My latest blog entry.

Enjoy

http://mtlscienfiticdiscgolf.blogspot.com/2008/09/ratin-about-ratings.html

From MTL's latest blog entry on player ratings:

"As the creator of the now standardized rating system the PDGA uses, Kennedy was able to seemingly fix one of the main problems within disc golf that ball golf never had an issue with; standardization of par."

* * * * * * * * * * * *

Standardization of par? :confused:

What about that doesn't make sense Jerry?

I have never considered the Player Rating system a "standardization of par".......two separate things (imo).

It's not. This is why I used the word "seemingly" and went to explain how it to some degree standardizes par b/c you can, to some degree, compare an SSA 49 course to an SSA 54 course based on the rating.

Thanks for reading, btw

Is too! :D

I now understand what you were trying to say in your blog, however, I disagree that the player rating system is a "standardization of par". Player ratings won't tell me who's in the lead after nine holes of play during a final round......a player's score in relation to par does. Player ratings can change from day to day, year to year.....course par does not.

I agree.

But as I said, I really think of the primary benefits of ratings is like what I talked about with the golf course / disc golf course example and was going along the lines of a lot more people understand I shoot 990 at Cedar Hills than they do I shoot 50.

Ideally, I would like to see the implementation of a course rating system in addition to the current player rating system. It would be difficult and time consuming to create a course rating system (particulary on courses where multiple pin placements result in par/SSA differentiation), however, a static course rating would nicely compliment the player rating system which is currently exclusive to PDGA membership.

Like the name drop MTL, good article.

It's apparent he still doesn't understand the ratings or he wouldn't have written the second half of the blog that way. Those who understand ratings do understand why the way we do it is better and in fact the only functional way to do it. But he's young and has many more years to eventually understand.

Perhaps my favorite example this came on at Glenburnie Park in New Bern, NC. This past year on the short tees, I put together my best personal round of 46. Mike Hofmann, now rated 1005, matched this round as well. However, North Carolina's newest 1000 rated player, Jeb Bryant, smoked the course and fired a course record 41. Since I had shot a 49 at a previous tournament on this course in similar conditions and received a round rating 1008, I was anxious to see a round rating of possibly 1040, which would be my third highest ever. However, since Jeb and others played really well that round, my 46 was rated LOWER than my 49. Exact same course. Same conditions. Three strokes better. Two points lower.


But he's young and has many more years to eventually understand.



I really don't wanna understand something like that....

The occasional scenario like that is miniscule in comparison to the error induced from weather variations every weekend using fixed SSA values like golf, let alone the impossible data management tasks involved in trying to have TDs figure out what SSA should be used in the report when more than 1/3 of the events have new courses or temp configurations with no SSA value. TDs have hard enough time doing the work to make sure players are in the proper division. Fixed SSAs would also mean no unofficial ratings during or right after the event. Ratings would only be provided during official updates.

Since you are forced to determine SSAs from propagators regularly anyway in order to figure out the official fixed SSA value, it's better to be consistent and do it that way for every round. Fixed SSAs would have brought the data management system in 1999 to its knees before it got off the ground and more divisions and players would not have gotten ratings. In fact, the way we did it was the only way ratings would have gotten underway since Roger and I did it free for five years with little PDGA backing. And we wouldn't have done it with fixed SSAs because the process would have been unmanageable.

But he's young and has many more years to eventually understand.



I really don't wanna understand something like that....

[/QUOTE]

Yet it makes sense to you that a 54 on a particular course, on a sunny, calm day should be rated the same as a 54 on the same course in pouring rain and high winds?

Do you think it remotely feasible to accurately identify the SSA for each course, with varying pin positions, tee pads, and an ever changing fairway? After all, most disc golf courses evolve with time, as trees grow and die. Holes change and the course rating would be outdated in no time.

Comparing a player's average to the field provides the best meaningful rating for that day. It is certainly not indicative of absolute performance, and you certainly risk anomolies that you describe above, but other systems would be more flawed and less meaningful.

I do disagree that outliers shouldn't be included in your ratings (they should be), and that bailing on a bad round should help your rating (it does, but it shouldn't), but the overall premise is sound.

The occasional scenario



That made me laugh.

There will absolutely be some variances like the one you point out. The good news is that because it's a statistical process, you get just as much 'good luck' as 'bad luck' so it won't matter overall. In fact, the odds of getting the exact same SSA within one decimal place on the same layout with the same players in the second round is less than 10%. That's why we now average the two rounds exactly for the reason having fixed SSAs is good - so players get the same rating for the same score. It doesn't take long to go thru the sequence of what would be required to try a fixed SSA process to see why you should thank your lucky stars the way we do it is possible. But even with your TD experience, you won't acknowledge that impossibility.

