It's fantasy football season! And just in time for the frenzy, Assistant Professor of Operations Management Michael Fry and MSQA student Andrew Lundberg published their results in the Journal of Quantitative Analysis in Sports. “A Player Selection Heuristic for a Sports League Draft” was based on work conducted at the University of Cincinnati, along with co-author Jeffrey Ohlmann from the University of Iowa.
Sports leagues conduct new player entry drafts in which franchises select, in a pre-determined order, players to complement their existing rosters. We model the decision-making process of a single sports franchise during a player selection draft. The basic premise of our model is that a team selects a particular player based on a combination of the player's estimated value, the value of the other players currently available, and the team's need at each position.
“Our results are robust,” says Fry. “It amounts to this: you have a set of choices that people can make. They all want the best player available and sometimes people just go for that player, regardless of what they really need. All you really want to know — in fantasy and in real drafts — is what set of players is not going to be available when your turn comes up.”
Football presents a good opportunity for such research because — unlike basketball, for example — players generally have set positions such as running back, quarterback or defensive lineman. Basketball players, on the other hand, often play multiple positions, which makes it harder to identify exact team needs going into a draft. Baseball presents its own set of challenges because there is huge uncertainty about how a drafted player will affect an actual team that season. Baseball players often spend many years in the minors and it is very difficult to tell how a player drafted now will affect your team in the future.
“Hence, baseball drafts are about 50 rounds long while NFL is 7 and basketball is 2,” says Fry. “Although baseball would be interesting because it is more statistically tracked.”
Mike Fry is a fantasy football fan’s best friend. In one conversation, he can talk “heuristic,” “tractable deterministic dynamic program” and “sequential decision-making” — then tosses out “league-wide passing stats” and “detailed injury reports.” Fry is one of “Cincinnati’s Numbers Crew” in the Department of Quantitative Analysis and Operations Management, featured in UC Research magazine (Summer 2007).
Fry’s team approached the idea in two phases. He’s a fantasy leaguer himself, so he understands the routine.
“First we decide who the good players are, the same way players are evaluated in real life by scouts,” Fry says. “Then, like sports teams’ general managers, we sit down and have a draft. We pick the player you want, and you look at who you need. You might want a cornerback, but you also need a linebacker.”
Fry says that every decision affects other actions down the road, so it is not humanly possible to anticipate all the ramifications and keep track of these decisions.
His advice for fantasy fans?
"Using commonly available fantasy football player evaluations for the 2007 season we have noticed a few trends from our model’s output:
A traditional maxim in fantasy football is to draft two running backs in the first two rounds because RBs are seen as being the most valuable. However, this year our model generally recommends against this strategy. There are only a few (really two) sure-fire RBs and then a deep crop of 2nd-tier RBs. Thus, our model often recommends taking a RB first, then taking a top QB or WR next and coming back for your 2nd RB later.
It can often be a good decision to draft one of the top defenses (generally the Chicago Bears or Baltimore Ravens) earlier than most people would suggest. If you miss out on these top defenses, then you might as well wait until the very end of the draft as there is too much uncertainty in how other defenses will perform. The same is often true for kickers.
But the main point is that if you feed garbage into the model, you get garbage out. If you are convinced, for whatever misguided reason, that the Cleveland Browns QB (whoever that turns out to be) is going to be a star and you rank him first, then guess what, the model is going to tell you to draft him very high.
"That’s not the model’s fault," Fry points out, "that’s user error."
But this is all fantasy — pretend. What difference does it make, I mean, really?
“The fantasy market alone is a billion-dollar industry,” says Fry, “especially with the Internet.”
OK, so Fry knows the numbers. How’s his game when it comes to fantasy play? He’s won his league once and made the playoffs seven times in nine years of playing. And his competition is no small peanuts.
“We have eight PhDs in our group,” he says, laughing. “And one NASA rocket scientist.”
Cite This Page: