8/3/09
Fantasy football analysis of players tends to focus on the total
fantasy points scored. This week we dig into historical performance
in more detail than a simple analysis of total fantasy points scored
can do. Our focus is on the QB position.
The quarterback is an interesting position in fantasy football.
Most league structures involve a team having to start exactly one
of them. And they are also one of the most likely positions to earn
you negative fantasy points in a game. The volatility of the starting
QB from week to week impacts the shape a fantasy team takes.
This week we investigate the value of consistency at QB and highlight
a few QB’s who stand out as being potentially more (or less)
valuable than their fatnasy points might indicate based on their
consistency.
We start with background on our idea and approach. We then include
relevant historical statistics. Then we step back and evaluate some
specific QB’s based on their historical performance with the
goal of providing perspective that might be useful to you on draft
day when targeting a QB.
The Idea
Typical analysis of past player results focuses on the total fantasy
points scored. This is obviously an important number in evaluating
a player in a fantasy context. In practice, a fantasy football manager
is impacted not only by the total points scored, but also by how
those points are distributed from game to game - consistency. Let’s
look at an extreme illustrative example.
Let’s say there are two QB’s that score 200 fantasy
points in a season. QB #1 scores an amazing 340 points in week one
and then scores negative 10 points for his remaining 14 fantasy
games for a total of 200. QB #2 also scores 200 total points for
the season, but does it by scoring about 13 points per game.
QB #1 is more useful in week 1, but is useless in weeks 2 through
16. QB #2 is less useful in week 1, but is more useful in every
other week. Although their total points scored are the same, many
would agree that the distribution of those points makes QB#1 useless
for a fantasy season and QB #2 very useful.
This extreme illustrative example provides some context for the
direction we will be going. We will dig deeper into the historical
statistics of fantasy QB’s to see if any stand out as being
more or less valuable than their historical totals alone might imply.
Scope of Our Historical Analysis
We focus on games in weeks 1-16 of the regular season over the last
four years. Some QB’s have played in more games over those
4 seasons than others. Some have been the starter for the full length
of time, some have always been the starter but have only played
for a season or two, and some used to be a backup getting some snaps
at the end of games and only recently moved into a starting role.
We want to compare all of these situations on as level a playing
field as we can.
With that in mind, we focus on QB performances in which a fantasy
manager might have considered starting them at the start of the
game.
Although Tom Brady had a bad first game of the season in 2008-09
due to injury, a fantasy owner would have considered starting him,
so we let this game’s statistics count. Contrast this with
Matt Cassel’s performance in week 11 of 2007-08. He did not
do much, but that was because he was put in at the end of the game
when New England was up by 40+ points. Cassel played in this game,
but since no fantasy owner would consider starting him (as he was
not the starting QB) we do not consider it in our historical statistics.
Scoring System
All of our analysis focuses on a fantasy football league that starts
1 QB. The base scoring system we use is 1 point per 20 completions,
1 point per 25 yards passing, -2 points per interception, +4 points
per passing TD, 1 point per 10 yards rushing, 6 points per rushing
TD, -1 point per sack and -2 points per fumble lost.
Sample Size
The chart below shows the QB’s whose historical performance
we will investigate, along with the number of games we will consider
as part of their historical performance. We include some but not
all of the lower quality QB’s available in fantasy drafts
this season.
Recall sample size here might be less than games played, as described
above. We only consider games where a fantasy owner might have
reasonably considered starting them at the beginning of the game.
The maximum possible sample size is 60, since we focus on weeks
1 to 16, which makes 15 possible games (after the bye) per season.
Peyton Manning, Drew Brees and Eli Manning stand out as being the
only QB’s with the full sample size of 60. They have a larger
sample size than Tom Brady mainly because he got injured last year.
They have a larger sample size than Matt Ryan because Matt Ryan
joined the league more recently.
Bigger sample sizes provide more information, and we don’t
have much information on QB’s like Aaron Rodgers, Joe Flacco,
Matt Ryan, and Tarvaris Jackson. Sometimes having more information
makes it more informative, and we should keep that in mind as we
evaluate.
Total Fantasy Points
We’ll start with the usual stats, which are based on total
fantasy points scored. We want to normalize these so that someone
like Peyton Manning doesn’t come up as being 4 times more
valuable than Matt Ryan, just because he’s been in the league
for all 4 seasons – so we focus on fantasy points per game
within our sample.
None of these should come as a big surprise. It’s worth noting
Kyle Orton performed much better last year than three seasons ago.
We weigh all of the past four seasons equally and that is why Kyle
ends up at the bottom. If we weighed recent seasons more heavily,
then he would not come out in last place.
Let’s Dig Into These In More Detail
Let’s take a look into these numbers a bit more. One undesirable
characteristic of a fantasy QB is their tendency to underperform
a QB who is retired or incarcerated. With that in mind let’s
take a look at QB’s sorted by their likelihood to get you
negative points in a game. We show the total number of games a QB
has scored less than 0 points divided by the total number of games
in the sample below.
Trent Edwards, David Garrard, and Jason Campbell are a few QB’s
who stand out favorably here – they’re ranked a lot
higher in this stat than their pure fantasy points scored. Matt
Schaub stands out too – and not in a good way. Although he’s
in the middle of the pack in total points scored, he is the most
likely QB to get you negative points. Other poor performers here
(relative to their rank in total points) include Jake Delhomme and
Philip Rivers.