I actually have said many times that doing the ratings for all rounds played one layout was a good thing and a step towards doing them the right way.

However, as soon as I said that I played in perfect conditions on Saturday and 50 MPH winds on Sunday. This lead to a very high rating Saturday and a very low rating Sunday, so I guess it averages out.

It's up to the TD to tell us not to combine the rounds.

And you think they should review a database of fixed SSAs and figure out whether an exisiting value is the same or different and how much to adjust it for the temp hole and the big tree that came down? And remember that a pin placement was hit by the mower and moved over. Then figure it out for the divisions playing different layouts and then adjust for wind factor we'd provide them to determine on a graph? :eek:

It's apparent he still doesn't understand the ratings or he wouldn't have written the second half of the blog that way.



That's what bothers me the most. Ignoring the whole dynamic SSA vs. fixed SSA argument, I look at the two paragraphs that begin with "Here is another level of stupidity about the ratings" and "It gets worse" and it's obvious Robert doesn't even understand how the PDGA ratings system works.

A responsible journalist would have done more thorough research before putting out such an article, and similar to his "PDGA broke its own rules" blog post, I think what we have here is another article which relies on sensationalism to attract readers. http://forums.slizone.com/style_emoticons/slizone/thumbsdown.gif

Three shots better. 2 Points worse.

I can not repeat this phrase enough.

And the crazy part is Chuck will be like "yeah that happens but its rare" but there are examples ALL over the place.

Here are the stats on the four courses played in my last tournament. This four things SCREAM consistency don't they....

http://www.pdga.com/tournament/course_ratings_by_course.php?RatingCourseID=343

http://www.pdga.com/tournament/course_ratings_by_course.php?RatingCourseID=102

http://www.pdga.com/tournament/course_ratings_by_course.php?RatingCourseID=1587

http://www.pdga.com/tournament/course_ratings_by_course.php?RatingCourseID=134

Three shots better. 2 Points worse.

I can not repeat this phrase enough.



I guess it's really not a "standardization of par" then... /msgboard/images/graemlins/smirk.gif

3 shots variance on a 50 SSA is still only 6% variance. That's about the difference in sales tax. First, it's statistically uncommon and you can just as easily get a benefit the other direction some other time. Many player's ratings precision expectations are out of line because each point seems so precise. But one rating point is just 0.1% variance and a cut thru can be 10 points or 2% variance in your score and one centimeter variance in a single throw at USDGC can produce a 3% variance. The SSA variance is less than those common occurrences per round and insignificant overall due to averaging. However, fixed SSAs if actually known are ALWAYS wrong in the same direction due to no weather adjustment which would occur in more than 1/3 of the events each weekend but impact certain regional areas more than others.

I don't know why people think I am in favor of a fixed SSA across the board.

Of course if I play a course in rainy conditions that course will play more difficult than the exact course in perfect confiditions - that I am not arguing one bit.

However, there needs to be a way to alter from the fixed SSA of the layout when weather is bad and I don't think basing it on peoples ratings is the way. I think if the course info had something along the lines of - Joe Doe Memorial Park - 30 MPH Winds from south instead of Joe Doe Memorial Park Joe Doe Memorial Tournament Round 2 - that could be a step in the right direction. The first example provides an explination to an outside eye why the course was that difficult that day. The second example and current use only is usefull to those who played the tournament or have access to the TD report (And since you love stats so much, Chuck) or only roughly .005% of the current membership. That is a pretty usefull stat - only .005% of PDGA members can explain why a course played 2 strokes harder that day.

The problem with the basing it off of what everyone shoots, I feel, is that when the conditions are consistent from tournament to tournament, much like the example I have listed in this thread, then players are penalyzed when the field plays good.

If ratings were in essence nothing more than just something we did as a sport for fun, I wouldn't care much. However, due to the many things I listed in in the blog, that is not the case.

I feel if you polled all PDGA members you would find that at least 50% (probably closer to 60%) of them think the ratings system is flawed. The main problem with that I think is how Chuck presents his arguments.

He is never wrong and never will be wrong.

I feel if you polled all PDGA members you would find that at least 50% (probably closer to 60%) of them think the ratings system is flawed.


I would be in that 60% and have identified the challenges. But the alternative is worse. Likely 90% think that the tax system is unfair in some areas.

The key problem with fixed SSAs is there's no basis upon which to determine their values. And we know how much agreement there is on determining par which is much simpler in theory. This is the key problem with ball golf course ratings - no objective way to do it. Second, there's no proof that a course plays with the same challenge in the morning and afternoon under the same conditions. There's no way to verify that it does play the same or it doesn't.

MTL - you presented the problem with the current ratings system, which has some valid concerns IMO. However you fail to present a feasible solution. CK seems to argue that yes the ratings are not perfect but they work under most situations and it's easy for ALL tournaments to have ratings no matter how dumb the TD is.

