First, we’ll need a way to determine which teams would have made the playoff each year. Unfortunately, over the first four years of the actual playoff’s existence, neither the AP poll nor our Elo ratings (which are designed, in part, to predict the playoff selection committee’s tendencies) have completely nailed the playoff field with their four highest-ranked teams going into the bowls. But a combination of both1 has been a perfect 16 for 16 in terms of predicting the real-life playoff teams.
So we’ll use that Elo/AP combo to pick the four playoff teams in each historical season.2 (Our Elo ratings can be calculated going back to the 1988 season, so that’s when our hypothetical exercise will begin.) I also found that, once the playoff field is set, the pre-bowl AP rankings alone have done the best job of matching the committee’s seeding for the teams, so we’ll set the seeds that way in our mythical playoffs.
Next, we’ll need a way to play out the theoretical playoff games themselves. For that, we’ll use Elo, which provides a probabilistic forecast for any given game based on the two teams’ pregame ratings. In most cases, we’ll use each team’s pre-bowl Elo ratings to give us the chances of each team winning both its semifinal game and the championship game (conditional on making it that far). The only exception is when a slated matchup happened in a real-life bowl that season, in which case we’ll use the actual result for that semifinal or final matchup.
538
