Why ratings are stupid to worry about

[ThrowBot]

Random thought:

Rather than looking for super props with a lot of rated rounds, why not look for players who consistently throw similar ratings?

It's reasonable that a player who has a lot of rated rounds is probably rated very close to their average performance over the time period when those rounds were recorded. However, having an accurate average for that player doesn't help to predict how good their next round is if there's a large standard deviation in the set of their round ratings. For an inconsistent player, a single round score could be much worse or much better than their "average" skill level.

On the other hand, a player with very consistent ratings will have a small standard deviation in the set of their round ratings. For such a player you could be more confident that their next round should be within a smaller range of ratings.

Just spit-balling here, but each propagator could be weighted in inverse proportion to the standard deviation of their round ratings. That would help to filter out some "loose cannons" who have a lot of rated rounds, but aren't consistent from one round to the next.

 

[Steve West]

Random thought:

Rather than looking for super props with a lot of rated rounds, why not look for players who consistently throw similar ratings? ...

Because first you have to prove it is not just luck that those players happened to have thrown similar ratings. There is not really any reason to think their next round will be any more consistent than the next guy.

Take 100 identical players, let them all play 8 rounds, and they will get a range of measured standard deviations. Those that happened to have fallen at the bottom of the range are not really any more consistent than any of the other identical players.

 

[Cgkdisc]

One of the characteristics to determine who were super props for our study was a lower standard deviation separate from number of rounds.

 

[ThrowBot]

Take 100 identical players, let them all play 8 rounds, and they will get a range of measured standard deviations.

Your sentence seems to contradict itself. If they wind up with different standard deviations, then they aren't identical players. The conclusion disproves the original premise.

 

[Steve West]

Your sentence seems to contradict itself. If they wind up with different standard deviations, then they aren't identical players. The conclusion disproves the original premise.

In case you hadn't noticed, there is an element of luck in disc golf. Even identical players will get different scores. Also, I didn't say they'd be playing the same courses.

 

[ThrowBot]

In case you hadn't noticed, there is an element of luck in disc golf. Even identical players will get different scores. Also, I didn't say they'd be playing the same courses.

Yes I agree that luck is involved in disc golf. What I disagree with is your premise that there are identical players. I doubt that I could find 2 identical disc golf players, let alone a hundred. On the other hand, for the sake of a discussion about statistical analysis, I suppose that I could entertain the theoretical concept of identical players.

To have that discussion we would need to agree on a definition of what identical means. A Google search rendered the following, which I find suitable: "similar in every detail; exactly alike".

Per that definition, any non-similar details would make different players non-identical. I guess I'd disagree with your statement that "identical players will get different scores." Players with different scores wouldn't be identical. You've also stated [and suggested] other non-identical traits for the "identical" set of players, such as different standard deviations in round ratings [and different courses played].

So really, this isn't a conversation about a group of theoretically identical players. It's a conversation about a group of players that aren't identical.

Unfortunately, that discussion isn't really the point of the thread :|

 

[joshmo65]

I just want to be called a super propagator...

 

[brentjacobs]

I just want to be called a super propagator...

Me too! My standards don't deviate very much, come on!

 

[ToddL]

I'd be happy if they just called me a pretty good propagator...

 

[cjdiscgolf]

The issue isn't that they are fixed afterwards by combinging them, the issue is this shows that they the calculation for them is flawed.

If the calculations were correct there would be no need for fixing them.

Addressing the problem isn't correcting the problem.

Ratings have and always are based on who is playing when. The only thing in golf you can't control is what your competitors shoot, yet that's exactly how ratings are calculated.

 

[Doofenshmirtz]

The only thing in golf you can't control is what your competitors shoot, and that's exactly why ratings are calculated how they are.

FIFY?