Unless you are offering a solution then you are just complaining. What do you want to change how do we do it?

Dave Feldberg is the highest rated player and also the best at disc golf. I think that indicates that the ratings work at some level.

did you not read 2 posts up?

Your proposed solution is to have TDs list weather conditions as part of the course info?

How many TDs will actually list weather? How do they know what to list? 5mph wind with 20 mph gusts for 20 seconds at a time, followed by a 20 min period of rain and then sunshine for the remainder of the round... you can see where that goes.

"step in the right direction" - maybe. What direction?

I think Mr. EsDubYa is presenting some skillfull moderating work here or in other words MB facilitation. You facilitate meetings and productive conversations for a living or something?

You complain a lot about ratings, but how close do you think a players overall rating is to there actual level? The good players around here have high ratings and the bad players have low ratings. Yes, every round rating is not an exact measurement but nothing ever will be. It sounds like your complaint is that your rating is based on your play against other players and not based on how well you played. This is true of any ladder system which is basically what ratings are. If you understand the whole process, then you mus be pretty smart guy because it is some pretty complex math. So as a smart guy, why don't you come up with another option.

I def. do not understand all the numbers and the calculations but I do understand the overwhelming complaining I hear just about every weekend about the ratings.

There is no possible way for me, with no background in statistics, to develop such a system.

The only thing I can do to help is point out the things I see as problems and hope those with knowledge of how to fix them can. If I am seen as a complainer, that is fine by me. I know what I am doing is standing up and voicing my opinion about something I don't agree with and I bet more than likely most people are with me on that.

Your first complaint is that your rating is based on how the players you played against that day did and you think it should be based on course history. The problem with comparing scores against previous tournaments is the conditions are different. You even use an example saying exact same conditions. Personally I doubt they were exact. Could you at least give us the dates of the tournaments, the temp., the average wind speed, and the humidity for the 2 tournaments. Or were they all the exact same? To start using these pieces of date in the rating equation a lot of data would be needed to see how each effects the scores during rounds. If enough data was gathered trends would probably exist that could be put into an equation to alter the SSA for a course. Courses also change when trees fall and trees grow. These I don't think could be measured so that would always be a problem. Also, as better discs come out courses get easier. I know players at my local course are shooting a lot better scores then they did 10 years ago. I am not sure how you would put that into the equation.

Your second complaint is about 20 players a 1000 player, a 990 player, a 980 player, ..., down to 800 (which fyi is 21 players). I am not sure I understand your argument, but I will try and address it anyways. You are saying complaining that it bases the SSA off of the idea that the 1000 rated players ability is 1000 and the 990 rated players ability is 990, etc. Yes, that is a problem if a player is a lot better than your rating or a player s injured and a lot worse than their rating. I think any system that bases the SSA on how players shoot and what their ability is based on previous rounds is going to have this problem. Even looking at previous rounds at a course would have this same problem. The overall percentage of these players would be the same, but yes it would not affect one tournament as much as does now. If you play enough tournaments it will average out and your overall rating will be the same.

Your third complaint is high ratings on courses with SSA 50 -54 and low ratings on courses with SSA over 60. I had to think about this quite a bit and I still do. I am not sure how you would fix this. Without varying the amount that each stroke is worth. The problem is the number of strokes between a 1100 rated round and a 1000 rated round is the same number of strokes between a 1000 rated round and a 900 rated round. You take a bunch of 900 rated players see what they shoot. Take a bunch of 1000 rated players see what they shoot. Add the difference to 1000 rated players average and that is 1100 rated round. Why players don't shoot that score on hard courses I am not sure. I would guess is the more throws that are involved the bigger the difference between these 2 groups because the 900 players are just so inconsistent. The bigger the difference that just means the more strokes a player has to shoot better than a 1000 rated round to get to 1100. On these hard courses it is just impossible to get there. Now that I have explained why this is I guess the question comes, why does this matter? I would guess that if a player played all easy courses their overall rating would be the same as their rating would be if they played all hard courses. They would have a bigger range of ratings on the easy course, but the average will be the same. So your complaint is just that the best rated round of all time will be on easy course? You know the best round in PGA history is on an easy course also, so guess that is just the way it is. If you wanted to figure out what was the best round in PDGA history, it would not necessarily be the highest rated round. I would have to think about it, but an equation could be written to compare rounds based on the SSA and the rating.

Your last complaint is that no one outside the sport knows what a rating is. Personally, I take my rating take the difference from 1000 divide by 10 and tell people that is my handicap. Anyone who is in a golf league totally understands that. Your comparing of rounds of 1010 and 1020 to an non player and trying to explain why one is better than another. Wow you must bore people with this story. I already addressed the issue of the round be rated without looking at other rounds. Also, in ball golf you can't totally compare scores from different courses either. Otherwise after the US Open I would be telling everyone I should of played, because I shot even at my local course that day. Or should we start explaining to non golfers that there are different tees, some greens are fast and some are slow, and some rough is rough and some rough is really rough.

