Boyd's World-> Breadcrumbs Back to Omaha-> A Different Perspective About the author, Boyd Nation

A Different Perspective

Publication Date: January 17, 2006

Standing on the Footstools, Anyway

There was a Robin Williams movie from the late '80's called The Dead Poets Society. It was one of those things that felt significant at the time but has turned out to be less so as time went on -- it probably just hit me at a time when I was sliding into adulthood and trying to figure out what that meant. Anyway, buried in all the existential teen angst was one useful scene -- Williams' unconventional teacher character has his students stand on top of their desks, because it's a useful thing to see things from a different perspective.

I've been playing around with a new rating system. It's not purely, or even mainly, my own invention -- the core ideas are creditable to Clay Davenport over at Baseball Prospectus. The general idea is called Third Order Winning Percentage. I haven't followed his methodology perfectly, so any flaws are mine and not his, but here's the general gist:

There are inconsistencies in the relationship between the component parts of a team's performance, the parts that they have some control over, and the actual outcomes of their games. Hits don't always turn into runs; runs don't spread out evenly over games; opponents don't behave the way they're expected to. For the most part, we call this "life" and go on about our business. However, it's sometimes useful to know how well a team has actually performed to a greater depth than their actual game record can tell us, and that's what Third Order Winning Percentage (and my version, which I guess I'll call College Third Order) tries to do.

Take a team's raw component stats, pick one of the many formulas which converts those stats into a projection of runs scored or allowed (I used one of the variants of runs created), and project their winning percentage based on a Pythagorean projection (Davenport computes the optimum exponent on the fly; I just use 1.9). Then take those second-order percentages to compute a strength of schedule (or two; I did separate ones for offense and defense) and redo the Pythagorean projections to get the Third Order Percentage. That gives you a rating that should represent the team's most likely winning percentage against the schedule they played given their raw offensive and defensive stats.

The results are interesting, for reasons I'll go into after the list. Here are the top 50 for the 2005 season:

 #  Team                       CTO

  1 Oregon State              0.833
  2 Texas                     0.830
  3 Nebraska                  0.822
  4 Rice                      0.811
  5 Tennessee                 0.808
  6 Mississippi               0.792
  7 Cal State Fullerton       0.789
  8 Arizona                   0.781
  9 Miami, Florida            0.778
 10 Oklahoma State            0.776
 11 Tulane                    0.775
 12 Virginia                  0.772
 13 Georgia Tech              0.764
 14 North Carolina            0.762
 15 Louisiana-Lafayette       0.751
 16 Baylor                    0.749
 17 Louisiana State           0.746
 18 Florida                   0.746
 19 Arkansas                  0.743
 20 St. John's                0.741
 21 Winthrop                  0.737
 22 Long Beach State          0.736
 23 Missouri                  0.736
 24 College of Charleston     0.734
 25 Vanderbilt                0.730
 26 Clemson                   0.725
 27 Auburn                    0.724
 28 Miami, Ohio               0.723
 29 Texas Christian           0.721
 30 Stanford                  0.720
 31 Wichita State             0.715
 32 Arizona State             0.714
 33 Florida State             0.714
 34 Boston College            0.713
 35 Central Florida           0.709
 36 Oral Roberts              0.703
 37 North Carolina-Wilmington 0.702
 38 South Carolina            0.701
 39 Washington                0.695
 40 Kansas State              0.690
 41 Alabama                   0.690
 42 Texas A&M                 0.689
 43 Michigan                  0.688
 44 Southern California       0.687
 45 Army                      0.681
 46 San Francisco             0.680
 47 Dallas Baptist            0.680
 48 Ohio State                0.678
 49 Central Michigan          0.677
 50 Rhode Island              0.673

Now, cross-referencing this with the end-of-season ISR's gives some interesting insights. Most teams stay around the same place, but the few that move show the potential insights that can be gained by climbing up on that desk and looking from a different vantage point. The ISR's attempt to measure what a team has done, purely at the game result level. The CTO's attempt to measure how good a team was. Now, if you're trying to see how good a season a team has had or trying to decide if they should be rewarded with postseason play, you want to know what they've done. If you're trying to decide what they're likely to do in the future, either during the rest of the season or the next year, it's quite possible that looking at how good they are is a better tack.

Just to pick a relevant data point, let's consider the 2005 Tennessee Volunteers. They were beaten up mercilessly by the ISR's, finishing the season at #24 despite making it to Omaha. In large part, that's because they had a distressing tendency to take days off -- losing to Furman, Belmont, Bethune-Cookman, and UNC-Asheville in the same year will hurt you in game-result-based systems rather badly. On the other hand, the component pieces were quite good statistically, and they finish at #5 in the CTO's, which was reflected by their reasonable level of success in the postseason.

Now, Tennessee actually scored and allowed about the number of runs that you'd expect from their component stats; they just clustered them badly in some games that really hurt them to lose.

I don't plan to replace the ISR's yet as my primary source of ranking guidance, but I do think the CTO's are worth having around to further the discussion, especially around tournament analysis time, so I'm going to add them to the list of weekly (actually, those may not be weekly this year; let me know how daily reports would go over with you) reports for this year.

Get Your Specifications Here

For those of you who offered to help with the data gathering project, or who didn't speak up but want to help, the specifications for the best way to help are up now.

If you're interested in reprinting this or any other Boyd's World material for your publication or Web site, please read the reprint policy and contact me

Google

Boyd's World-> Breadcrumbs Back to Omaha-> A Different Perspective About the author, Boyd Nation