Rookie Points
|
Posted 9/2 by Chase Stuart, Exclusive to Footballguys.com
|
As we get ready for our fantasy drafts, rookies will be among the most exciting
players to pick. Which Jones should you take, Julius or Kevin? How about which
Williams, Roy or Reggie? While I'm not going to tell you which rookies will
do well-I'd like to see when they're going to do well.
Remember LaDainian Tomlinson's first season? He averaged 21.0 Fantasy Points/Game
(FP/G) in September, 12.2 FP/G the next two months, and just 9.9 FP/G the final
quarter of the season. Many assumed that his decline coincided with him hitting
the proverbial rookie wall. How about Andre' Davis? "Taking opponents by
surprise," the Browns rookie scored fifty points the first five weeks of
the 2002 season. Yet "defenses must have figured him out," as he couldn't
even score thirty points the last eleven weeks. We've all heard these explanations
when a quick-starting rookie falls off his hot pace.
What about William Green? In his rookie year he had six games of at least ten
fantasy points-and they all came in the last seven weeks. He had "proved
himself" as the season wore on, and the Browns began to lean on him for
their playoff push. What about Marvin Harrison, who scored over seventy percent
of his points in the second half of his rookie year? Well certainly, the game
had begun to "slow down" for Harrison, and he was finally figuring
out opposing defenses.
This got me to wondering-which cliché holds water? Do rookies get better
or worse as the season goes on? Fortunately, we've got more than enough data
to find out. Let's start with the running backs. Here's what I did:
- Find the weekly stats of all rookie RBs for the last nine years (from 1995-2003)
- Include only those who ranked in the top thirty in fantasy points, and played
at least twelve games.
- Using a weighted average system (as described by Doug Drinen here),
conclude as to whether talented rookie RBs perform better early or late in
the year.
Why use a weighted average system, instead of simply first and second half
splits? First half splits don't assign any more weight to what happens in week
one, than to what happens in week seven. Second half splits have the same problem.
A player that declines steadily throughout the year might have better first
half splits; but someone that has two huge games in weeks six and seven, will
show similar splits. In order to capture the true feeling of how the season
progressed, here's the formula used:
Weighted Average = |
1*(Game 1 FPs) + 2*(Game 2 FPs) + 3*(Game 3 FPs) +
... + 16*(Game 16 FPs)
1 + 2 + 3 + ... + 16
|
This weighted average would thus give more weight to later games, and help
solve the problem caused by first and second half splits.
To determine how back-loaded (or front-loaded) a season was, we simply calculate
the Weighted Average FP/G minus the normal FP/G. As Doug Drinen did, we'll call
this the E/L (Early/Late) factor. A positive E/L factor means a player is a
fast finisher; a negative E/L factor is for strong starters. Obviously, the
higher the number in either direction, the faster the finisher (or stronger
the starter) truly was.
Example
The chart below shows the sample points distribution for a fictitious rookie.
All sixteen weeks are shown, along with his fantasy points for each week.
Week
|
1
|
2
|
3
|
4
|
5
|
6
|
7
|
8
|
FPs
|
5
|
5
|
5
|
5
|
10
|
10
|
10
|
10
|
Week
|
9
|
10
|
11
|
12
|
13
|
14
|
15
|
16
|
FPs
|
15
|
15
|
15
|
15
|
20
|
20
|
20
|
20
|
This rookie has scored 200 FPs in the season, averaging 12.5 FP/G. It's clear
that he improved as the season went on-what the E/L factor tries to do is capture
by how much did he improve. Using the formula described above, this player scores
15.4 Adjusted Fantasy Points per game. When you subtract his regular
FP/G from his Adjusted FP/G, you are left with 2.9 FP/G. The E/L factor for
this rookie is thus 2.9. Remember, positive E/L ratios indicate fast finishes,
while negative E/L factors show strong starts.
Running Back Results
There's not much weight behind the "hitting the rookie wall" philosophy.
Of the thirty-two RBs that met the qualifications above:
- Twenty-three had a positive E/L ratio, signaling strong second halves. Of
these, fifteen were of 1.00 or above. The average was 1.45
- Nine had negative E/L ratios, with just two of -1.00 or below. The average
was -0.55.
- Overall, the thirty-two RBs combined for a resounding average E/L ratio
of 0.89 per player.
- If we limit this to just the past five seasons, the average E/L factor
increases to 1.09. If we count just the last three years, it rises again
to 1.32. If we make it only the last two, it jumps all the way to 2.52-based
on the way Clinton Portis (4.07) and William Green (3.47) finished their
rookie years.
- If we limit this to RBs drafted in the first three rounds, the average
E/L rate stays about the same-0.85. First or second round? 0.94. First
round picks only? The E/L factor for them drops to 0.54. This isn't surprising-a
first round pick is more likely to start immediately, and thus won't appear
to finish as fast. For a later round pick, earning that starting job could
take awhile.
- If you only care about "stud" RBs and recent history, consider
this: Running backs drafted in the first two rounds for the past three years
have an E/L factor of 0.96.