 

[geoloseth]

The way Chuck explained the ratings calculations to me a while ago (I've lost all my emails since then) is that IN=OUT. This means that the course becomes irrelevant in the ratings calculations. The average rating of the propagators during a specific round will equal that average rating for that round. And you can take any round you'd like from an event. Take the average of the players in a round that played a certain layout and it will be pretty close to the average round rating of that round.

This means that if you take 10 players a 910,915,920,925,930,935,940,945,950,955 (Average of 933) and put them on a course. If each player happens to shoot a 50. Then that round would be a 933 rated round.

Now take the same course under the same conditions but put a different pool of players. the top 10 in the world. Lets say that their average player rating is a 1033. They happen to all shoot that same exact 50. That round would instead be rated a 1033.

Chuck will have to correct me if I'm wrong, but that's the jist of conversation we had some time ago. Like we can all assume, these exact results would never actually happen in tournament play. But I know of at least one event where the same score from an Am weekend to a Pro weekend was ~40-50 points less for an identical course/conditions. The only thing that changed was the average player rating of the propagators. The fact that the formulas used in calculating ratings allows it to happen means that there is an obvious flaw in them.

 

[teemkey]

Per that definition, any non-similar details would make different players non-identical. I guess I'd disagree with your statement that "identical players will get different scores."

Now wait a minute. I often play solo with two sets of discs, keeping score for each set. I rarely get the same score for each set.

Are you saying I'm not identical to myself?

 

[sprintermatt]

With ratings, you get out what you get in. Average of player propagator ratings = average of ratings for tournament with scoring spread somehow based on par. Higher pars, less points per stroke.

 

[Horsman]

My dgcr Rating is 35 points higher than my actual rating

 

[Yetistomp]

Great conversation and Math hurts my head! I have always been saddened that in a sport where we play against the course we actually get rated against the quality of players present and playing the same course at the same time with everything smoothed over by averaging it all out in the end. I totally get why we can't have a true SSA for a course due to a hole/pin or three changing year to year. Our Ratings system is based on the Law of Averages. The same law that says while most players find wind and rain more difficult to play in, there are actually enough players that play above average and enough that play below average in those conditions that it all means poor conditions don't mean much as Chuck has said in the past. Of course, this means even less if the TD doesn't report large condition changes from one pool or tee time group to another in order to not average them together, note to TDs.

Here is a great or not so great example of averages from this last weekend's US Masters:

Bradford Shorts (Par 55) Shooting a -9 (46) with same conditions for all rounds. This is birdies on half of the holes during a round.

Pro GrandMasters and Pro Master Women (no 1000 rated players)
Rd 1 (1015)
Rd 2 (1026)
Rd 5 (1015)

Pro Masters (8 1000+ rated players)
Rd 3 (999)
Rd 4 (1003)

It all averages out to 1011.6 which I suspect will be the final number for everyone. Better than the disparaging difference between a 999 round and a 1026 round.

 

[cjman]

.....

 

[Smigles]

I'm merely stating that they are only accurate within a +\- 20-30 point variance. Which isn't bad, but I think the system caused some issues. Tournaments where entry times are based on ratings, divisional breaks, etc.

But I do think that a person who plays enough tournaments has a fairly accurate indication of their rating.

It just seems to fluctuate too easily. And it's frustrating when someone says I just shot 1000 rated round then you realize you shot the same score last year or last week and your rating was only 970.

Really I was just bored and haven't posted in forever.

Guess I just miss the old days when we had no ratings. When you moved up a division when you felt you were ready. But I understand it from a sandbagging point of view, though I think sandbagging was rectified by other players making fun of them for winning 25 int events in a row.

Man, I need to get out and throw some discs.

All I see is complaining.

Please propose a better system.

 

[chris deitzel]

Better system? How about no ratings?

 

[RyanisCorriganis]

Better system? How about no ratings?

Normally when someone makes a break through in the science/math community they publish their findings for others to scrutinize.

IIRC, the formula to calculate PDGA ratings is a well kept secret. Jeverett and some others spent lots of time essentially "redesigning the wheel" to get us DGCR ratings.