Does the preseason matter?

Yes. For the Tampa Bay Buccaneers, there is a correlation between their preseason performance and their regular season record.

I started off doing this research to prove to a buddy that the preseason does not matter. For decades I have scoffed at anyone that makes a big deal out of preseason performance and have always said “Who cares? The preseason does not matter.” I went into this to prove that point but instead walked away dumbfounded because, yes, I have been wrong for all those years. The preseason does matter.

Pearson Correlation Coefficient equation

Pearson Correlation Coefficient equation

Correlation

In doing this research, I went back to 1995 and gathered all of the Buccaneers preseason and regular season records. Now, what I expected was to find complete randomness in comparison to the two records, preseason and regular season. The first thing I did was to compare the records of both the preseason and the regular season to see if there was a mathematical correlation. I did this using the Pearson Correlation Coefficient which is just a fancy math label for comparing numbers to see if there is any relationship. What I found surprised me.

Preseason and regular season correlation points

Preseason and regular season correlation points

The way the Pearson Correlation Coefficient works is it calculates sets of numbers and spits out a number, called a coefficient, which represents how close the numbers relate to each other. It works on a scale between 1 and -1. The closer the coefficient is to 1 then the more the two numbers have a positive correlation with each other. If they have no correlation then the coefficient is 0. If there is a negative coefficient, say -.63, then it means that there is a correlation of the two numbers but they are reversed. In our case, it would mean that a losing record in the preseason would give us a winning record in the regular season. But that was not the case. The correlation coefficient for the Buccaneers preseason and regular season records is .45. That surprised me, to say the least, because what it means is that there is a positive and moderately strong correlation between the preseason and the regular season. Forty-five percent of the time our preseason record will correlate with our regular season record.

Were it up to chance, the correlation coefficient would be closer to .20 because there are five different possible preseason records (0-4, 1-3, 2-2, 3-1, 4-0) and predicting the correct preseason record randomly would occur approximately 20% of the time.

The correlation within the past decade has been even closer. Since 2004, our correlation coefficient has been .54. Only twice in the past ten years has our regular season and preseason not matched within 3 games. If we remove 2004 out of the equation then we get a whopping .71 correlation. That is pretty damn strong.

Predicting Preseason Win-Loss Record Based on Regular Season Record

This was not anything at all like I expected so I decided to try it out a different way. I took the regular season records and, without looking at the preseason record, predicted how the preseason record would look. This was a simple matter of dividing the possible regular season records into five categories to match the possible preseason records. Below is the chart I used in comparing the records.

Preseason record Regular Season Record
0-4 0-16
0-4 1-15
0-4 2-14
1-3 3-13
1-3 4-12
1-3 5-11
1-3 or 2-2 6-10
2-2 7-9
2-2 8-8
2-2 9-7
2-2 or 3-1 10-6
3-1 11-5
3-1 12-4
3-1 13-3
4-0 14-2
4-0 15-1
4-0 16-0

As you can see, there is an overlap in the 10-6 and the 6-10 regular season records. There are 4 games in the regular season that match those records and I did all kinds of calculations using all of the different variables and the results still came out close enough that I am confident that those overlaps do not pose a danger to the outcome being incorrect.

In any case, when you take the regular season records and then try and predict what the preseason record was by using this chart, 9 of the 19 preseason records, or 47% of the time, you can predict the preseason record by looking at the regular season record. If the preseason had no impact on the regular season record then the percentage would have been closer to 20% due to randomness. This method shows that almost half of the time the preseason and regular season records match. Damn.

Season Regular Season prediction Preseason
1995 7-9 2-2 3-1
1996 6-10 1-3 or 2-2 1-3
1997 10-6 2-2 or 3-1 1-3
1998 8-8 2-2 2-3
1999 11-5 3-1 4-0
2000 10-6 2-2 or 3-1 3-1
2001 9-7 2-2 1-3
2002 12-4 3-1 3-1
2003 7-9 2-2 4-1
2004 5-11 1-3 3-1
2005 11-5 3-1 2-2
2006 4-12 1-3 1-3
2007 9-7 2-2 3-1
2008 9-7 2-2 3-1
2009 3-13 1-3 1-3
2010 10-6 2-2 or 3-1 2-2
2011 4-12 1-3 2-2
2012 7-9 2-2 2-2
2013 4-12 1-3 1-3

When Going 1-3 in the Preseason

So then I decided to look at each individual record set and see what I could see. The Buccaneers did not have a 0-4 record in those 19 years but in those 19 preseasons, the Bucs went 1-3 six times. Of those 6 times we only had two winning seasons, a 9-7 season in 2001 and a 10-6 season in 1997. In the other 4 seasons going 1-3 in the preseason we went 4-12 in 2006, 3-13 in 2009, 6-10 in 1996, and 4-12 in 2013. According to those numbers, when we have a 1-3 preseason record our chances of having a losing season is 67%. That is not random.

