division 3 football computer rankings
division 3 football computer rankings
division 3 football computer rankings
division 3 football computer rankings
division 3 football computer rankings

My Favorite Sports Blogs

Playoff Projections

November 1, 2016

Projecting which teams will make the playoffs in DIII is pretty tricky. There's a lot of factors to consider, such as teams' regional rankings, win-loss record (overall, versus DIII opponents, and versus other regionally-ranked teams), strength of schedule (according to the NCAA, which does a terrible job of approximating SoS), as well as the human factor of the voters in the committee. The systematic approach taken by d3football.com to project the bracket has done a pretty good job of picking the teams who receive at-large bids, but sometimes the national committee does some counterintuitive things in picking teams for the field of 32 (2015 Ohio Northern anyone?), so I decided to take a probablistic approach to projecting my bracket. Here's how I did it (and here's where you can click to skip the methodological details).

 

Step 1 - Gather information on previous playoff participants (2015-2005).

 

Shoutout to Keith McMillan from d3football.com for doing most of the legwork on this. He provided me with a list of his that had every team, and which pool they got in through, dating back to 1999--the first season of the Pool format. I was only concerned with seasons back to 2005, which was the first season with 32 teams in the tournament. The only work I had to do was to double-check some of the Pools for teams in 2005 and 2006. The results are below:

 

Step 2 - Decide which teams are even "eligible" for Pool C consideration (bubble watch)

 

There are a few things the committee has never done, or has stated they won't do again. They have never selected a 3-loss team (overall record, not just DIII), they have never selected more than three teams from the same region, they have only once selected a team that finished third in their conference, and they have only once excluded a region from having a Pool C team. The two exceptions from above were both in 2007, when the third place team from the E8 got in through Pool C, and the South region didn't have a single Pool C team (but they did have two Pool B teams and two more automatic bids than they do now, so I think that explains why). The committee chair a few years ago said in no uncertain terms that "the days of three teams from a conference getting in are over."

 

With these exceptions used to eliminate teams from the bubble watch, I was able to limit the number of Pool C-eligible teams to anywhere from 20-30 teams depending on the season.

 

Step 3 - Compare teams left on the bubble to Pool C selections

 

I didn't have a very robust method for how I compared teams. I basically used the spaghetti approach--just throwing things at the wall until something stuck--so I'll spare you the details and just tell you the results. The four factors that most closely correlated with a team's odds of being selected for Pool C were:

 - A team's total losses

 - Winning percentage relative to other bubble teams in their region

 - NCAA SOS

 - Conference rating

 

By taking the geometric mean of these values (with some slight re-working of the numbers so that low values were good, and bad values were bad), I was able to rank teams based on their likelihood of being selected for the tournament (a "Bubble Rank"). In every year since 2005, this method for ranking teams would only miss on 1 or 2 teams.

 

Step 4 - Assign a probability of being selected based on a team's Bubble Rank

 

Based on how frequently a team with a given Bubble Rank was selected since 2005, I developed a smoothed fit to assign each rank a probability of being selected for Pool C with 6 bids available. It works out so that the team with the best bubble rank has about a 90% chance of being selected, the 6th ranked team has about a 50% chance, down to about 1% for the 15th ranked team.

 

Step 5 - Run simulation and tally results

 

I've been running simulations all season to project Pool A probabilities, and all I had to do was include a few extra columns in Excel to tabulate my results for Pool C (with UMHB likely to remain undefeated, Pool B is a forgone conclusion). You can view the results for every team here.

 

Step 6 - Estimate teams' odds of advancing through the playoffs

 

Because I'm not a great programmer, and because guessing matchups is exponentially harder than guessing selections, I took some major shortcuts here. I assumed that every game would be played on a neutral field, and that every team got a perfectly average draw. Once the actual bracket comes out, it will be much easier to factor in home-field advantages and likely opponents. The projections as they are now are meant as just a rough benchmark, so they don't really tell us much more than what we already knew - one of the teams in the Top 5 of the polls are probably going to win the Stagg Bowl (~5-in-6 chance), and nobody from the East is a real contender (3-in-1,000).

 

 

Please reload

This site was designed with the
.com
website builder. Create your website today.
Start Now