Winning in extra innings is more luck than skill
Posted by Andy on July 29, 2010
The Mets and Cardinals played 13 innings last night after playing 20 earlier this season. As of yesterday, there have been 276 team games in extra innings this year, so an average of about 9 per team. The Cardinals have been in 9 games, going 4-5, while the Mets (visible on the same link) have played in 12, going 4-8.
Incidentally, on that detailed standings page, you can also see that the teams with the best winning percentage in extra innings are the Royals (7-2), Astros (6-2), and Orioles (7-3) despite the fact that each team is in the bottom 7 in overall W-L% this year. Weird. Neil put it best in a chat window to me just now: "Further proof that winning close games is more luck than skill".
July 29th, 2010 at 11:06 am
According to the stats on the detailed standings page, the average team is 5-4 in their last 10 games, 10-9 in their last 20, and 14-15 in their last 30.
July 29th, 2010 at 11:10 am
Those are rounded of course and off-days may mess up the calculations. The average of all teams is going to be .500 over any period of time.
Somebody actually asked me how many wins is average for an MLB team each year...I pointed out that each year, the average number of wins is 81...
July 29th, 2010 at 11:23 am
Although random factors undoubtedly play a significant role in determining extra innings records, the fact that teams with poor overall records have better extra innings records doesn't necessarily prove that these games are basically a flip of the coin.
Generally speaking, the teams with the worst overall records have weak starting pitching. In extra inning games the starting pitching has been more or less neutralized (or the game would not have gotten through 9 tied). A team with a stronger bullpen (relative to its starting pitching) should perform better in extra innings relative to its overall record.
July 29th, 2010 at 11:39 am
Andy,
Not to nitpick, but 81 wins isn't always accurate because teams occasionally play fewer than 162 games because of rainouts that aren't rescheduled (although Bud Selig has made a point of emphasis of cutting down on this) or more than 162 games when an additional game is needed to break a tie for a division or wild card position.
July 29th, 2010 at 11:43 am
OMG Evan. That's pretty obviously nitpicking.
July 29th, 2010 at 12:02 pm
Evan,
Bad teams, on average with great variation, are bad at everything. This includes starting pitching, but also relief pitching, bench and every other aspect of the game. Just possibly the eleventh or twelfth pitcher on a bad team may be closer in ability to the 11th or 12th on a good team, than the top starters, but there is no reason to expect them to be better on the bad team.
If you took large numbers of good, (say, .600 W-L), and bad (.400), teams I would expect the good teams to be better than the bad teams in extra innings games, but that both groups would have regressed towards the mean. So we might expect to see .550 and .450 so some such. Would need very large sample sizes to get rid of the noise.
July 29th, 2010 at 12:27 pm
Is there a way to easily find out the most teams over .500 in a season?
July 29th, 2010 at 12:43 pm
#7 If so, I don't know how to do it. It is easy, however, to find most wins in a season:
Generated 7/29/2010.
July 29th, 2010 at 1:03 pm
To me, extra innings is usually about who can score first. Not always, but usually. I wonder how the Royals/Astros/Os do in this regard? The obvious difference between regulation and extra innings is who scores most versus who scores first. I'm not suggesting those 3 would be the ML leaders but maybe they're better at scoring first than scoring most.
July 29th, 2010 at 1:05 pm
#7/"Bryan Mueller Says: Is there a way to easily find out the most teams over .500 in a season?"
I'm guessing that this would occur in a league with a team (or two) with an unusually bad record, so that the wins would be "spread" out amongst more teams. For instance, in the 1904 AL, the Senators were 38-113, and 5 out of 8 teams were above .500. In the 1962 NL, there were TWO bad teams, of course the Mets but also the Cubs lost 100+ games, and there were 7 out of 10 teams above .500.
July 29th, 2010 at 1:08 pm
The bottom line about extra innings is that randomness plays a much bigger factor since most extra inning games are decided in an inning or two. Even the lowly Orioles have a shot at beating the best teams in baseball when they need score only 1 run. I would suspect that over the years, record in extra innings shows little-to-no correlation with record in regulation-length games.
July 29th, 2010 at 1:17 pm
#8 is missing the team that was the most games above .500 all-time, the 1906 Cubs who went 116-36 and then lost the WS in the only all-Chicago WS ever and also the last WS where both teams were making their first appearance in the WS.
Back to this topic, what a typical SABR loving theory that anything that cannot be explained by some stat is just luck....yet the #1 thing that SABRs love to point out about just about everything is that the sample size is too small. So 3 teams playing 10 games in extra inning in 1 season is enough to "add further support" to a theory? First of all, was there any extra digging into these games? The Royals are terrible but they have 1 of the best closers in the game...who saved all 5 of their road extra inning wins. The Orioles had 6 home extra inning games, and they went 2-2 in road extra inning games so not that dominant. The Astros are pretty weird going only 1-1 in home extra time games and 5-1 on the road.
This does not mention of course the next teams in extra inning games: LAD, TBR, PHI, MIN, TEX, SDP, NYY...all playoff contending teams. The most extra inning games from this group is 12 with the Phillies, rangers and Padres at 7-5....only 1 extra inning game "behind" the Orioles.
July 29th, 2010 at 1:22 pm
Yes, sorry, #8 is 1920-present and of course favors teams playing a 162-game schedule.
To be clear, in #11 above I'm talking about randomness, not luck. I'm saying that a team like the Orioles does score runs, and also does prevent runs from being scored sometimes, and there are going to be lots of 1-2 inning occasions when they manage to do that, so for them to come out ahead in a small number of extra-inning games is not surprising.
One wonders if all games were 2 innings long--what would records look like at the end of the season?