I just thought I would add a few of my thoughts to your complaints. Note, I do not think that the rating system is perfect. I think it does a good enough job that the PDGA should stick with it. Actually, I was disappointed to here that each round doesn't stand by itself anymore. I understand that the more rounds the more accurate the rating, but I just think there are too many other factors involved as listed above to make this fair.

You do come off as a complainer, but it is good that it is fine for you. And yes most people are like you in that they are not good at statistics, would not know how to specifically improve the ratings system or make something better, and yet they complain about it anyways. /msgboard/images/graemlins/smirk.gif

Your third complaint is high ratings on courses with SSA 50 -54 and low ratings on courses with SSA over 60. I had to think about this quite a bit and I still do. I am not sure how you would fix this.


Read my By The Numbers article in issue 4 of the FDM when it comes out for a new perspective on this.


Actually, I was disappointed to here that each round doesn't stand by itself anymore.


Actually they do if they are far enough apart. If the SSA difference is less than 1.5 between two rounds on the same layout, then they are combined. If the diff is more than 1.5, they are kept separate. We don't feel the process can differentiate whether a 1 throw SSA difference is due to statistical variance or wind so we combine them. Once you get to 1.5 difference, the assumption is that weather made the difference.

I may have just found my first article for 1000rated. I hear Brian wants to interview Chuck (has he got ahold of you yet, Chuck?), so I could put it up in conjunction with that.

Brian asked and I told him I would do the interview after I got back from my Summit trip week of Sep 21.

However, there needs to be a way to alter from the fixed SSA of the layout when weather is bad and I don't think basing it on peoples ratings is the way. I think if the course info had something along the lines of - Joe Doe Memorial Park - 30 MPH Winds from south instead of Joe Doe Memorial Park Joe Doe Memorial Tournament Round 2 - that could be a step in the right direction. The first example provides an explination to an outside eye why the course was that difficult that day. The second example and current use only is usefull to those who played the tournament or have access to the TD report (And since you love stats so much, Chuck) or only roughly .005% of the current membership. That is a pretty usefull stat - only .005% of PDGA members can explain why a course played 2 strokes harder that day.




I like the idea of listing the weather as part of results, just as other sports do:

Minnesota 24, Green Bay 19 - 72 degrees, 5 MPH winds.

http://scores.espn.go.com/nfl/boxscore?gameId=280908009

Sony Open in Hawaii (Jan 10 - 13) Round 4 - 25 MPH Winds from SSW

http://sports.espn.go.com/golf/leaderboard?tournamentId=418


wait a second...why wasn't that information shown in the box scores?

In general, sports score reports show....scores. Write-ups, on the other hand, would certainly include mention of weather which impacted the game - and the same is generally true for disc golf (for those events that have write-ups).

Just like college football rankings are imperfect, so are disc golf ratings. In sports, it is generally hard to absolutely quantify results in any real, meaningful terms. A team's record may not be a true indicator of how good they really are, unless toughness of schedule is looked at.

The only objective results from a tournament are: a player shot X score. Or player Y won by X strokes. While it may not be meaningful, nor are scorelines in other sports barring data surrounding the players or teams involved. Ratings serve as a guide to give an idea of how good or bad a round is. It is an inexact science, but is more sophisticated and reliable than alternate methods to quantify performance.

Well worth a read is this BBC article (http://news.bbc.co.uk/2/hi/uk_news/education/7545529.stm) about school league tables. Conveniently they base their values from 1000 as well, so it seems to apply even more closely to the PDGA ratings.

The key point to take from this article is that a single round rating is like a single CVA, i.e. that because of all of the weather condition and players shooting hot/cold rounds factors, the confidence interval is great enough to cover every single complaint that MTL makes about the system.

The second point is that this doesn't matter! The whole point of statistics is to take a variety of data points with large confidence intervals, and to combine those in order to establish a trend. The larger the number of data points, the tighter the confidence interval of the end result gets.

The fact of the matter is that you simply cannot have an accurate round rating (other than by chance distribution within the confidence interval), they are simply means to an end, data points towards your far more accurate overall rating.

Obviously the system needs to do it's best to minimise the confidence interval on individual rounds, and things like not using players >60 points off their rating as propagators are useful in achieving this.
With unlimited resources and time I'm sure more could be done, but this needs to be balanced against the reward.

