Who’s going to make the expanded College Football Playoff? What’s the best price on the national championship? Now you can start to answer those questions with Unabated’s College Football Futures Simulator.
Designed by Rufus Peabody and Cade Massey – the duo behind the Massey-Peabody ratings – along with the Unabated data science and engineering teams, the simulator uses pre-loaded or custom power ratings to model out the coming season.
As with any simulation work, it all starts with the ratings you bring to the table.
Premium members can use the pre-loaded Massey-Peabody ratings currently. Or you can use your own, whether you’re creating your own numbers or bringing them in from another trusted source.
“I typically use a blend of ratings, in order to regress Massey-Peabody numbers toward the mean,” Peabody said. “ESPN’s FPI, Bill Connelly’s SP+, and TeamRankings all have free power ratings available that you can use.”
(And if you need a refresher on how to blend ratings, we can help you out with that.)
Using the College Football Futures Simulator
When you choose to upload your own, our template includes all FBS and FCS teams. The default rating for FCS teams is -17.5, but you can update that to your specifications if you have a strong handle on the FCS landscape.
You can also edit the home-field advantage value on a team-by-team level.
Once you’ve put together your ratings, hit the big green “Simulate” button and let the Monte Carlo games begin.
You’ll see a list of probabilities for each team to make the CFP, win their conference, win the national championship, and their projected number of wins.
When you expand the menu for a team’s win total, you’ll get the full distribution of wins for that team, along with fair prices for each half win based on your power rating.
Change over to “View in Futures” mode to see how your projected fair prices stack up to current numbers across four markets. Bets that show a positive expectation are highlighted in green.
You can also combine multiple markets into one view. Use the search bar to see a snapshot of how any single school is positioned across books and markets.
How the College Football Futures Simulator Works
The simulator starts by using your ratings to derive an implied point spread for each matchup. If, for example, you have Alabama rated at 22 playing at Georgia, rated 28, you’d add your home field advantage into the equation.
If you were using 2.5 for home field, the point spread implied by your ratings would be (28 + 2.5) – 26, or UGA -4.5. We then convert that implied spread to a probability of winning.
“Since power ratings are point-agnostic (each point is worth the same), unlike point spreads, we use a different, more continuous approach,” Peabody said. “Using the same trading tools we have on Unabated does not work.”
The simulator generates a random number between zero and one. A team “wins” if the number is less than the probability of the implied spread.
In our example, if Georgia -4.5 worked out to a 65 percent implied probability of winning and the random number generator returns .649 for the game, the sim credits the Bulldogs with a win. If it comes up .651, they get a loss.
After each simulated week, we update each team’s power rating based on simulated data. We then run 10,000 iterations of the sim.
Not All Wins Are Created Equal
“If a 7-point favorite wins by 14, we expect their power rating to improve a bit,” Peabody said. “We can actually estimate an exact amount the team’s rating should move, based on what I call simulated surplus (simulated score differential minus expected score differential) and the number of games the team has played. Ratings update more aggressively early in the season, when we have less information on a team.”
There’s uncertainty baked into the process as well. A team that wins by 14 in real life may not have dominated the game. Maybe they were the beneficiaries of great turnover luck.
To account for that, the rating update function simulates from a distribution of possible ratings movement based on that simulated surplus.
Say, for example, the median move for a 7-point favorite coming off a 14-point win was to increase their power rating by one. But there’s a distribution for that. Some amount of the time it may be zero, or sometimes two, or more. We simulate the uncertainty around those outcomes in our process.
We also account for rating adjustments for teams coming off bye weeks. Sometimes learning more about a team’s previous opponents makes us re-evaluate our opinion of the team in question. These moves tend to be smaller, though.
This process is then repeated for every week in the season. We determine conference championship matchups and winners based on our simulated standings and tiebreaker rules.
Simulating the College Football Playoff
Just because the sim says team is the favorite to win the SEC doesn’t mean we’re done here.
The real fun, especially in 2024 when there’s less market certainty in the first year, comes with the expanded College Football Playoff.
The 13-member committee ranks the teams. The top five conference champions and the next seven highest-ranked teams are the ones that go to the playoffs. You can see where subjectivity creeps into the process. But we can account for that.
“We have a rich history of committee behavior over the last decade,” Peabody said. “These are humans making decisions, so they’re not perfectly consistent or logical all the time, but Cade Massey and I still built a model to try to predict how the committee will assess the strength of each team.”
The model uses the things the committee talks about when they talk about how they assess each team: strength of schedule, whether a team won its conference, and how many losses they have. It also incorporates the simulated end-of-season power rating to predict a logit function. We generate our committee rankings using that function.
“Because there is (again) uncertainty — our model wasn’t perfect at predicting the committee’s behavior in the past — we have to sample from a distribution of possible logit values and then order teams based on the simulated logits,” Peabody said.
Once we’ve determined which teams are in at the end of our simulated season, it’s back to basics. We simulate individual 12-vs-5 or 11-vs-6 matches or what have you, all the way through to the end of the postseason.
Building for Reality
At the end of the day, if you’re going to put real money into play based on these simulations, you need to be able to trust the tool. Why should you trust the College Football Futures Simulator?
“For any simulation to be worth it’s salt, it needs to model reality,” Peabody said. “That means being properly calibrated both in terms of magnitude and variance in rating change, both on a week-by-week level and a season-long level.”
If you want to check the College Football Futures Simulator out, you can schedule a free one-on-one demo and get a five-day trial of Unabated Premium.
