Hey hey ho ho round ratings have got to GO!

[DiscFifty]

Originally Posted by ballgolfconvert

Biggest problem with ratings is that they reflect an average. Handicaps for example show how well a player is capable of playing throwing out the worst 10 if last 20 scores. . ratings could work the same way if they discarded the worst 50% of the ratings for each player. 20 rated rounds. The top 10 count in your rating. This would move a bunch of players who shouldn't be in Rec, out of Rec, and same for other divisions. It would go a long way to prevent sandbagging, which is why ball golf has the system they have.

I don't know much about ball golf (pops does, not me) but that sounds extremely logical.

[Cgkdisc]

Originally Posted by ballgolfconvert

Biggest problem with ratings is that they reflect an average. Handicaps for example show how well a player is capable of playing throwing out the worst 10 if last 20 scores. . ratings could work the same way if they discarded the worst 50% of the ratings for each player. 20 rated rounds. The top 10 count in your rating. This would move a bunch of players who shouldn't be in Rec, out of Rec, and same for other divisions. It would go a long way to prevent sandbagging, which is why ball golf has the system they have.

Averages of "good" scores are more useful for predicting performance whether they're your average performance or top end average performance like ball golf. Ball golf uses mostly self-reported scores. If they only used tournament scores like disc golf, they wouldn't need to trim the bottom half of rounds (I believe only best 8 out of the last 20 are used with the new R&A/USGA combined system.) They also use Equitable Stroke Control ESC to adjust abnormally high hole scores downward based on your handicap before determining your score for handicap reporting. I'm an advocate of hole score capping in DG at par+4 initially since it's the worst score we allow even when not playing a hole but it probably should be more like Par+3.

Their handicap team experimented with a handicap ratings process like we developed for DG but discovered how inconsistently players complied with score reporting guidelines. (I met with them around year 2000 and heard this from them in person at USGA HQ.) What they found was about 45% of players tended to pad their scores and about 45% reported lower scores with just 10% appearing to report actual scores. Of course, one 45% group was trying to boost their handicap since they played a lot of handicap rounds. The other 45% group presumably liked to brag about their low handicap since they rarely used it.

Whether you use overall performance averages or best performance averages, you would still have rating breaks where the same people currently in a division ratings range would fall into the same division as their current rating indicates. We already double weight 25% of a player's most recently rated rounds to better indicate their current performance. This adjustment is useful when more players in the system are improving versus declining, although once you are over 40, the stats indicate a steady slide that seems to accelerate in your late late 50s. But currently only 1/3 of PDGA members are at least 40.

[Cgkdisc]

Originally Posted by DiscFifty

I was fine and all things groovy until....

The weather impact is in the scores, the scores go into the formulas. We have no adjustments in the formulas for weather. The TDs can tell the PDGA about weather and it may provide help when there are few props (5>) and sometimes manual adjustments may be needed. Otherwise, the weather and all other factors are embedded in the scores of the props.

[Steve West]

Originally Posted by Cgkdisc

Odds of a relatively stable propagator at 850 rating beating a stable 1000-rated propagator on a normal course both are familiar with is about 1 in 1000. Add another pair of players with these two ratings where both 850s beat the two 1000s in a single round, it's 1/1000 times 1/1000 or 1 in a million. Each new pair extends the odds another 1000 times. In Jim Carrey's voice, "So you say there's a chance?"

You used a lot of qualifiers: relatively stable propagators, normal course, relatively familiar. Also, the basic assumption that round ratings are normally distributed around player ratings, and that performances are independant.

I looked at the Championship layout on Vasset Discgolfpark during the Sula Open 2018 presented by Discmania - EPT#2. Of the 1494 possible pair-wise match-ups between players with at least a 150 point ratings difference, 38 were won by the lower-rated player. That's 2.54%, or about 25 times the 0.10% your theory would suggest.

I would concede that Vasset may not be a normal course, the players might not be propagators (or stable), and they have an unknown familiarity with the course.

On the other hand, I would think if any 850-rated player beat a 1000-rated player, something might be going on to make it more likely that a second player would, too.

Yes, the chance of multiple low-rated players beating multiple high-rated players gets real small real fast, but not {1/1000 raised to the number of players} fast.

[Jugular]

Originally Posted by Cgkdisc

The weather impact is in the scores, the scores go into the formulas. We have no adjustments in the formulas for weather. The TDs can tell the PDGA about weather and it may provide help when there are few props (5>) and sometimes manual adjustments may be needed. Otherwise, the weather and all other factors are embedded in the scores of the props.

Earlier Chuck it seemed like you were considering two separate rounds. I think the example you were referring to was considering one long round including low and high rated players where high rated players tee off later in the day. Weather conditions change dramatically thus raising the chance of the hypothetical we've been discussing. You're saying the rating system would account for this but that doesn't seem consistent and it seems that would only occur if they are separate rounds or divisions. Are we missing something?

[adamgservo]

Weather isn't in the calculation. Only the scores and ratings for a given course are in the calculation. If the weather changes significantly, it becomes a different course, so the scores are separated into different buckets in order to perform the calculation.

Change in weather example:
https://www.pdga.com/tour/event/45796
For each round, 1000 rated round = [66 68 69]

No change in weather example:
https://www.pdga.com/tour/event/43983
Ratings are the same for each round.

