Friday, July 30, 2004

Are the Steelers rebuilding or was 6-10 a fluke?

Bill Cowher's new contract runs through 2007 but he has no time frame on how long he plans on coaching. He's going into his 13th season as the head Steeler and his 2004 squad will be much different than the one that ended the season last December with a loss to the Ravens.

All this talk got me to thinking about what many naysayers who claim to be in the know have been saying these past few months -- namely that the Steelers are rebuilding. And while I saw some pretty disgusting displays of football by the Steelers last season (Cleveland at home comes immediately to mind), I'm still not ready to concede that this team can't compete for the AFC North title and go deep into the playoffs.

And there two reasons I feel this way (quick stroll down memory lane -- one of my first posts to this site used a time-series model to predict how many games the Steelers will win in 2004 -- which turns out to be very similar to what I find here):

  1. I'm biased -- I like the Steelers. But maybe more importantly, Pittsburgh, despite the fact they only made one big free agent moves, did a lot to improve the team simply by releasing aged, overpaid veterans and giving younger guys a chance to prove they can play. The fact that they've replaced a lot of experience and waning physical skills with guys who possess unlimited athleticism and little experience should prove exciting (exciting in a good way, that is).
  2. The NFL is a lot different than any other professional sport for a couple of reasons. Obviously football is a lot more physical than other sports, but as a consequence there are many fewer games. As a result, teams that are evenly matched on paper often have very different records (I'll explain in detail below).

It's seldom the case that baseball teams with the most talent don't end up near the top of the league standings (barring injury or some other extreme circumstance). And the reason this is usually the case is because there are 162 games in a season, and over the long haul, talented teams have more opportunities to win games based on skill rather than luck. So while some team like the Kansas City Royals might win 10 games in a row, over the long MLB season, they won't come anywhere close to winning the division.

On the other hand, NFL teams can get on a mini-run that could be more a result of good fortune rather than superior talent. And any team -- good or bad -- able to string together a five or six game winning streak has substantially improved their chances of making it to the post-season. The other side of the coin obviously has the opposite effect. Just ask the Steelers. Last season they lost six in a row -- many games decided by one pivotal play that turned the tide in the oppositions favor -- and they never had a realistic chance of making the playoffs. So if the NFL had a 162 game season (and there were no injuries, etc...) we should expect to see what we see in MLB or the NBA -- a situation where the best teams finish near the top of the league.

So, back to my initial question: "Are the Steelers rebuilding after going 6-10, or was it just a statistical fluke?"

I used some statistical tools to take a look. Specifically, using something called the binomial distribution I was able to estimate the probability that a team would obtain X number of wins in a given season based on their past winning percentage.

[Let me make this disclaimer: These are just estimates and are in no way definitive. And that's primarily because I use the 2001, 2002 & 2003 seasons as a predictor of how teams will perform in 2004. For simplicity (and lack of time) I don't consider trades, drafts, age and anything else you can think of that affects team performance. More than anything this is a statistical exercise, but we can still make some comments based on the findings.]

Ok, let's get to the good stuff. I decided to look at the four teams in the AFC North and see how the division should stack up using the binomial distribution to estimate how many games each team will win in 2004 based on 2001, 2002 & 2003 performances (if it makes you feel any better, I've given the 2003 season more weight in determining 2004 performance than the 2002 & 2001 seasons).

Here's a look at how each team has finished the last three seasons:

200120022003
PIT 13-310-5-16-10
BAL 10-67-910-6
CLV 7-99-75-11
CIN 6-102-148-8
Using these win-loss records I estimated how many wins each team can expect in 2004. Here's what I found:
winsPITBALCLVCIN
00.0000.0000.0000.000
10.0000.0000.0030.004
20.0010.0000.0160.020
30.0060.0010.0500.057
40.0200.0060.1060.117
50.0510.0210.1660.176
60.1020.0530.2000.202
70.1580.1050.1870.181
80.1940.1610.1380.128
90.1870.1960.0800.071
100.1420.1880.0370.031
110.0840.1410.0130.011
120.0380.0800.0040.003
130.0130.0340.0010.001
140.0030.0100.0000.000
150.0000.0020.0000.000
160.0000.0000.0000.000
So what does all this mean? The first column are the number of wins a team can have over the course of the NFL season. Each subsequent column corresponds to each of the four teams in the AFC North. More importantly however, are the numbers under each of the four teams. These numbers correspond to the probability that a particular teams wins that many games. For example, the probability that the Steelers win exaclty eight games is 0.194 (or 19.4%).

Remembering that these numbers were generated based on the three previous seasons, we can get an idea of how each team should do in 2004. Despite all their offensive woes, the Ravens have been more consistent than the Steelers since 2001 and as a consequence they have a 19.6% chance of winning exactly nine games. The opposite can be said of the Bengals. They were awful in 2001 and 2002, but had a strong 2003. But because of their slow start, they're highest probability is associated with a six win season (0.202).
This graph shows the win distribution of each team in the division. As you might expect, the Steelers and Ravens are expected to win more games than the Browns and the Bengals.

And while it's fun to look at graphs with color and all that, I decided to take it one step further. Specifically, I wanted to take a look at the probability of each team winning at least eight games, at least nine games (and all the way to going through the season undefeated). Here's what those numbers look like:
probPITBALCLVCIN
w>70.6620.8130.2720.244
w>80.4680.6520.1350.116
w>90.2810.4550.0540.045
w>100.1390.2670.0180.014
w>110.0540.1260.0040.003
w>120.0160.0460.0010.001
w>130.0030.0120.0000.000
w>140.0000.0020.0000.000
w>150.0000.0000.0000.000
In the table above, prob refers to the probability of winning 'X or more games.' For example, the probability of the Steelers winning nine or more games is 46.8%. But the probability of the Steelers winning 12 or more games is only 1.6%. We can look at the other three teams in the division to get an idea of how they would fair in certain situations as well. Below is a graphical display of the table above:
What immediately sticks out is that it looks like the Ravens have a slight edge over the Steelers for first place in the division while the Browns and Bengals will be battling for last place. But as I mentioned at the beginning of this post, this was strictly a statistical exercise not accounting for personnel decisions. So while we can draw some conclusions, we have to take them with a grain of salt.

Ok, now that that's out of the way, these findings suggest that the Steelers are indeed not rebuilding but are probably closer to maintaining. And by that I mean that instead of posting another 5-7 win season in 2004, there much more likely to regress to their mean wins per season -- somewhere around 8.3 to 8.7 wins. And while that may please some, I suspect most Steelers fans would not be overly excited about an 8-8 season (myself included).

But as we've heard before (ad nauseum), the AFC North is a crappy division and 8 to 9 wins should produce a division title. We'll see. Either way, you've read it here first. The numbers say the Ravens hold a slight edge over the Steelers, but hey, that's why we play the games -- to see what actually transpires. And not only that, the Steelers were the run away favorites in 2003 -- anybody remember that?

Here's what the model predicts the AFC North final standings will look like in 2004:
2004WL
BAL97
PIT88
CLV610
CIN610
And while this was an interesting exercise, I'm still banking on the Steelers reeling off 16 in a row.