Year Preseason Prediction Regular Season
2006 1-3 1-3 4-12
2009 1-3 1-3 3-13
1996 1-3 1-3 or 2-2 6-10
2001 1-3 2-2 9-7
1997 1-3 2-2 or 3-1 10-6
2013 1-3 1-3 4-12

When Going 2-2 in the Preseason

We went 2-2 in the preseason 5 times. In those years our regular season records were 7-9, 10-6, 4-12, 11-5, and 8-8. If you compare the numbers to the chart that is two seasons of near .50 records, one season right on the border of 2-2 and 3-1, and two seasons that are off on the correlation. So with a 2-2 preseason record, there is a 60% that the Bucs will have a regular season record between 6-10 and 10-6. That is not random.

Year Preseason Prediction Regular
Season
2012 2-2 2-2 7-9
2010 2-2 2-2 or 3-1 10-6
2005 2-2 3-1 11-5
2011 2-2 1-3 4-12
1998 2-3 2-2 8-8

When Going 3-1 in the Preseason

The Bucs have gone 3-1 six times in the preseason since 1995. Of those six times we had winning seasons 4 times. A 5-11 season in 2004 and a 7-9 season in 1995 were the two losing seasons. With a 3-1 preseason record we have gone on and had a winning season 67% of the time. That is not random.

Year Preseason Prediction Regular Season
2002 3-1 3-1 12-4
2000 3-1 2-2 or 3-1 10-6
2007 3-1 2-2 9-7
2008 3-1 2-2 9-7
1995 3-1 2-2 7-9
2004 3-1 1-3 5-11

When Going 4-0 in the Preseason

We have gone 4-0 once and that was in 1999. Our regular season record that year was 11-5. We went 4-1 once in 2003 and ended up with a 7-9 record. Winning 4 games in the preseason gives us a 50% chance of a winning record in the regular season. It is a small data set, to be sure, but that outcome can be considered random.

So:

  • a 1-3 preseason record gives us a 67% chance at a losing record in the regular season;
  • a 2-2 preseason record gives us about a 40% chance of a losing record in the regular season;
  • a 3-1 preseason record gives us a 33% chance of a losing record in the regular season;
  • and a 4 win preseason gives us a 50% chance of a losing record in the regular season.

Looking at that data, it is fairly easy to see that as our preseason record gets better, our chances of having a losing season decreases. The opposite applies as well. The better our preseason record, the better our chances of having a winning regular season record.

By Wins/Losses

I then decided to look at general wins and losses and see if a winning preseason could predict a winning regular season and vice versa. Excluding all of the 2-2 preseasons, of which there were five, 9 out of 14 times our preseason record has matched our regular season for a winning or losing season. That means that if we win more games than we lose in the preseason then there is a 64% chance that our regular season will also have more wins than losses. It also means that if we have more losses than wins in our preseason then there is a 64% chance that our regular season will also have more losses than wins. It is actually a pretty low chance, around 36%, that we will have a winning preseason and then a losing regular season or a losing preseason and then a winning regular season.

To say that I was not expecting these type of results would be an understatement. For nearly 40 years I have claimed that the preseason does not matter. My math and analysis could be totally ass backwards. If there is anyone that sees errors in my work, please let me know. But until then, the bottom line is that, while it is not an exact correlation or even a very large one, there is a correlation between the preseason and regular season records. Football is a very unpredictable sport and there is never going to be a 100% correlation between anything. We all know that on any given Sunday that a winless team can beat a team with no losses. But more often than not, the winless team will lose. Not always, but more often. With the numbers I have calculated it shows that more often than not, our regular season record is a reflection of our preseason record. Call me dumbfounded but it appears that the preseason does matter.

This entry was posted in Analysis, Stats. Bookmark the permalink.