[Jugular]

Originally Posted by adamgservo

Weather isn't in the calculation. Only the scores and ratings for a given course are in the calculation. If the weather changes significantly, it becomes a different course, so the scores are separated into different buckets in order to perform the calculation.

Change in weather example:
https://www.pdga.com/tour/event/45796
For each round, 1000 rated round = [66 68 69]

No change in weather example:
https://www.pdga.com/tour/event/43983
Ratings are the same for each round.

Adam, you seem to have misunderstood also, we're talking about the same round, one round.

[DavidSauls]

Originally Posted by ballgolfconvert

Biggest problem with ratings is that they reflect an average. Handicaps for example show how well a player is capable of playing throwing out the worst 10 if last 20 scores. . ratings could work the same way if they discarded the worst 50% of the ratings for each player. 20 rated rounds. The top 10 count in your rating. This would move a bunch of players who shouldn't be in Rec, out of Rec, and same for other divisions. It would go a long way to prevent sandbagging, which is why ball golf has the system they have.

Ah, the mythical sandbagger.

Such a change would raise everyone's ratings -- except the very top players -- and more than likely have one of two results:

(1) The division thresholds would change, so just about everyone would still be in the same divisions; or

(2) A bunch of current Rec players would get bumped to Intermediate, but a bunch of Intermediates would get bumped to Advanced, so you'd have essentially the same result as you'd have with the current system, if you just moved the division thresholds.

No matter what, some players are going to do well in their divisions, and be labeled "sandbaggers."

[araytx]

Originally Posted by jakebake91

Agreed. But isn't the number still calculable?
Maybe it's not an infinite number of monkeys. Maybe that number is 100,000,000,000,000,000,000,000 monkeys given 100,000,000,000,000,000,000,000 years?

Isn't that the thing we are arguing about here? Things being "mathematically possible"?

My point in all this is ......the original argument taking place, in my opinion, is just silly.

Nope. That's my point exactly. I NEVER couched my arguments in terms of 'mathematically possible'. My context was actually possible in the real world -- not in a mathematical formula. I've always maintained that you can put those numbers into the ratings computer. It's just that the results you get are meaningless. Garbage in = garbage out.

Originally Posted by txmxer

There's common use of language versus the language of science. Since the discussion drifted in to statistics and probability, the choice of words is expected to be precise. Calling a non-zero probability "impossible" may be functionally accurate because it is never going to happen, but, in the world of numbers and science, people typically choose to caveat the term.

There are reasons why. To use your example of .000000001%--is that the line where we say something is impossible? Why not .00000001%...or .001%? Where is one allowed to accurately use the term?

The odds of winning powerball are about 1 in 300 million or .00000033%.

But it happens.

Nobody would object to "it's never going to happen" or "for all practical purposes, it is impossible"...

ahhh...but the original scenario didn't include a bunch of caveats. It was simply 800 rated players having a better round than 1000 rated players. So...there is a tournament on day 2 where the lower rated players play in the morning and the higher rated players play in the evening. At noon, a front rolls in. Wind gusts 40 MPH and temperature drops from 60 degrees to 40 degrees. Those scores in the afternoon are going to suck. If one only looks at the numbers, then there is no explanation other than the 1000 rated players had a really bad day.

So, when stating, "it's impossible"...the next thing you will read is a bunch of caveats basically stating that all other variables are equal. Once that caveat is inserted in to the discussion over word choice, then the issue is settled. If caveats are necessary to make the statement accurate then it's just a matter of choice as to how one caveats the statement.

Someone already beat me to the punch. In the scenario above (even if you didn't intentionally manipulate it) those rounds aren't rated together. They can't be. I'm talking about same course/same day/same conditions/same time/same layout/same everything. That's where it's not possible for an entire set of 1000-rated players to play worse than an entire set of 800-rated players. Not without manipulation. And that's not a "caveat". That's my statement. Because D50 has been talking about/trying to get to or show that the system has a flaw when rounds that are rated separately but at the same time don't depend on the players incoming ratings. All I'm saying is both things can't be true at once.

Originally Posted by DiscFifty

Perhaps you're not looking at it in a small enough bite. Certainly for one day a worse player can beat a better player. It happens every tournament. Now the odds of every worse player beating the better players in 1 round are incredibly small, however...we're talking about avg rated round, so what increases those odds (perhaps alot).. is... some of the worse players shoot lights out and some of the better players shoot incredibly bad. If you think this scenario is flat out impossible... I would 100% disagree with that. Being human + weather conditions add random variables to the equation that will never provide a static result regardless the stability of the other variables.

I'm with you on that first part. One-offs or tiny sets can happen -- yes. The point I made (or that I thought I made) was, any set which is large enough to meet the statistical definition of reliable and valid won't be possible for that to occur, not without some outside manipulation making the experiment either unreliable or invalid.

[adamgservo]

Originally Posted by Jugular

Adam, you seem to have misunderstood also, we're talking about the same round, one round.

I see. That I'm not sure about.

I think I remember hearing that if say the early tee times had nice weather, and then a bad storm came through that it would be possible to separate the ratings based on the early/late tee times. I could be mistaken though.