Thursday, September 12th

When the dust settles on the first Sunday of the NFL year, almost all of us are disappointed by the performance of at least one of the players we drafted. Disappointment can turn to panic, and panic can cause us to make decisions we might later regret.

This column is an effort to step back, count to ten, and take an objective and unemotional look at what week one performance generally means in the grand scheme of things. So step away from the ledge, at least for long enough to read this. There may still be hope.

Running Backs

I took a look at all running backs from 1995-2001 who fit the following profile:

  1. top 20 RB from the previous year;
  2. got at least one carry in week one;

That amounted to 127 running backs. I looked at each of their week one performances and sorted them from best (Emmitt Smith 1995) to worst (Marcus Allen 1997) in terms of fantasy points. Then I cut the list in half. The guys at the top of the list were said to have had a "good week one" and the guys at the bottom were labelled as "bad week one."

So what happened to these two groups for the rest of the year? 


                                  ----- fantasy points ---
                                  Prev yr   Wk 1   Wk 2-17
----------------------------------------------------------
Avg RB who had "good week one"      214      17      184
Avg RB who had "bad week one"       204       5      137

The first column (previous year fantasy points) is there to demonstrate that the two groups had been roughly comparable the previous year, with the "good week one" group being just a shade better. The second column tells you the obvious: that the "good week one" group did better in week one. But the interesting thing is that it didn't stop there. The "good week one" group, on average, went on to outscore the "bad week one" backs by 47 points over the rest of the season.

According to the raw data, a bad start by your star RB just might be something to be concerned about.

A glance through the complete data set, though, should ease your mind a bit. In particular, you'll notice that a lot of the "bad week one" backs who went on to bomb for the rest of the year were marginal backs for whom the dropoff was actually quite predictable. Even though Rodney Thomas had a pretty good season in 1996, for example, everyone knew that he didn't have a job in 1997. The fact that he followed up his unproductive week one with an unproductive season tells us nothing about what to expect from Corey Dillon this year. And the bottom of the list seems to be populated with similar stories: Derrick Loville, Ironhead Heyward, Lewis Tillman, etc.

I'm definitely abandoning my promise to be objective, and possibly using too much hindsight, but my impression is that, even though their numbers were similar in the previous year, the "good week one" backs were actually significantly better than the "bad week one" backs. And I'd bet that that explains the difference in the week 2-17 performances of the two groups. My general impression is that most of the backs who had bad week ones did about as well as they were expected to do (barring inury).

But don't take my word for it. Browse the data set yourself and draw your own conclusions.

Wide Receivers

This is a much easier case. It's clear that an established WR's week one perforance should not affect your opinion of him. Using the same methodology, we have the following: 


                                  ----- fantasy points ---
                                  Prev yr   Wk 1   Wk 2-17
----------------------------------------------------------
Avg WR who had "good week one"      182      16      136
Avg WR who had "bad week one"       170       4      124

The "good week one" WRs were, on average, 12 points better than the "bad week one" WRs in the previous year, and they were 12 points better over the rest of the test year.

So my default opinion is that, for WRs, a bad week one is absolutely no cause for concern. I may override the default in special cases, but I'd make sure I had a very compelling reason to downgrade a WR after week one.

Just to highlight one quick example, consider the Big 3 [TM] receivers: Terrell Owens, Marvin Harrison and Randy Moss. All three had uninspiring openers last week. But note that they are, as a group, ahead of where they were at this time a year ago. 


Big 3 combined numbers in week 1:

Year         REC    YD     TD      fantasy points
-------------------------------------------------
2001          12   156      0            16
2002          14   133      2            25

Quarterbacks

With QBs, I only considered those who were top 10 the previous year, rather than the top 20. Other than that, the methodology was the same for QBs as it was for RBs and WRs. 


                                  ----- fantasy points ---
                                  Prev yr   Wk 1   Wk 2-17
----------------------------------------------------------
Avg QB who had "good week one"      298      24      231
Avg QB who had "bad week one"       285      10      162

This looks like the RB data, but is more extreme. The previous year, the two groups were, at least statistically, very comparable. Their post-week-one performance though, differed by an average of 69 points. That's a serious chunk.

Can I wave my hands and explain this away like I did with the RBs? I'm not so sure. It's far from universal, but a quick pass through the list indicates to me that a bad week one has foreshadowed an unexpectedly bad season in a lot of cases. Again, I'll invite you to examine the data and draw your own conclusions, but I am not going to offer any reassuring words to Jeff Garcia owners (a group to which I belong, by the way).

One last thing

I can't resist asking one last question:

If you took all these piles of data to a numbers-whiz who knew absolutely nothing about football (many people fitting that exact description reside in offices right down the hall from mine), what would he or she conclude about the importance of week one? In particular, there are specific statistical tests that can be run to answer questions like "is week one performance a significant indicator of week 2-17 performance?" and "is week one performance a better indicator than last year's performance?"

Those tests, obviously, are blind to things like 40 times and injuries on the offensive line. Thus, they fail to take into account lots of relevant factors. On the other hand, they fail to include the biases that we all bring to whatever we're trying to analyze. And yes, we all do bring biases, some of which we're aware of and some of which we're not. I don't pretend that some sophisticated statistical tests are going to answer anything definitively. They simply provide one more viewpoint.

That said, here is what the number-savvy footballaphobe would tell you about the data.

  1. For running backs, previous year fantasy points and week one fantasy points are both unquestionably useful in predicting week 2-17 points. Roughly, they are equally useful.

  2. For wide receivers, previous year points are useful in predicting week 2-17 points. But, if you know last season's points, then week one points are not useful at all in predicting week 2-17 points.

  3. For quarterbacks, week one fantasy points are useful in predicting week 2-17 points. But, if you know week one points, incorporating last season's points does nothing to improve your predictions of week 2-17 points.

It's interesting that you get three very different answers for the three positions.

Unless otherwise noted, all stats come from football-reference.com and the disclaimer applies