July 29th, 2010 at 2:11 pm
Kds @6
My point was and is that teams are rarely, if ever, equally bad, good or mediocre at everything. Under the premise that the factors which influence performance in extra inning games do not completely correspond with those which influence overall record, some teams are more or less likely to excel in extra inning games based-upon strengths and weaknesses.
e.g. if team A has average offense, well-below average starting pitching and slightly below average relief pitching, it is better positioned to compete in extra inning games because it is now utilizing comparatively better pitching than it does in a general game where the starting pitcher is highly influential.
My overarching point is that you can't assume that poor teams performing well in EI games means that EI games are randomly determined.
I would expect most poor teams to have better relative relief pitching than starting pitching for two reasons:
1) Capriciousness: relief pitchers seem to have much greater variation in performance from year-to-year, so it is more difficult to for an already above-average team to sign tailor its bullpen with free agents or trades than it would be to add a quality starting pitcher;
2) Self-fulfillment: Starting pitching is generally a better proxy for overall record than relief pitching - good starting pitching is one of the reasons good teams are good in the first place
July 29th, 2010 at 2:33 pm
I would suspect that over the years, record in extra innings shows little-to-no correlation with record in regulation-length games.
I'm guessing there is a decent correlation. Not a strong one, but certainly more than "little-to-no." Certainly in any one season a good team could play badly in extra inning games, and vice versa. If you look at all records of all teams over several years, I bet good teams play above .500 in extra innings. Basically I agree with Kds #6. Anyone want to actually dig up the numbers?
July 29th, 2010 at 3:14 pm
@Andy #2
I assumed it was rounding and just posted because I thought it was kind of funny looking, but the more I think about it, the less I understand how that happened. How are game record averages calculated? I would think you'd just add up the wins, add up the losses, and divide by total games, which when looking at league-wide standings would always give a perfect .500 record. Especially since in the last10 column, for example, every team has played exactly 10 games. Looking at the other columns, total W-L, pyth W-L came out to a perfect 50/50, which is surprising since I'd expect the rounding error there since not all teams have played the same amount of games. R vs. RA came out perfectly as did Home+Road splits, Inter, ExInn, 1Run, and vRHP+vLHP. The only other column(s) with a discrepancy is >=.500 and <.500 add up to a 49-50 record. So are the basic calculations for the averages for all of these columns the same and by some fluke rounding error some came out to 50/50 and others didn't? And if so, what's the calculation being used?
(I'm not trying to be critical or anything, you guys consistently do an amazing job on every facet of this site. It's just that as an engineer I have a compulsive need to know how mis-matched data came to be that way)
July 29th, 2010 at 3:42 pm
Djibouti @16,
Remember that league-wide records for Last 10, 20 etc. don't necessarily have to produce a .500 record. Because of variations in off days Team A's victory over Team B might be included in one team's Last 10, but not the other's.
If you add up the combined records of all teams L10 (fairly easy if you sort by L10 record) you get 152-148 (as of the morning update for 7.29.2010). Which would make the average record 5.0666-4.9333. The code which produces the league averages must be truncating this number instead of rounding it.
July 29th, 2010 at 5:03 pm
The Cardinals won because the Mets were stupid enough to pitch to Pujols. Check out http://www.thefantasybaseballguru.com leave comments ask qs and tell me what you think.
July 29th, 2010 at 11:07 pm
The only teams I found with perfect records in extra-inning games were the 1995 Indians (13-0), the 1928 Athletics (8-0), and the 1948 Reds (5-0). At least as impressive were the 1949 Indians (18-1) and the 1959 Pirates (19-2, the most extra-inning wins of any team since 1901).
At the other extreme, the 1969 Expos were 0-12, and the 1982 Twins were 1-13.
The 1943 Red Sox played 31 extra-inning ballgames; the 1936 Browns, only 3.
July 30th, 2010 at 2:47 pm
Was going to try comparing records to extra-inning records but it's too hard to pull the numbers into Excel. The extra-inning records are given in one column on the league pages (8-5), and Excel keeps turning that into a date (5-Aug) when I copy them in. I don't have the time or patience right now to enter them all manually. Anyone got any Excel tips to avoid that?
July 30th, 2010 at 11:20 pm
JT,
Before copying into the excel sheet, click on the column header you want to put the EI records into. Then select format, cells (I forget the exact terminology) and format them as text. It should then treat whatever you enter as text instead of trying to interpret the numbers as dates. If this doesn't help or is confusing, post back here and I will repost from a computer that has Excel installed so I can give better instructions.
July 31st, 2010 at 10:38 am
I was trying something along those lines but not exactly the way you described. This seems to work. Thanks. Will try to find time later to run the numbers and see what they say.
August 1st, 2010 at 12:10 am
Alright, I ran some numbers. Some caveats: (1) Even with Evan's helpful hint, I still had to separate the extra innings records into 2 columns and could have mistranscribed a number; and (2) I'm a bit drunk.
I looked at the records since 2006, so 5.5+ seasons. First, the correlation between overall record and extra-inning record is .218. Not a great correlation, but certainly a positive one. I had thought it might be about 0.3. It would probably be better to compare records in 9-inning games, instead of overall records, to extra-inning games, but I'm too lazy for that right now.
The average winning % of teams over .550 is .584, and their extra-inning record is .556.
The average winning % of teams under .450 is .410, and their extra-inning record is .472.
The average winning % of teams between .450 and .550 is .506, and their extra-inning record is .489.
August 1st, 2010 at 8:16 am
Good stuff, JT. Looks like if you took the extra-inning games out of the overall numbers, the correlation would get a little worse but still be real.
It goes along with what I was thinking--in the sense that there's enough randomness to make almost any outcome a strong possibility. Over the years I'm sure there are tons of example of teams that win only 40% of their games overall but won 60% of their extra-inning games, etc.