- How about counting only rookies that have finished in the top twenty
since 1995? The E/L factor jumps to nearly one. Top 12? Makes it 1.17. Top
8? 1.39. Top five? 2.16.
- This is a bit surprising. It says that the more points a rookie RB scores,
the more skewed his stats will be towards late season success. Clinton
Portis, Mike Anderson and Fred Taylor all started their rookie years as
backups, but finished as stars. While that's not entirely accountable
for the interesting result, it's certainly a large part of it.
- How about this: All rookie running backs that:
- Were drafted in the past five seasons
- Ranked in the top twenty-four their rookie years, and
- Were first day draft picks (top three rounds)
- That leaves just five players: Edgerrin James, Jamal Lewis, Clinton
Portis, Anthony Thomas and LaDainian Tomlinson. Average Early/Late
factor? 1.57.
Conclusions
- Rookie RBs generally score more points in the second half of the season
than the first-don't give up on your rookie too early.
- The trend is getting stronger. In recent years, rookie point distribution
is coming more towards the end of most seasons. Definitely, don't give
up on your rookie too early.
- We would expect to see high draft picks have more consistent stats, while
low draft picks might start to play and produce mid-season. While there are
some stats to support this, the correlation isn't as strong as I would have
predicted.
Wide Receiver Results
Using the same general idea, I applied this to Wide Receivers as well. All
rookies since 1995 were included, provided they finished in the top sixty that
season. This left us with thirty players.
- The average E/L factor was just 0.17 for WRs-an almost negligible
number. Seventeen had positive E/L factors, two had zero E/L factors, and
eleven performed better in the first half than the second.
- Eighty-percent of the rookie WRs had their Adjusted FP/G finish within one
point of their regular FP/G. This continues to show that there wasn't a strong
partiality towards late (or early) season heroics.
- If we limit this to the past seven seasons, the E/L factor actually
turns negative. For the last five years, the ratio becomes -0.25. Since
2001, the E/L factor has been -0.32. The only two players from 2003, Andre
Johnson (-0.9 E/L) and Anquan Boldin (-0.5 E/L) both classify as strong
starters.
- Does the E/L factor change if we look at where the players were drafted?
For players drafted in the first three rounds, the average rate is 0.22.
(Remember, for all players it was 0.17). First two rounds? 0.16. First
round picks? 0.22. Basically, there is very little correlation (if any)
between draft pick and how your rookies FPs are distributed. Additionally,
all players drafted in the top sixteen of the first round have an E/L
of exactly
zero.
- Rookies ranked in the top forty had a slightly larger E/L factor
of 0.34. In the top thirty, it rises again to 0.52. In the top fifteen,
there are just four players-with an E/L ratio of 0.28. A note about the
top thirty: Six of those ten receivers were from either 1995 or 1996.
Removing those rookies drops the E/L metric to 0.06.
- The last seven years, Chris Chambers was the fastest finisher-he averaged
9.7 adjusted FP/G, and 8.9 regular FP/G.
- How about this: All rookie wide receivers that:
- Were drafted in the past five seasons
o Ranked in the top forty their rookie years, and
- Were first day draft picks (top three rounds)
- They had an Early/Late factor of
0.00. Additionally, none of the
ten had an E/L factor greater than +/- 1.00.
Conclusions
- I think that last category (last five years, first three rounds, top forty)
is the most relevant to current fantasy predictions. For the important players
we profile, rookie WRs tend to be fairly consistent in their scoring. While
running backs tend to improve as the season goes on, rookie WRs don't go up
or down.
- It doesn't matter whether the rookie is a first round pick or not drafted
at all. Draft status and how fantasy production is distributed are surprisingly
unrelated.
And then what happened?
You may be wondering if how a player scored his fantasy points has any
correlation to how many he'll score the next year. In short, the answer
is essentially no. It shouldn't really bother you that Andre Johnson saw his
production dip at the end of last season; likewise, Dom Davis' strong second
half isn't a guarantee of future production. For those that are interested in
all of the analysis, keep reading. For the rest of you, just know that rookies
tend to be similar to veterans in at least one respect: When they scored their
fantasy points isn't anywhere near as important as how many they scored.
I thought the best way to test this theory would be to correlate two numbers:
a player's E/L factor and their change in FP/G in year two. These two numbers
capture exactly what we're trying to see: Does a strong finish lead to improved
play the next year (or vice versa)?
Of the thirty-two RBs in the study, twenty-eight of them had follow-up seasons.
Domanick Davis hadn't yet had his, while Dominic Rhodes, Jamal Lewis and Robert
Edwards missed the next year due to injury.
The correlation coefficient of E/L Factor and change in FP/G was -0.13. For
those not familiar with that statistic, a negative correlation means that as
the E/L factor increases (goes from strong starter to faster finisher), their
FP/G in year N+1 decrease. However, -0.13 is essentially statistically insignificant.
