Wednesday, July 25, 2012

There are no Cut Scores in Sabermetrics

Imagine a scene where a major league general manager and his statistics wonk sit down to decide which of two minor league infielders to move up to their major league roster. The stats guy brings a bunch of numbers -- "multiple measures" you might say -- including a single aggregate number which compares the overall value of the two players. The GM glances at the report, says, "This guy only is only hitting .238 and I have a rule about calling up anyone with a batting average, under .240, so we'll take the other guy."

It may well be the right decision, but it kind of makes a mockery of the supposed sophistication of the process.

Or for that matter, imagine that your fantasy football commissioner announces that any team starting a quarterback who throws four interceptions automatically loses, regardless of the rest of the teams' scoring. Would you think that was a good idea?

From Zach Mezera's report on a recent City Council Education Subcommittee meeting:

One that was ID’d as “Focus” that probably caught peoples’ attention was Nathan Bishop. The primary reason for this, it seems, was that only 9% scored proficient (in what?) in the most recent NECAP. The RIDE rules state that if a school has a proficiency score of less than 10, it automatically is placed in Focus regardless of other scores.

It isn't actually 9%, but appears to be 9/30 for overall performance on the new classifications. Bishop does have an overall score slightly higher than some schools rated above "Focus," so this appears to be accurate.

I'd note that I don't really know one way or another if Bishop should be a "Focus" school.

However, I'm highly dubious of this particular rule. Let's just scroll down through middle schools and pick out some with similar overall achievement scores and see how they did overall. These will be listed by name then overall achievement/total points/classification ("Typical" is better than "Warning" is better than "Focus"):

  • Nathan Bishop: 9/51.33/Focus
  • Dr. Earl F. Calcutt: 9/50/Focus
  • Riverside: 13/46.33/Warning
  • Highlander Charter: 14/75.5/Leading
  • Nicholas A. Ferri: 13/51.5/Typical
  • Frank E. Thompson: 12/54.83/Typical
  • Goff: 11/45.17/Warning
  • Esek Hopkins: 8/50.50/Focus
  • Segue Institute for Learning: 10/53.17/Warning
  • UCAP: 12/55.71/Typical
  • Woonsocket: 11/51/Typical

There are a lot of schools that were on the bubble, some with lower overall scores than Bishop. I think Highlander is particularly telling though because they demonstrate that a school with an overall performance score under 50% and within 5 points of Bishop can still get one of the highest total scores in the new classification system.

And piling on top this arbitrary cutoff, the overall performance score is itself based on numbers determined by cut scores. It looks like Bishop lost at least 1 point by coming in one percentage point under the 45% cut score in 7th grade reading. That's one kid tipping the scales to trigger identification.

No comments: