Remember me
Forgot your username/password?
 
Print
Home>Articles

How Meaningful are the First Ten Games?

By John Dewan
April 19, 2012

It’s rather remarkable that such a small sample size can show a trend, but sure enough, based on the last ten years of data, the first ten games of the season do just that.  Since 2002, 44 teams have started the season with 3 or fewer wins in their first ten games.  Only 4 of them (9%) have gone on to win 90 games for the year.  On the flip side, 46 teams have started the season with 7 or more wins and 21 of them (46%) have gone on to win 90.

Nevertheless, there are exceptions and last season had more than its fair share.  Three of the four instances in the last ten years where a team wound up with 90 wins despite a horrible start occurred last year.  Both the Boston Red Sox and Tampa Bay Rays started the season at with a 2-8 record, and both went on to win 90+ games.  The Tigers started 3-7 and wound up with 95 wins.  Of the five teams that started with 3 or fewer wins in their first 10 games, three of them went on to win 90 or more.

Among the fast starters, only two of the teams that started with 7 or more wins in their first 10 went on to win 90 or more.

In 2011, more poor starting teams did well by season's end, compared to the fast starters.

This year, the slow starters with only 3 wins in their first 10 games are the Kansas City Royals, Minnesota Twins, Chicago Cubs, and Pittsburgh Pirates.  The San Diego Padres only won two of their first 10.  If historical norms hold, it is unlikely that any of these teams will go on to win 90 or more.  Based on pre-season projections for these teams (the only team we had near .500 was the Padres), this comes as no surprise.

The fast starting teams are the Los Angeles Dodgers (9-1) and the Texas Rangers (8-2).  The Detroit Tigers, New York Mets, St. Louis Cardinals, Arizona Diamondbacks and Washington Nationals started with 7 wins in their first 10 games.  If past trends hold, four of these teams will finish with 90 or more.

Here’s the complete chart for the last ten years:

Performance in First Ten Games (2002 - 2011)

Wins in
First 10
 Teams Finished
90+ Wins		 Pct of
Total		 Made
Playoffs		 Pct of
Total
0-3 44 4 9% 5 11%
4 68 14 21% 14 21%
5 78 20 26% 23 29%
6 64 20 31% 19 30%
7-10 46 21 46% 19 41%

 

 
 

COMMENTS (9 Comments, most recent shown first)

tangotiger
http://www.insidethebook.com/ee/index.php/site/comments/true_talent_levels_for_sports_leagues/
11:28 AM Apr 21st
 
lidsky
Hi Tango,

Forgive me if this is a newbie question, but why did you chose 70 games for your regression toward the mean calculation?
2:09 AM Apr 21st
 
tangotiger
Regression toward the mean would suggest adding 35 wins and 35 losses, to whatever record you happen to have.

So, a 10-0 team would have an expected rest-of-season winning percentage of 45/85=.529. Over 152 games, that's 80.5 wins. Add in the 10-0 start, and on average, that's 90.5 wins.

Obviously, that's an average, and the distribution of expected 90+ wins would have to be calculated.

What we care about is if the regression toward the mean approach is different based on the number of games, or the record. And I can't tell this by John's data.

Bill James did something like this back in the 1985 Abstract, when trying to explain the Tigers hot start. I would not be surprised if a regression toward the mean approach as described here would have explained the rest-of-season winning percentage in Bill's study.

2:25 PM Apr 20th
 
Robinsong
Given 79 teams who finished at 90 wins or above. Assume average winning percentage is .580 - 94 wins. Expected number of teams in any 10 game period with:
7+ wins- 26
6 wins - 20
5 wins - 17
4 wins - 10
3- wins - 6
So this interesting analysis shows us that top teams are likely to get off to a slightly more average start than expected based on their eventual winning percentage. We care about the first ten games because we want to predict the future, but John's data is quite consistent with the notion that the beginning of the season is like any other ten-game period. A 10 game interval is a pretty significant sample, and tells us a lot about team quality.
1:52 PM Apr 20th
 
jemanji
Re: same pattern at any 10-game stretch ... perhaps. But later 10-game stretches occur after teams have placed "shims" on their early problems.

The 10-game stretch out of the gate may be more directly measuring the 25 players that are most core to the various 30 franchises. Later nicks, dings, injuries, role changes, etc., might camouflage a team's talent pyramid more; the first 10 games might show the quality of an org's talent pyramid more nakedly.

Or not. But the 46% vs 9% gap is considerably wider than I'd have guessed it was.
.
6:16 AM Apr 20th
 
jemanji
Those: you might expect that, but then you are *acknowledging* that this first .700+ stretch is something characteristic of 90-win teams.

John asks, is there any predictive validity to the first ten games, and then he shows a 9% vs 46% comparison.

90-odd teams isn't a huge number, and a binomial calculator would show we don't have a Grand Theorem here, but think about it. How surprised would you be if it had shown 46% of 3-7 teams had won 90, vs. 9% of 7-3 teams winning 90?

Nice article. A team starting 2-8 may be in a little more trouble than we thought they were.
6:12 AM Apr 20th
 
tangotiger
I agree, what we care about is how they do in their next 152.


10:16 PM Apr 19th
 
those
I'm skeptical that this is more than a tautology. All this shows is that 90-win teams are more likely to have success over a given 10-game period than 90-loss teams. Wouldn't we expect to find the same pattern at pretty much any 10-game stretch in the season, say games 51-60 or 98-107?
3:42 PM Apr 19th
 
Trailbzr
This would seem to lend itself readily to a weighted average prediction for the full season. Like, expected winning percent over subsequent 152 equals: X% the record (as a percentage) first 10 games; Y% the record last season; and (100-X-Y)% of .500 ball.


10:10 AM Apr 19th
 
 
©2024 Be Jolly, Inc. All Rights Reserved.|Powered by Sports Info Solutions|Terms & Conditions|Privacy Policy