Currently I'd estimate that an individual round rating is accurate to better than 5% in general (there will always be outliers), and that with ~10 rounds, the actual player rating is accurate to better than 0.5%

Halving the inaccuracy in the individual round would take a massive effort by both the ratings team and reporting TDs, with detailed info about weather, injured players, and stats to work out how these things effect things. And the result? Maybe 1 rating point difference in your player rating.

The fact is that individual round ratings are nothing more than ego stroking, they're meaningless, other than as a means to an end, an accurate player rating.

One other item that MTL doesn't account for: Even if a course layout hasn't changed and the weather is more or less the same, a round from a couple of years ago may be rated higher than one from this year purely based on changes in disc technology and throwing techniques. Holes that may have been just unreachable before might now yield lower scores. I don't think this is a huge difference year over year, but over time some courses are being "obsoleted".

From MTL's blog:


Tournaments are starting to turn two 27 hole rounds into three hole rounds just so the ratings won't be off.



Please name 10...wait name 5 events where this happened. (OK, how about naming 2 not including any you may have been involved with.) I run events at 27 hole courses and I can assure you I have never ever considered ratings when deciding the format. I have never had players ask me to consider not having 27 hole rounds because it may adversely affect ratings. I don't think most TD's would ever consider ratings when deciding a format unless they were playing less than 18 holes in a round. I think only a TD that puts undue importance on their own personal rating would even consider this.



I feel if you polled all PDGA members you would find that at least 50% (probably closer to 60%) of them think the ratings system is flawed.



I think most of this statement is correct: however, wouldn't feel comfortable drawing some percentage out of thin air. While many may agree with you that the system is flawed, very few complain about it and even fewer want to change it. The ratings system does exactly what it was designed to do: provide a measure of performance.
Ratings have proven to be a fair record of past performance as well as a fairly accurate indicator of future performance.


If ratings were in essence nothing more than just something we did as a sport for fun, I wouldn't care much. However, due to the many things I listed in in the blog, that is not the case.



I think you over-exaggerate the importance of the "many things" listed in your blog...


...However, if we are going to base who gets sponsored, who gets recognition and by all means who can play in what division on our ratings, shouldn't they be a little bit more accurate and not based on such a silly system?



While ratings are used to determine divisions, this is for lower rated Amateurs who are playing for prizes and hopefully fun. I don't think it is a serious problem that perhaps a handful of players in these groups may be forced to play up a division or have to face a player who should be playing up. It is common knowledge that ratings may lag behind a quickly improving player. A new system would not change that. Only providing more frequent updates would help solve that problem. IMHO, it would not be worth the extra resources. These young guns can wait patiently for an extra month or two to see just how "big" they are.

Why should the PDGA care if a potential sponsor wants to use ratings to determine sponsorship? Why should the PDGA care if certian players place an inordinate amount of importance on ratings to boost their ego or prove how "big" they are. I think sponsors know how to view a player profile and see if a player rating is inflated or if someone is dropping or tanking rounds. Sponsored players get bonuses for winning events, not for tanking rounds or quitting.

If you are hearing players complaining week after week about the ratings, it must be a very small circle of players because I don't hear it and I don't see it on here either.

The only people locally I hear complaining about ratings are the ones who really don't understand them.

From MTL's blog:


Tournaments are starting to turn two 27 hole rounds into three hole rounds just so the ratings won't be off.



Please name 10...wait name 5 events where this happened. (OK, how about naming 2 not including any you may have been involved with.)



Wasn't involved in either one..

Tupelo Bay 06. Two 27 hole rounds. The TD broke it down into three rounds.

And a tournmaent in Texas this past year - My buddy play in it - broke down a larger course into smaller rounds JUST b/c of the ratings.

And a tournmaent in Texas this past year - My buddy play in it - broke down a larger course into smaller rounds JUST b/c of the ratings.



Crowley, perhaps? I believe I recall reading about that one....

I feel if you polled all PDGA members you would find that at least 50% (probably closer to 60%) of them think the ratings system is flawed.



I think most of this statement is correct: however, wouldn't feel comfortable drawing some percentage out of thin air.



Neither would most responsible journalists. But in the interest of sensationalism to attract readers, it's perfectly acceptable to pull numbers out of the air.


If you are hearing players complaining week after week about the ratings, it must be a very small circle of players because I don't hear it and I don't see it on here either.



Neither do I. The only complaints I hear is when a tournament director in my state doesn't upload the results to the PDGA server within a few days after their event. People are usually banging doors down to get their ratings, not complaining about them.


The only people locally I hear complaining about ratings are the ones who really don't understand them.



<font size=4>BINGO.</font>

It is perfectly acceptable in an opinion style column, such as this, to guesstimate.

I clearly said "I feel" at the start of the sentance which further clarifed the stat was completly based on my own.

I'm pretty educated about things in journalism being that I have a degree in it and know what is acceptable what is not acceptable.

