17 games or 30 games or whatever is even less meaningful. Oh, they're 6-11 after 17 games? Doesn't matter, they're in a lousy division, they could easily still make the postseason. Under the current format, mediocrity is encouraged and rewarded, so mediocrity is what you get.
   5. Sunday silence Posted: May 01, 2014 at 03:34 PM (#4698363)
a really interesting question would be: is pythagorean runs differential any better than w/l record at predicting the final season record?

is there a point in time when one is better than the other? Id really like to see the answer.
   6. Crispix Attacksel Rios Posted: May 01, 2014 at 03:42 PM (#4698373)
What exactly does the word "inform" mean here? "Influence"?
   7. geonose Posted: May 01, 2014 at 06:20 PM (#4698479)
What exactly does the word "inform" mean here? "Influence"?

That's what I was wondering, so I looked it up:

To be a pervasive presence in
To impart some essential or formative characteristic to

I suppose "influence" is a decent synonym. Still, it's an odd use of the word, one that's decidedly out of style.
   8. Walt Davis Posted: May 01, 2014 at 07:47 PM (#4698504)
I think I like "inform" here. "Influence" is causal ... and in a sense going 6-12 has a causal effect on your final record. But here he is talking about what "information" that early record gives you. The team is no better or worse having gone 6-12* so it's a question of does it change your mind about your assessment of team quality. Granted "what does early season performance tell you about a team's playoff odds" might be simpler.

I focused more on the question of what people mean when they ask "what kind of team are they?" Playoff odds are not necessarily the same thing. We might have come into the season wondering if the Cubs could be a near-500 team but everybody could have told you in March 2014 (or even March 2012) that the 2014 Cubs weren't a playoff team.

But what kind of team are the Cubs? Well, there were several questions coming in. Would Castillo continue to be an average or better C? Would Rizzo and Castro bounce back? Does Mike Olt have anything to offer? Does the OF stink as bad as it looks on paper or, if we squint, we can almost see a non-embarrassing OF put together out of platoons and defense. Will Hammell stink or be this year's Scott Feldman? Will Samardzija perform well? Will he get traded or re-sign? What about Travis Wood? Will the pen be non-putrid?

So far some of those questions have been answered favorably, some have not. While the current record might be predictive of later record, it doesn't tell us anything about the answers to those questions and so not anything about the "kind of team the Cubs are." Overall performance might well stay the same but some of the questions with favorable answers will turn unfavorable or neutral and vice versa.

I don't like this focus on playoff odds. Focus on playoff odds justifies crap like the Astros and Cubs. Focus on playoff odds diminishes the regular season. Focus on early season effects on playoff odds encourages fans to give up on their teams in May and even tells the fans of teams performing well that "it's over" and they can tune out until Sept.

If you're itching to check the latest Vegas odds on teams making the playoffs, this is useful information. Otherwise, sit back and enjoy how the season unfolds.

#5, that's been done although I don't remember the answer. But you're not quite asking the question correctly -- pythag win% can only predict future performance not change past performance (obviously). Last I saw the Angels were 500 with something like a +30 run differential. At best that +30 tells you they'll play something like 600 ball from here on out but, even if they did, final season win % will be under 600.

So along those lines, does the Cubs 9-17 record mean that they are pathetic or does their -.2 SRS (-9 run differential, 12-14 pythag, whatever you want to use) mean they're OK? By SRS, there are 3 teams in the NL West plus Philly who are worse and the first-place Giants are nearly as bad. According to SRS, this lousy Cub team is better than the $200 M Dodgers which makes you wonder if SRS has ever seen the Cubs play.

My guesstimate on this issue is basically this. After 30 games, a team has about 1200 PA so their offensive performance is probably about as projectable as two seasons would be for an individual batter with possibly more variance due to roster/playing time decisions (i.e. compositional change). Pitchers will have about 270 IP (or 1200 BF) which is about 1.5 starter-seasons. We could probably carry over other lessons from micro-analysis like K-rates stabilize fairly quickly, again remaining mindful of possible compositional change.