I think a more accurate study would be done if we tweak our data set to:
- Rookies from 1998-2002
- That started at least eight games the next year
This removes Olandis Gary (one game in 1999), Mike Anderson (seven games started
in 2001), Ron Dayne (7 games started in 2001) and Travis Prentice (zero games
started in 2001).
The remaining eleven RBs:
- Had an average E/L factor of 1.17.
- Averaged 12.5 FP/G their rookie year; 13.67 Adjusted FP/G
- Averaged 14.2 FP/G their sophomore years.
The correlation between these RBs E/L factors and their N+1 production was
-0.21. Statistically speaking, that's still not worth much. There are only eleven
people in the sample, and -0.21 isn't very strong. However, a negative correlation
means that the stronger you finish, the less likely you are to improve
in the following year. Exercise extreme caution with this: Travis Henry for
instance, had the third lowest E/L factor-and he improved by more than anyone
but Ricky Williams. While at first glance that might make you think fast starters
are more apt to improve next season, Henry actually had a positive E/L factor.
Remember, nine of the eleven players were fast finishers-Henry was just a little
slower than most.
Conclusion
- How a rookie RB distributes his FPs in one year doesn't mean anything in
regards to how many points he'll score the next season.
And then what happened? Part 2
Once again, let's look to the following year to see if any conclusions can
be drawn for wide receivers. Of the thirty receivers, twenty-seven have
had a follow up year to date (Sylvester Morris, Anquan Boldin, Andre Johnson
have not). Those twenty-seven receivers had a correlation (comparing E/L factor
to change in FP/G) of 0.12-which borders on being statistically irrelevant.
In the RB comparison, we edited the list to only include players that started
at least eight games the following year. Unlike RBs, wide receivers have much
better chances to play well even if they're not starting. Instead of using games
played to pare down our list, we're going to only include:
- Rookies from 1998-2002
- Rookies that averaged at least six points per game
That leaves us with eleven players. They
- Had an average E/L factor of 0.28
- Averaged 8.14 FP/G their rookie years
- Averaged 7.4 FP/G their second years
On average, they declined by 0.74 FP/G. It's become well known in the fantasy
community that rookie receivers that perform well tend to slip the following
year. The correlation coefficient between the Early/Late factor and difference
in points per game in year N and year N+ 1 is just 0.14-not very significant.
If it says anything, it says that players that are fast finishers might appear
to be a bit more likely to hold value better the next year. It's interesting
that this is the opposite of what the RB data said. For what it's worth, if
you raise the limit to averaging at least seven points per game, the coefficient
increases to 0.54. However, with just seven players in our sample size, I'm
not sure a ton can be drawn from that.
Conclusions
- How a rookie WR distributes his FPs in one year doesn't really mean much
in regards to how many points he'll score the next season.
Final Conclusions
- WRs had an average Adjusted FP/G of 7.74, compared to their average Regular
FP/G of 7.57. That represented a 2.24% increase.
- RBs had an average Adjusted FP/G of 12.85, and a Regular FP/G of 11.97.
That's an increase of 7.35%.
- Those two factors show that productive rookie RBs are significantly
faster finishers than their WR counterparts. Somewhat ironic, since RBs
adjust quicker to the pro game, while receivers generally take at least
a couple of years before they excel.
- Take a lot of the N+1 data with a grain of salt. There were small sample
sizes and not terribly overwhelming results. I just included it because I'm
sure several people would have e-mailed me asking if the data meant anything
for the following year. As far as I can tell, it really doesn't.
- Anquan Boldin and Andre Johnson really declined as the season went
on. Domanick Davis really improved as the season went on. Neither probably
means much for 2004.
- Don't give up on Kevin or Julius Jones this year-it's possible that they'll
be "eased" into the offense. Additionally, players like Stephen
Jackson, Chris Perry and especially Tatum Bell might not represent a ton of
value early on, but could get more touches later in the season. Obviously
for Jackson, that would assume a healthy Faulk to start 2004.
- The rookie WRs this year appear to be extremely talented. Reggie Williams,
Rashaun Woods and Lee Evans appear to be in situations where they might start.
Larry Fitzgerald and Roy Williams are great talents that look to start immediately.
However, history shows not to wait too long for the rookies to show signs
of life. Generally speaking, if you're not seeing something out of your rookie
WR in the first half of the season, they're not going to turn it on in the
second half. In the last seven years, only Chris Chambers has been able to
start off real slow but still hold fantasy value.
- Don't forget an important fact about this study: This looks at the top
rookies after the fact. Someone like Thomas Jones disappointed his
rookie year, while Mike Anderson outperformed expectations by about 1400 yards.
This study only looks at the top players once we know our rankings-thus, holding
onto a bum RB won't magically turn him into gold.
If you have any questions or comments about this article, please shoot me an
e-mail at [email protected].
Incredible thanks, as always, goes to Doug Drinen. The owner of www.pro-football-reference.com
provided all the statistics you've seen today. Without his work or that site,
this article simply could not have been written.
|