Had this been a report on the ratings, no such number would have been in the blog without stats to back it up.

What is so hard about that to understand?

Also..

this is a PERSONAL blog that I enjoy sharing. Had John Doe written such an article with no journalistic back ground, such claims of irresponsible journalism practices would not be made.

It is merely a coincidence that my thoughts in my blog come from someone who has a journalism degree as anyone could go into blogspot and write their own blog with the same ideas and that would be fine.

Had I published this in a paper or media outlet, then you are somewhat on to something. However, it would be allowed b/c opinions are subjective and are allowed to have "sensastionalism to attract readers" in them. It is a style or writing and one that I am the best at.

And a tournmaent in Texas this past year - My buddy play in it - broke down a larger course into smaller rounds JUST b/c of the ratings.



Crowley, perhaps? I believe I recall reading about that one....



http://www.pdga.com/tournament/tournamen..._ratings=1#Open (http://www.pdga.com/tournament/tournament_results.php?TournID=7748&amp;year=2008&amp;incl ude_ratings=1#Open)

Rounds 1 and 2 were actually 1 round of 30 and rounds 3 and 4 were another round of 30. They turned these in as 4 rounds of 15 simply b/c as I was told "E-Mac killed it the year before and he was screwed on his rating b/c the SSA was so high."

Another thing I don't like about ratings.

A player could play courses all in the SSA 50 - 54 range (Which numbers prove produce the highest SSAs) and play about the same level of golf as someone who plays only SSA 60+ courses and the first would have a better rating.

I can understand seperating these types of SSAs for reasons of tracking the best rounds ever within those, but then the raw numbers used for ALL SSAs are lumped together to arrive at the players rating.

If we recognize that different SSAs can not be compared together, why do average them out for a players rating?

A player could play courses all in the SSA 50 - 54 range (Which numbers prove produce the highest SSAs) and play about the same level of golf as someone who plays only SSA 60+ courses and the first would have a better rating.



Not correct. Statistically, we have verified that everyone shoots their rating on average on courses over the full SSA range. You will not shoot any better on average on an SSA 50-54 course than on SSA 60+ or even those under 45. However, your standard deviation will be higher than if you only play SSA 60 courses. It's only the ratings spread, not the average that is affected by the SSA.

You're in denial CK. This condition is most evident when you look at course records on high SSA courses. My CR at Hawk Hollow was a 1058, Markus' CR at the whippin post was also low. Compare that to the multiple 1080's and 1090's given to high rated players shooting in the low 40's on lower SSA courses, and the disparity is obvious.

Assuming that everyone always shoots their rating does not allow for recognition of exceptional performances. And also tends toward providing lower ratings for the exceptional performances of players without the highest ratings.

This is a flaw in the formula. It's not an insult - it just............is

But a 990 on an SSA 45 is not as good as a 990 on SSA 60, right?

They are both as good but the 990 on the SSA 60 is marginally less likely than on SSA 45 unless you are a 990 player. Then they are both normal.

So if they are basically the same then why do we group them seperately?

They're only grouped separately because we're ranking the statistically best rounds. If a certain SSA range has more chance for someone of the same rating to shoot an exceptional round compared to another SSA range, then the ranges need to be kept separate. However, players with the same rating will end up averaging the same rating regardless of the SSA they play. Of course, some play wooded courses better than open and some play elevation better. But on average, everyone of the same rating shoots the same rating on average regardless of the SSA.

So Mike Johansens 37 at Kentwood Park this year was rated 1075 and the SSA for the round was 42.37 was better than Brian Schwebergers 57 at Renny Gold rated 1073 with an SSA of 70.16?

I mean, 1075 is larger than 1073.

Apples and oranges. It's like comparing a high school guard who shot 75% on 3-pointers with an NBA guard who shot 73% on 3-pointers. Pretty close but not really comparable.

But then when we calculate a players rating, we count all rounds regardless of the SSAs.

But I thought we can't compare apples and oranges

The difference is individual rounds can't be compared in terms of which is better when the SSAs are far apart. However, each round is just one datapoint among several in a standard deviation spread that cumulatively reflects a person's average performance.

Stop focusing on the individual rounds as others have pointed out. You play enough rounds that the average of those rounds gives a good representation of what you did. As a journalist looking at a photo with your nose touching the print might give you a good look at the pixels but doesn't give you much of an idea about what the whole picture looks like.

I'm seriously confused.

I'm trying to understand, so far I don't.

I'm not questionning my rating as I play enough where I think its pretty accurate.

Think of it as impressionist painting Robert. Lots of little blobs of paint with no well defined lines - melded together to make a picture which gives you the impression of your skill level.

That was beautiful, Craig.

He said Lots of little blobs! :D:D

Think of it as impressionist painting Robert. Lots of little blobs of paint with no well defined lines - melded together to make a picture which gives you the impression of your skill level.



