7/1/03
Average Value Theory (AVT) is a tool I devised
in 2002 with original article co-writer Christopher Annunziata. Its
purpose was a simple one, to provide all the usability of a Value
Based Drafting (VBD) draft board without using player projections.
I found that a major weakness of VBD was the requirement that owners
develop individual player projections, which are -- for those of you
lucky enough not to have done them before -- season-end stat line
guesses for every draftable player in the NFL. Player projections
are at the foundation of VBD. The fantasy point totals calculated
from player projections are used (via baselines) to compare running
backs to wide receivers (or apples to oranges). Without projections,
those comparisons that have made VBD famous are not possible. Yet,
think of what is involved in creating player projections; think of
how complex that task is. Here is a single example of what might be
involved in creating player projections for a player:
- Get statistics covering the last three (or four, or five, or
two) years of that player's NFL career.
- Analyze those statistics to look for consistencies and/or trends
(are his numbers "going up"? are his numbers "going
down"? if so, how can you apply that trend mathematically
toward guesses about next year's statistics?)
- Examine off-season moves by that player's team; determine whether
those moves help or hurt his situation; attempt to translate "help"
or "hurt" into some sort of numerical value that can
be applied to guesses about his future statistics.
- Examine that player's team's schedule; determine whether it
is an easy schedule or a difficult schedule for his particular
situation; attempt to translate "easy" or "difficult"
into some sort of numerical value that can be applied to guesses
about his future statistics.
- Calculate the above list of guesses and numerical translations
into a season-end prediction of what his statistics will be.
Now that you've finished that player's projection, repeat the process
for the other 199 draftable players in your league.
You can see how that long string of guesses can add up to monumental
potential for error, consequently throwing your draft day board into
complete disarray. And, what's worse, should you at some point want
to alter your projections, there is a cascading effect that takes
place. Alter a quarterback's passing yard total, and you're going
to have to sift through all his running backs and receivers and tight
ends and somehow make the numbers add up.
The task is daunting, and not as scientific as it sounds. What it
amounts to is a chain of guesses ("help", "hurt",
"easy", "difficult", "going up", "going
down") translated into a mathematical process somehow, with the
final output being a numerical representation of how good you think
that player is relative to his peers. A numerical representation that,
unfortunately, has a strong relationship to other players on his team;
wrong about one, wrong about all. And we're all wrong about most of
them, I admit. Especially when we're attempting to pin down a player's
future statistics.
Wouldn't it be more simple to merely do what is at the heart of the
player projection process, and not get caught up in the mathematical
voodoo we practice every offseason? Why not rank players at each position,
and go from there? No, you can't do that, because you can't translate
pure rankings into a VBD board. You need projections to get the fantasy
points to establish the baselines to make the comparisons.
Well, that's where AVT comes in.
AVT takes your player rankings (Top-30 QBs, Top-40 RBs, Top-50
WRs for example) and turns that into a complete VBD draft board.
It already has all the fantasy points, it has already made all the
comparisons, it just needs the names. There are several versions
of AVT, but here is how the most simple one works:
Using historical data, AVT calculates what each player-slot will
score on average. By player slot I mean WR1, WR2, WR3, WR4, etc.
Each one of those player slots is given a fantasy point value, customized
to fit your scoring system. These values are obtained by taking
historical season-end statistics, calculating the fantasy point
totals through your scoring system, removing the names, replacing
those names with player slot ranks, doing this for several consecutive
years, and taking the average. Very simple, very clean, and standardized
while still being customized.
Let me give you an example (using Fantasy Asylum scoring):
In 2002, Marvin Harrison was WR1 (the top-ranked WR) with 239.2
fantasy points.
In 2001, Marvin Harrison was WR1 with 242.7 fantasy points.
In 2000, Randy Moss was WR1 with 234.2 fantasy points.
In AVT, the fantasy point value for WR1 will therefore be: (239.2+242.7+234.2)/3
or 238.7
No matter if you decide that Marvin Harrison or Randy Moss or Terrell
Owens will be the top wide receiver, WR1 will have the fantasy point
value of 238.7 because that's about what WR1s usually score. Did
any of our above mentioned WR1s score 238.7 points? No. But they
were all close. Close enough, some might say. And, I've found, often
closer than what projection-makers will come up with.
By replacing individual player projections with uniform player slot
averages, calculated from actual historical data, AVT is not only
far easier, but it standardizes your VBD draft board. If there is
error involved (and there will be some error on every single player,
remember, due to the nature of the calculations), at least that
error is standardized within the calculation system so as to nor
wreak havoc on the rest of your draft board. What happens if your
baseline wide receiver's projection is off by a sizable amount...
say 32 fantasy points? Forget for a moment the actual name of the
player. Instead, focus on what a baseline player means to VBD. If
your baseline wide receiver's point total is off by that much, it
is going to throw off the relational values of every wide receiver
on your board. Now, your top wide receiver's VBD value of 93 doesn't
make as much sense; he's really 32 points more valuable than that.
As a result, the relative values of all other positions will be
off. Now, WR8 isn't about even with RB22; WR8 should be closer to
RB17. That has huge consequences on draft day. You might be taking
the wrong position at the wrong time because of your faulty player
projections.
However, if you let history be your guide -- as AVT does -- you
know that all the relative and relational player values are in accord
with historical norms, no matter the presence of fantasy point error.
An AVT-enhanced VBD board does what VBD boards are intended to do.
It makes a suggestion about which position to take at any particular
time in the draft. It is a complication of VBD that it must rely
on player projections. That need makes it to easy for us to screw
things up. The AVT process is easier, is more structured, is fully
customized to your scoring system, and it will better maintain the
apple-to-orange comparisons that VBD seeks to provide.
|
