Introducing Similarity Rankings

In case my seven-part series on scatter-plotting teams' F/+ ratings didn't convince you, I am kind of obsessed with finding new ways to compare teams to one another. I'm pretty excited about this one. It's something I've had in mind for awhile. Frankly, I don't think it's a finished product just yet, so I'd really like to get your feedback on how to improve it.

Here's what I'm attempting to accomplish: Ideally, I'd like to find a way to take the "DNA" of a team, plug them into a spreadsheet, and find the 20(ish) most similar teams to them in college football over the past few seasons. I think this will be valuable for a few reasons:

• Let's say we are in week 8 of the college football season. We'll be able to take a team that we're not quite sure about, plug them into the system, and see what teams from prior seasons that have similar profiles were able to accomplish.
• With teams that have already finished their season, we will be able to compare their results to their peer group to see if they over- or under-performed.
• Finally, (this is what I'm still working on at this point), we will be able to look at an upcoming matchup and see the outcomes of previous games that featured games with similar profiles. I'll give an example here in a moment.

Here's what I included in each team's profile: Offensive F/+ rating (25% weighting), Defensive F/+ rating (25%), Run/Pass Ratio (22.5%), Plays/Minute (22.5%), Special Teams F/+ rating (5%).

Every team from 2007-12 is included in the system. This is a simplification -- if you want the details, let me know -- but teams are compared against each other by determining how many standard deviations away from the mean they are in each category.

It's easier to explain by using an example. Let's look at Kansas State's 2012 team. These numbers will mean almost nothing in isolation, but bear with me for a moment. Here's K-State's profile:

• Offensive F/+: +1.05 standard deviations above the mean
• Defensive F/+: +1.34
• Run %: +0.92
• Plays/Minute: -1.41 (K-State was a very slow team, especially for the Big 12)
• Special Teams F/+: 2.27 (they had one of the best special teams units on record)

Okay, so using the weightings I listed above, here are the 20 most similar teams to this K-State team from 2007 - 2012:

You'll notice that I listed each team's actual statistic in each category rather than the deviations from the mean figure, mainly to make this all easier to digest. So here's a group of 20 teams that have a very good offense, an excellent defense, run the ball a lot, and are S-L-O-W on offense (and I mean slow -- K-State was in the fourth percentile in plays/minute).

You may notice that I left off special teams. This is simply because the table is busy enough as it is without another column. Special teams is factored in when calculating similarity ratings, but as the least important factor it does not make the cut in the visual.

What can we learn from this list? A few things.

The average record for this group of teams is 10.3 wins and 2.9 losses (78.0% winning percentage). The best outcome on the list was 2010 TCU at 13-0, winners of the Rose Bowl. The worst team is 2012 Wisconsin at 8-6.

I found it interesting that four out of the top six on the list (including K-State) lost BCS bowl games. Is there something about this profile that makes teams unsuccessful against other really good teams?

I guess you could say K-State over-achieved their peer group, but just barely. 11-2 is better than the 10-3 average of their peer group.

Here's where I would like to go with this project: we could have looked at this list prior to the Fiesta Bowl, when K-State was matched up with Oregon's up-tempo offense. We could have looked at how teams similar to K-State did against teams similar to Oregon. I'm interested to see what we can learn from that type of analysis.

Finally, here's where K-State and their peer group compares to BCS title winners:

Generally speaking, this group of teams is very good, but not quite good enough to win a national championship.

Thoughts? I'll come back to this in future posts. I'd be happy to run your team for you. Let me know what you think in the comments.