Awesome comparison.

You just forgot to mention that some blobs can't be compared to other blobs :D

You just forgot to mention that some blobs can't be compared to other blobs


Different number of strokes... of a brush that is... ;)

Chuck, I don't know if you can look this up or not, but you can you look up the worse rated rounds by over 1000 rated players? The worse rounds should come on low SSA courses. If the data shows this to be true, then maybe that will open the eyes of some people that their is not an advantage to playing one type of SSA course. Note, bad rated rounds could be from penalties that could skew the data.

That's the problem. You can shoot a lower rated round on any course. Although it takes more throws to do it on a high SSA course. OB penalties cost fewer rating points on Winthrop Gold even though a player might take more than on a typical 54 SSA course.

Neil, thank you for your post. It helps me to see more clearly why it is not so important to rationalize why a great round at a pushover course is rated higher than a great round at a tough course. I was impressed while reviewing the results of this years masters division at the worlds. There were 10 players with ratings of 1000 or more and they were all in the top 11 as listed below. The rating system is obviously not perfect but it surely does give us a resonable method of comparing skill levels after enough data has been accumulated.
1 Al Schack 3407 1013 50 52 48 50 62 52 51 50 32 447 $3,000
2 Joe Mela 2607 1009 51 52 55 52 59 49 56 47 35 456 $1,600
2 Jim Oates 3351 1014 47 54 50 52 62 49 52 53 37 456 $1,600
4 Brad Hammock 5912 1000 47 57 51 52 64 45 56 48 37 457 $1,200
5 Stan Mcdaniel 2938 1013 46 53 50 55 61 52 53 51 421 $850
5 Eric roady Roadhouse 10017 1005 46 60 48 55 63 52 49 48 421 $850
7 Dean Tannock 4028 1003 51 51 53 55 61 50 53 48 422 $750
8 Brian Mcree 7883 1006 51 53 50 52 64 50 51 52 423 $675
8 Mike Raley 7846 1005 51 55 52 52 64 49 49 51 423 $675
10 Michael Sullivan 3028 990 50 53 54 52 64 47 53 53 426 $588
10 Dan Ginnelly 3591 1003 55 50 50 54 63 52 51 51 426 $588

The rating system is obviously not perfect but it surely does give us a resonable method of comparing skill levels after enough data has been accumulated.


qft... well said

MTL can you name any more events that opted out of a 27 hole format or reported the scores differently? So far 2 events out of 500 doesn't really seem to be defining a trend for me.

Next, please explain how to break up my 2 round 27 hole event into three 18 hole rounds for ratings. We played one round on Saturday and the other on Sunday. In order for rounds to be compared wouldn't everyone in every division would have to play the same holes in the same 18 hole breakdowns? There was a shotgun start, with groups on every hole. Since the event was overfilled, there were a few mixed division cards. How do you keep track when groups are shuffled between round 1 and 2? For example a player could have started on Hole 9 the first round and Hole 15 the second round. And what about the last place Intermediate who played Round 2 with the Advanced Masters on holes across the course from the rest of his division?

Don't bother trying to actually do this, as this is simply hypothetical. We didn't play the same layout both days. But I still want to know how a TD would go about doing this and why they would want to do so much extra work.


Currently, I would simply have to input the round scores onto my Excel file. This would take about 15 minutes (even less with help). It would take a minute or two to input the course data. Then I could copy and paste the results to the PDGA website in seconds. Another thirty seconds to attach my TD report to an email to the PDGA and would be all done!! Please explain to me why I would want to spend hours and hours going over scorecards and breaking out 9 hole scores for each player?

I suspect that any difference in ratings between two 27 and 3 - 18 hole rounds would be slight in a player's overall rating.

Please prove me wrong and tell me again why would a TD want to spend literally hours and hours extra time to do this?

High SSAs = lower ratings. That is a fact.

I don't know why a TD would do all that work to break it down. You will have to ask them.

Last attempt to make you understand this...

If you shoot 49 on an SSA 50, it'll be rated about 1012
If you shoot 59 on an SSA 60, it'll be rated about 1007

First of all that's less than 0.5% difference, so is well within the noise as far as your overall rating goes, but that isn't really my point here.

In both cases you are 1 shot under SSA, but in case 1, you are 1/50th under SSA or 2%
In case 2, you are 1/60th under, or 1.6666666etc%

Does this make it clear why the former actually deserves a better rating? They are about 0.4% more under SSA by this measure. Unsurprisingly this is reflected in the ratings...

And before anyone jumps on the exact figures, I made up the 1012/1007 numbers, which would explain the slight difference. :D

I understand totally why the points fluctuate based on SSA. I would much rather be down 5 shots from someone going to an SSA 64 than an SSA 45 b/c there are more chances in the SSA 64 to make up those 5 shots.

