Tuesday, May 22, 2007

Does the best team win the Superbowl?

I've often been fascinated with the notion that tournaments and sports seasons don't always produce the "best" team... I decided to construct a spreadsheet (which I can send to anyone who wants it) to test this...

My model was the NFL... Here's my assumptions:

Modeled after the current schedule (i.e. two games against each division opponent, each division plays one division in the other conference round robin, each team plays 6 pseudo-random teams in their own conference)

Assumption number 1: No home field advantage. There are two reasons for this one... First of all, it would make the model incredibly complicated and make the scheduling virtually impossible... Secondly, the home field advantage in the NFL is only worth about 1.5 points... so it's relatively small... The only situation in which I see this making a big difference is the playoffs...

Teams stay consistently good (or bad) throughout the year: No fluctuations in strength for injuries, momentum, giving up at the end of the season, etc.

A team has a % strength... the chance of winning is the ratio of Team A's % Strength vs. Team B's.
Example 1: Team A has a 95% strength and Team B has a 5%. Team A has a 95% chance of winning
Example 2: Team A has a 50% strength and Team B has a 25%. Team A has a 66.66% chance of winning
No Ties were allowed

----------

The only test "control" for this type of experiment is to see if we can duplicate historical results. Since there's no way to know who the "best" team was in any given year, the next logical thing was to determine if a reasonable win distribution could be discerned. Specifically, the thing I focused on was number of undefeated teams and number of winless teams. The number of undefeated teams was given more weight because winless teams tend to fight a bit harder and often against opponents with nothing to gain... On the other hand, undefeated teams fight hard for the undefeated record and their opponents fight hard to claim the undefeated "scalp"...

Based on the methodology outlined above, a set of 250 years worth of results was run... Here's the summary of results:

Strengths range from 40% - 60%
The "best" team won the supervowl 10 times (4%)
The average strength of the two teams in the superbowl was 13.292 (i.e. between 13th and 14th strongest teams in the league)
Out of the 500 years (250 per conference) the best team in the conference made the playoffs 234 time ( 46.8%)
There were no undefeated teams during that time period
There were no winless teams during that time period

Strengths range from 20%-80%
The "best" team won the supervowl 31 times (12.4%)
The average strength of the two teams in the superbowl was 8.428
Out of the 500 years the best team in the conference made the playoffs 234 time (71.8%)
There were no undefeated teams
There were no winless teams

Strengths range from 0% - 100%
The "best" team won the superbowl 33 times (13.2% of the time)
The average strength of the two teams in the superbowl was 7.462
Out of 500 years the best team in the conference made the playoffs 416 times ( 83.2%)
There were 4 undefeated teams
There were 149 winless teams

The 0-100% group appears to be fairly accurate (though it does have a somewhat high preponderance for winless teams)

Also tested:
Strengths range from 20%-80% but strengths are squared before comparing ratios (results in the theoretical best team having a 94.11% chance of beating the theoretical worst team)
The "best" team won the superbowl 37 times (14.8% of the time)
The average strength of the two teams in the superbowl was 6.234
Out of 500 years the best team in the conference made the playoffs 417 times (83.4%)
There were 3 undefeated teams
There were 25 winless teams

This seemed to be a pretty good sweet spot with undefeated teams and winless teams being fairly accurate:

What does all this mean?
If you're trying to figure out the best team, the NFL format is not the one to go with... With only 16 games, you're bound to get abberations... Here's some interesting ones

In 250 years the best team in the league failed to record more than 8 wins 16 times... In the same time period, the worst team in the league recorded 8+ wins 6 times

I've long contended that the NFL is skewed for maximum excitement and minimum accuracy due to the incredibly short season... These results seem to bear that out...

1 Comments:

At 2:23 PM, Blogger Nick Manning said...

I would think basketball because of the make-up of a team would produce the most champions that had the best regular season record. I know they play 4 rounds but because the best players account for most of the scoring and defense and because an opponent would have to defeat them 4 times out of seven in a given series and because players get injured less in that sport the top regular season teams would win or at the very least appear in the finals more often than not.

Looking at the tournament format you might think hockey would be second but there are so many more players on a team, injuries are common, and regular season leaders can be upset by a hot late season team quite easily.

Baseball probably produces a large number of league leaders in the World Series but then selecting a winner from the two league leaders would be difficult in the last 10 years since interleague play started because they still play relatively few games against the other league and nearly impossible prior to interleague play becaus ethere is no regular season comparison.

The NFL and NCAA basketball tournaments would seem to produce the most champions from the field as opposed to the regular season leader because of the one and done likelihood in their respective play-offs.

I have done no mathematical calculations but I bet I'm right.

 

Post a Comment

<< Home