Alternatively, from a simple weighting perspective, after 81 games you'd guesstimate the final win % would be the average of their actual win % to date and their projected win % (preseason projection, pythag, SRS, whatever). Unfortunately rosters are very unstable in the 2nd half. And all of the variance in that estimate of final record is coming from the uncertainty about future performance.

*Obviously starting 6-12 can have all sorts of longer-term consequences for the final season record. First, there may have been injuries. Second, being 55-55 at the end of July is a lot different than 52-58 in terms of deadline trades, cups of coffee, etc. Third, the 6-12 start might lead to panic moves or at least different playing time decisions.
   9. Joyful Calculus Instructor Posted: May 01, 2014 at 08:38 PM (#4698526)
That's actually pretty surprising. I guess it's my experience as a Dodger fan that is biasing how I think about such things. Dodger playoff teams in my lifetime have typically had losing records deep into the season. Of the five playoff teams I've followed ('04,'06,'08,'09, 13), only the 2004 and 2009 teams started the season strong. Last date of losing records:
2004- April 5
2006- August 4 (47-55 on July 26!)
2008- September 2!
2009- April 10
2013- July 9 (but 30-42 on June 21!)
   10. Joyful Calculus Instructor Posted: May 01, 2014 at 08:51 PM (#4698536)
Also, fivethirtyeight had an article about this recently:

Using an unbelievably useful methodology from arch-sabermetrician Tom Tango, I calculated the number of games necessary in each sport to regress a team’s record halfway to the mean — meaning, we’d know half of its observed outcomes were due to its own talent (while the other half results from randomness). For pro basketball and football, the numbers are similar: In the NBA, it takes about 12 games; in the NFL, 11 games. But in baseball, it takes a whopping 67 games for half of the variance in observed winning percentages to come from the distribution of talent and half from randomness.
   11. OsunaSakata Posted: May 01, 2014 at 11:55 PM (#4698634)
The comments in the 538 article suggest that starting pitchers and playing series against the same opponent are why baseball conveys less information than the other sports in a given number of games.
   12. SoSHially Unacceptable Posted: May 02, 2014 at 12:38 AM (#4698645)
The use of pitchers plays a role, but it simply takes longer for true talent to identify itself in baseball than it does in basketball and football (hockey is closer to baseball). All sports differ in this regard. Some sports are very predictable even at the individual game or event level. Others not so much. If you had a tennis or golf tournament with equal numbers of professionals, you would have far better luck forecasting the outcome of the tennis tournament than the golf tournament.

In baseball, a college team could beat an MLB team in a single game if things broke right (even an MLB team starting its best pitcher). A college team would not beat an NBA team. The sports are just different in this regard.

My suspicion would be the sports that are closer to simple tests of athleticism (such as sprinting) would be quicker to identify talent than the sports that involve fine-tuned and/or multiple skills (such as baseball).

   13. DKDC Posted: May 02, 2014 at 01:17 AM (#4698655)
Or baseball is just more random or has fewer discrete events in a game than other sports.
   14. SoSHially Unacceptable Posted: May 02, 2014 at 01:30 AM (#4698666)

Or baseball is just more random or has fewer discrete events in a game than other sports.

I believe the first part (though I think the randomness is a byproduct of the complexity), but I don't see any reason to think there are fewer discrete events in it.

   15. Joyful Calculus Instructor Posted: May 02, 2014 at 02:02 AM (#4698677)
There are about 80 PAs in a baseball game. Basketball has about 200 possessions per game. The NFL has about 140 plays per game.
   16. SoSHially Unacceptable Posted: May 02, 2014 at 02:12 AM (#4698679)
There are about 80 PAs in a baseball game. Basketball has about 200 possessions per game. The NFL has about 140 plays per game.

There are also about 250 pitches (in a sense, individual plays) in a baseball game. I don't think any of these numbers are all that telling.

