AI to Advise NFL on What to Pay Players
AI to Advise NFL on What to Pay Players
Company owned by broadcaster, ex-player Cris Collinsworth
has found that some teams are overpaying star running backs
Cris Collinsworth, a former Cincinnati
Bengals player who is now an analyst for NBC’s ‘Sunday Night Football,’ took
majority ownership of the analytics company in 2014. PHOTO: JONATHAN
ALCORN/REUTERS
By John Murawski July 22, 2019 5:30 am ET | WSJ PRO
An analytics company owned by former pro football player and
current broadcaster Cris Collinsworth is testing an artificial-intelligence
system to generate recommendations on player salaries for all 32 National
Football League teams.
Early results show that some teams are overpaying star running
backs, among other players, by millions of dollars.
Cincinnati-based PFFA Acquisition LLC, which does business as Pro
Football Focus, sells data on every player’s performance in every game,
presented in a searchable format with accompanying videos, to every NFL team
and to more than 60 college programs. Teams can use the data service, also
called Pro Football Focus, to analyze plays, prepare for games and recruit
players. Agents, scouts and sports media also use the system.
“We’re building tools where you can plug in players, you can plug
in salaries, you can plug in our grades, and decide whether or not a player is
worth it,” said Mr. Collinsworth, PFF’s chief executive, who took majority
ownership of the company in 2014. Mr. Collinsworth, a wide receiver for the
Cincinnati Bengals in the 1980s, is an analyst for NBC’s “Sunday Night
Football.”
Mr. Collinsworth said pro football is becoming “a war of the
mathematicians,” much like baseball before it. Teams are looking for every
possible advantage and hiring data scientists.
The new salary component PFF is developing, expected to be
released within a year, builds on the company’s data repository. Its
machine-learning algorithms will give teams a player’s market value based on an
analysis of historical data and projected performance. Metrics include standard
statistics such as the number of times a receiver drops a pass considered
catchable and the difficulty level of completed passes by a quarterback.
Eric Eager, a senior data scientist at PFF, said the company’s
analysis indicates that running backs tend to be overrated and overpaid because
they are a focal point on the field. But running backs are relatively easy to
replace because there are many of them and their performance is largely
dependent on other players.
A spokesman for the NFL Players Association, which represents
players in labor talks, said the current salary-negotiation system is
effective.
The system is being developed in two phases. The first part of the
system, which has been provided to all 32 teams for feedback, uses regression
models to weight a player’s historical performance in each game and to project
how that player would play against new opponents under various hypothetical
conditions. Such a condition could be a player’s projected performance in a
third down and 15-yards-to-go situation against an opponent he has never faced,
in the last 2 minutes of the game. This analysis allows PFF to predict how
valuable a player is.
The second phase,
which the company is testing internally, overlays a financial value to that
metric for that particular position. The salary recommendation requires
analyzing the difficulty of filling a particular position on the team. It also
requires factoring in the diminishing returns on paying a premium for some
positions, for example, an above-average offensive lineman, when that money
would be better spent on another player in a key position.
Comments
Post a Comment