Originally Posted by Script
I'm gonna chip in...
Sugaw is a great example as to why the ratings system is not as accurate as one would think. When the clash came, lower rated players played Sugaw. The lowest scores were 2 people with 50's (rating came out to around 980ish)
When the highest rated pros play there, they don't shoot in the mid to low 40s. It just doesn't happen. MJs handicap at this course is around a 47 and he plays at a 1030ish rating. It just doesn't make sense that 3 strokes is near a 50 point swing. Also, look at Worlds and previous tournaments there. A 49 is always a 1000 rated round, BUT NOT AT THE CLASH. You can blame whether all you want, but a few degrees difference doesn't change this course. You either play it well, or play it poorly. Hardly is there an in between.
In my opinion, one significant problem is the 'compression' formulas the PDGA uses to convert SSA into rating-point-per-throw: the formulas the PDGA uses are just incorrect. They use two linear formulas (based on the SSA), when instead they should be using the slope of the [probably] linear regression used to calculate the SSA itself. Just for reference, the way the 'compression' formulas work, the higher the SSA the smaller the rating-points-per-throw. So a course with a very low SSA (i.e. a very easy course) in the compression formula will come out with a value for rating-points-per-throw quite a bit more than 10.
From looking at data, the 'compression' formulas appear incorrect, for all events, inducing a small amount of error into all non-exactly-1000 rated rounds. For reference, I wrote up a lengthy post on the topic here previously:
Note: the post above includes a slight error with the SSA second-pass calculation.. the only rounds that should be dropped from the SSA calculation are those from players who's rounds (after the first-pass) are rated >60 points below their rating. This doesn't affect the conclusion, however: the 'compression' formulas are not accurately reflecting the relationship between player (initial) rating and scoring spread for an event/round.