Let’s look at some desirable fantasy statistics and see how
the QB’s rank.
Getting 10 or more fantasy points consistently is a desirable
stat for a fantasy QB. Let’s see how QB’s stack up.
Below shows the percentage of games in our sample where the fantasy
QB scored 10 or more points.
Let’s split this into two pieces and look at likelihood to
score 10-20 and 20+.
Observations At The Top And The Bottom
The Top
Tom Brady,
Peyton Manning,
Aaron Rodgers,
Drew Brees,
and Tony
Romo look good across all stats. They score a high amount
of points per game, are unlikely to get you negative points, and
are among the most likely to get you more than 10 fantasy points
in a game. Of the stats we consider they underperform in the 10-20
FPts range because they are so likely to score more than 20 FPts
per game. These QB’s have performed well across the board over
the last few seasons. Aaron Rodgers has the smallest sample of
data available among these.
The Bottom
Kyle Orton,
Tarvaris
Jackson, Trent
Edwards, Joe
Flacco, Chad
Pennington, and Jason
Campbell finish near the bottom in total points scored per
game. Some bright spots worth mentioning:
- Trent
Edwards has never scored negative points.
- Joe
Flacco is the 3rd most likely QB to score 10-20 points.
- Jason
Campbell is tied for 8th least likely to score negative
points, and is the 12th most likely to score 10-20 points.
- Kyle
Orton performed much better last year than in prior years.
In this analysis, we apply equal weight to performances over
the past four years, and his older performances drive the poor
results.
These QB’s are at or near the unfavorable end of most stats.
Trent Edwards, Joe Flacco, Jason Campbell, and Kyle Orton have some
bright spots that stand out favorably compared to the others.
Digging Into The Middle Of The Pack –
Who Stands Out And Why
Garrard: Historically consistent.
David
Garrard is in the middle of the pack in fantasy points scored,
ranked 10th out of 22. Unlike most around him, he has never produced
negative points in a game. He is the 7th most likely QB to score
10 or more in a game, doing so 73% of the time. And he is the most
likely to score 10-20 points in a game at 61%. He also has upside
beyond that, scoring more than 20 points 12% of the time, performing
in the middle of the pack at 14th. David
Garrard stands out as being more consistent than many of the other
QB’s in the middle of the pack, while retaining some
of the upside. These are attractive characteristics for a fantasy
QB.
Matt Schaub is also in the middle of the pack in fantasy points
scored per game ranked 13th of the 22 we consider. Over the past
4 seasons, he is the most likely QB to score negative points in
a game with 14% of games turning in performances below 0. This
might be surprising to some. In terms of likelihood to score 10
or more fantasy points he is again near the bottom of the list
at 18th, and he is the second least likely QB to score 10-20 points
at 21st out of 22. The one bright spot on his resume is he is
the 7th most likely QB to score more than 20 points. You might
say that he usually performs badly, but he has spots of upside.
He reminds us in some ways of our unattractive “QB #1”
from our extreme illustrative example at the start.
Philip Rivers
is also in the middle of the pack in fantasy points scored per game,
ranked 11th out of 22. He scored negative points about 7% of the
time, placing him only at 15th best at avoiding negative points.
He’s in the middle of the pack at ability to score 10+ points placing
14th. He’s one of the least likely to score 10-20 at 18th, but he’s
9th most likely to score 20+. His profile resembles a positive shift
up from Matt Schaub’s in most respects – but he shares the texture
of Schaub’s performance. That includes average overall performance,
with a lot of points of downside and upside. His upside is better
than Schaub’s historically, and his downside is less. Rivers looks
like a “better version” of Schaub when viewed in this context –
but he still historically resembles our unattractive “QB #1” described
at the start of the article.
More Details On QB Distribution Of Scores
This table shows the percentiles of scores for each QB, and sorts
it by their median. Aaron Rodgers having a 90th percentile score
of 23.3 indicates that 90% of the time he scores less than 23.3.
Donovan McNabb’s median score 14.5 indicates 50% of the
time he scores more than 14.5 and 50% of the time he scores less
than 14.5.
This table helps provide additional perspective on the distribution
of historical scores per game by QB.
Another stat that might be interesting is to simply look at the
volatility of points scored per game for each QB. Volatility is
a measure that gives you a sense of how wildly a QB’s scores
vary from their average. That is shown below.
Care should be taken in interpreting this. In particular the historically
high volatility of Tom Brady is more attractive than the historically
high volatility of Matt Schaub. And the historically low volatility
of David Garrard is more favorable than the historically low volatility
of Kyle Orton or Jason Campbell.
Conclusion
Players are often ranked based on fantasy points they did (or are
expected to) score, with minor comments accounting for “upside
potential”. This article digs into these types of minor comments
in more detail and elaborates on how two QB’s with similar
average fantasy points scored can bring two very different types
of performance to the table. One QB that stood out favorably in
this analysis was David
Garrard. Two QB’s who stood out unfavorably in this analysis
included Matt
Schaub and Philip
Rivers.
There are pluses and minuses to any QB. Have perspective on the
texture of your QB’s historical performances before you make
your choice this year. Their history sometimes gives an indication
of what their performance might be like in the future. Sometimes
it’s OK to ride a roller coaster – as long as you prepare
yourself appropriately in advance. |