PDGA Discussion Board - View Single Post

ck34 · Aug 29 2007, 05:29 PM

Nope. That's unnecessary math that doesn't provide any better ratings because the actual performance of similar rated players provides the curve in producing the score distribution to produce the SSA and the ratings that follow. In other words, a 1080 round occurs as infrequently today in the current calculation process as it would if the rating for that same score came about from each throw having an increasing rating point value like you're proposing.

We don't know if some courses of the same SSA produce flatter or more spiked distributions depending on length or terrain, for example. However, that's not necessary under our current process because we use the actual scoring distributions on these courses to produce the ratings.

And the math would be much more complicated and harder to explain with no payoff in accuracy. The thing is, we don't have any metrics to tell us how much each throw should be worth other than the stats and distributions we have already, so we would be adding more complexity on top of the current base calculations with no more accuracy.

Besides the the loss of accuracy from using too few rounds, we lose accuracy because the smallest data unit is one throw. If there were such things as quarter or half throws in our scoring, the accuracy would be increased, especially on low SSA courses where one spitout has a bigger ratings swing than at Winthrop.

Aug 29 2007, 05:29 PM	#575
ck34 PDGA Member Join Date: Nov 2003 Location: Twin Cities, MN Posts: 9,529	Re: Ratings are approximate Nope. That's unnecessary math that doesn't provide any better ratings because the actual performance of similar rated players provides the curve in producing the score distribution to produce the SSA and the ratings that follow. In other words, a 1080 round occurs as infrequently today in the current calculation process as it would if the rating for that same score came about from each throw having an increasing rating point value like you're proposing. We don't know if some courses of the same SSA produce flatter or more spiked distributions depending on length or terrain, for example. However, that's not necessary under our current process because we use the actual scoring distributions on these courses to produce the ratings. And the math would be much more complicated and harder to explain with no payoff in accuracy. The thing is, we don't have any metrics to tell us how much each throw should be worth other than the stats and distributions we have already, so we would be adding more complexity on top of the current base calculations with no more accuracy. Besides the the loss of accuracy from using too few rounds, we lose accuracy because the smallest data unit is one throw. If there were such things as quarter or half throws in our scoring, the accuracy would be increased, especially on low SSA courses where one spitout has a bigger ratings swing than at Winthrop.