PDGA embarks on statistical evaluation of courses

[Lund]

"Is there a numerical value that can be calculated to objectively measure the "goodness" of the courses used to determine winners not only in Open but for each division regardless of skill level?"

Stat nerds can download the full .pdf of their methods and discussion from pdga website here.

[Jason Philips]

Just when I thought I was going to get any real work done today. Thanks a lot. Ha!

[jeverett]

For any non-stat-nerds, the document basically uses the 'Pearson Correlation Coefficient' calculation, and compares the ratings of all players going into an event with the sum of their round ratings coming out of the event. i.e. Did the highest-rating player get the highest-rated rounds overall? How closely did the event results fall in line with the ratings of the players involved?

The correlation coefficient ranges from (theoretically) zero to one, with a value of 1.00 meaning that the results of an event were 100% predicted by the ratings of the players going into the event. While I imagine a value of 1.00 being neither possible (due to the inherent randomness of the sport) nor entirely desirable, as a goal getting close to that value does sound beneficial. The document calculated an estimated average correlation coefficient of ~0.62, but found that some of the courses used for majors and NT events had coefficients as high as 0.91 (Highbridge Gold), while a few others had *very* low coefficients.

tl:dr - reducing randomness via course design increases the effect of player skill/rating on the observed distribution of ratings for an event vs. the initial rating of all players involved in the event.