   17. Russ Posted: May 02, 2014 at 09:32 AM (#4698735)
I think it's less about the number of pitches/plays/possessions and more about the influence that individual players can have over the season. Teams in the NL averaged 37.91 plate appearances per game. Joey Votto lead the NL in PA with 726, so he was getting only 4.5 of those 38 PA (or 12%). Even a guy like Andrew McCutchen was down around 4.2 PA per game (11%). And those are the very best players in the league last year who played a ton. Obviously for even other great players, their influence is mitigated further by injuries, less than full playing time, etc.

If you use Player Impact Estimate (PIE), you see that the top players in the NBA like LeBron James or Kevin Durant are up around 20%. Wainwright led the league IP last year and he threw 241 innings or an average of only 1.48 per 162. That means that he's pitching 16.4% of the 9 inning average per game (really 8.99). Now, a starting pitcher of course has more influence over a particular game, but where you see more variability is that technically Wainwright could only affect 34 out of the 162 games... so 80% of the games he has zero impact on.

Each batter and reliever has pretty low leverage in a particular game and each starter has zero impact on 80% of the games, which means that the result is due more to the sum of small contributions from everyone, which is obviously going to lead to a lot more variability in results.
   18. bookbook Posted: May 02, 2014 at 12:41 PM (#4698887)
#10, If you look at this in terms of percentage of the sport season, it looks a bit different.

In NBA basketball, 15% of the season will tell, in MLB baseball 41%, but in the NFL it takes a whopping 68% of the season to tell a good team from a bad one!

But, man, another reason to hate home-and-home series in the baseball schedule--like we needed another one beyond the sheer boredom, and undue influence of streaks/injuries on in-division outcomes.

You must be Registered and Logged In to post comments.



<< Back to main


All News | Prime News

Old-School Newsstand

BBTF Partner

Dynasty League Baseball

Support BBTF


Thanks to
for his generous support.


You must be logged in to view your Bookmarks.

Hot Topics

NewsblogOT - 2017-18 NBA thread (All-Star Weekend to End of Time edition)
(5338 - 1:44am, May 28)
Last: maccoach57

NewsblogWeekend OMNICHATTER for May 26-27, 2018
(142 - 1:38am, May 28)
Last: Bote Man

NewsblogTaking Back the Ballparks - Orioles voting thread
(18 - 1:02am, May 28)
Last: Jarrod HypnerotomachiaPoliphili (TeddyF.Ballgame)

NewsblogOT - Catch-All Pop Culture Extravaganza (April - June 2018)
(2497 - 12:19am, May 28)
Last: PreservedFish

Gonfalon CubsPadding
(55 - 11:46pm, May 27)
Last: Moses Taylor, aka Hambone Fakenameington

NewsblogOTP 2018 May 21: President takes British royalty to a baseball game, May 15, 1991
(1314 - 11:39pm, May 27)
Last: Lassus

NewsblogOT: Winter Soccer Thread
(2130 - 10:13pm, May 27)
Last: J. Sosa

Sox TherapySo That Happened
(4 - 9:29pm, May 27)
Last: The Anthony Kennedy of BBTF (Scott)

NewsblogPosnanski: Modern Romonce
(116 - 8:54pm, May 27)
Last: PreservedFish

NewsblogBautista must be last over-the-hill vet the Mets take a chance on
(53 - 8:21pm, May 27)
Last: Howie Menckel

NewsblogZack Cozart is concerned the Rays 'have an ulterior motive' in starting Sergio Romo
(13 - 8:01pm, May 27)
Last: Colin

NewsblogAlex Colome, Denard Span traded to Mariners
(17 - 1:55pm, May 27)
Last: snapper (history's 42nd greatest monster)

NewsblogCan't think of anything funny to put up for OMNICHATTER! for May 25, 2018
(68 - 9:15am, May 27)
Last: bobm

NewsblogThe Shohei Ohtani experiment isn't just working -- it's been perfect
(56 - 8:05pm, May 26)
Last: dave h

NewsblogPrimer Dugout (and link of the day) 5-25-2018
(9 - 5:31pm, May 26)
Last: RMc's Unenviable Situation

Page rendered in 0.1981 seconds
47 querie(s) executed