Goodbye, Guesswork A team of Ph.D.'s says it has unlocked the secret of NFL draft success

June 06, 2004

There's no better illustration of the vagaries of the NFL draft
than Ryan Leaf, the No. 2 pick in 1998. The big, rangy Washington
State star looked like a can't-miss prospect, but he went to
pieces in the pros, throwing interceptions and temper tantrums
with alarming regularity before retiring in 2002. While his
flameout was conspicuous, it was hardly unusual. A study of the
draft from 1990 to 2000 by a team of analytical Ph.D.'s found
that the success rate of teams--defined as a player taken in the
appropriate round, given his subsequent NFL performance--was a
little over 20%.

The study is the work of scientists from Can He Play Corp., a
company formed by experts in such fields as statistics,
biomechanics and psychology that for five years has been
developing a mathematical model to predict NFL performances of
college football players. CHP's Jeff Brazell, a Ph.D. with
expertise in forecasting, says the formula is "horribly complex"
(the company website describes it as "a combination of Bayesian
mixture fusion technology with dynamic weighting and adaptive
optimization algorithms in a threshold performance framework"),
but in essence CHP evaluates the career of every player drafted
since 1990 and retroactively figures where he should have been
drafted. It then extrapolates from that data--and the
characteristics of each of those players--to determine the
combinations of factors that lead to success or failure.

The CHP model evaluates more than 50 variables, from objective
measurements such as 40-yard-dash times and Wonderlic test
scores, to subjective evaluations of players by scouts (which are
quantified). CHP contends that its model identifies legitimate
first-rounders more than half the time and is even better at
pinpointing players who won't pan out--for quarterbacks, Brazell
says, the model tags future Ryan Leafs with up to 80% accuracy:
"Right now the [NFL] decision makers seem to be looking at some
of the right information, but they don't weigh, filter and
integrate it in such a way that it's giving them a good
prediction."

A handful of teams have expressed interest in CHP's data, and one
unnamed club used it this year. If the model is right, a couple
of teams that went with wideouts in this Year of the Receiver may
not have chosen wisely--CHP predicts that while Larry Fitzgerald
(No. 3 overall), Roy Williams (7th) and Rashaun Woods (31st) will
justify their first-round status, Reggie Williams (9th) and
Michael Clayton (15th) will be good but not spectacular.

Will this gridiron version of Moneyball--the team-building
strategy employed in baseball by Oakland A's G.M. Billy
Beane--catch on? Former Pro Bowl tight end Todd Christensen, a
CHP adviser, notes one crucial difference between the major
leagues and the NFL. "Economically, Billy Beane was forced to
change," says Christensen. "In the NFL you don't have that.
Everybody's in the black. There's no incentive. This program is
for teams that are willing to be dynamic." CHP's predictive
success may be hard to ignore, though. "We're still wrong some of
the time," says Brazell. "But the value of not making the Ryan
Leaf mistake is gigantic."

COLOR PHOTO: MATT BROWN/NEWSPORT CATCH CAN? Reggie Williams went ninth overall in April--too high,according to the Can He Play mathematical model.

HOLE YARDS PAR R1 R2 R3 R4
OUT
HOLE YARDS PAR R1 R2 R3 R4
IN
Eagle (-2)
Birdie (-1)
Bogey (+1)
Double Bogey (+2)