My thoughts are it is easier to shoot a 49 on an SSA 50 than a 59 is on an SSA 60 b/c there are more chances to screw up and thus this should be rewarded better. I also don't like the fact that most if not all the best rated rounds of all time will never come on disc golf's premier courses.

It really is a double edge sword. It is easier to screw up an SSA 60, but it also provides you with more chances to recover that screw up.

I understand totally why the points fluctuate based on SSA. I would much rather be down 5 shots from someone going to an SSA 64 than an SSA 45 b/c there are more chances in the SSA 64 to make up those 5 shots.

My thoughts are it is easier to shoot a 49 on an SSA 50 than a 59 is on an SSA 60 b/c there are more chances to screw up and thus this should be rewarded better. I also don't like the fact that most if not all the best rated rounds of all time will never come on disc golf's premier courses.

It really is a double edge sword. It is easier to screw up an SSA 60, but it also provides you with more chances to recover that screw up.




You're looking at it as more opportunities to screw up, that's why you get it wrong.

Take 2 equally rated players playing identical golf. They both go round an SSA 50 course in 46 shots (4 under).
One stops, the other plays the first 9 holes again. His course is now SSA 75, but if he comes in with 71, his form has clearly dropped, as the form he had in 18 holes suggested he should pick up 2 more shots, and finish on 69, 6 under.

Once you're under SSA, you can't look at it as more places to screw up, it's more places to gain shots.

Maybe it would help if the calculation of ratings based on deviance from SSA weren't a linear formula (i.e. the farther you get from SSA, the more each stroke is worth). That would actually make more sense, since each successive stroke becomes harder and harder to obtain.

As a very basic example, Imagine an open 18-hole course, total length of ~4500' -- every hole is less than 300'. Let's say SSA for this course is ~40. Someone could shoot a "perfect" 36. And get a rating of ~1056 (4 strokes x 14 pts). However, if someone takes an ace and shoots a 35, their rating would only be 14 points higher (1070). Even three aces in a sick-round that defies all probability wouldn't quite hit that magic 1100 number (7x14=98). I know ratings get more exaggerated as SSA goes up, but the example is still relevant. Getting two aces in an an otherwise "perfect" all-birdie round is statistically FAR LESS LIKELY than one ace (or zero aces), yet those extra strokes are worth the same amount as all the other strokes.

On an SSA 60 course, its obviously impossible to start talking about scores in the 30s, but the principle is still the same. There is still a theoretical limit (just like 36 is the theoretical limit in the example above), where the odds of shaving each additional stroke off the score card become exponentially less likely.

And to talk about the other end of things, let's take a course with an SSA of 54 to make it easy. Every stroke is worth 10 points. A couple of juniors shoot 100 and 102 -- for SSAs of 540 and 520. A couple of pros shoot 50 and 51, for SSAs of 1040 and 1030. The PERCENTAGE difference between each the two sets of scores is the same (2%), however the PERCENTAGE difference between the two sets of SSAs is very different (less than 1% for the pros, and almost 4% for the juniors.)

Whereas, I think if most people were watching from the sidelines, they would say that it's a whole lot easier for a junior to pick up two strokes than for a pro to pick up one stroke at that point. At the high-scoring end, every stroke is a tremendous coup; at the low-end, every stroke doesn't matter much.

Really, the way ratings should work to be most accurate, is that there should be an algorithm that weights scores so that every stroke is worth more at the low end and less at the high end. To work best, this would have to be combined in some way with the present system, so that on a course with a lower SSA, the weighting process would accelerate more quickly because of the score compression mandated by the easier design.

All this is not to say that the current system doesn't serve it's function -- it does work well to show who is playing better than their friends on a consistent basis. All I'm saying is that it could be a much more accurate system, one that would solve the issues of low-SSA and high-SSA courses having different probabilities of resulting in awesome, top-10 scores. This would balance the system so that the most challenging courses would have an equal (or better) chance of producing the highest-rated rounds.

hauntingly familiar...

Yep, I believe in sticking up for accurate math.

With the current calculations, I think even Chuck would agree that it is impossible to compare apples to apples with round ratings on high SSA vs. low SSA courses.

But in my thinking, if we admit the system is flawed -- that the math ceases to be useful at some point -- that is an indictment that the system isn't accurate.

Part of the purpose of math is that it should be a universal language. If the math gets bendy and ceases to be reliable at the ends of the spectrum, then that suggests there is a better way to do the math.

I know that statistics can be distorted by outliers, but that isn't the case here. We're not talking about an injured 1000-rated player shooting 800-rated rounds with their left hand; we're talking about every single player being subject to the same distortion an inaccuracies on courses that are at the fringes of the SSA spectrum.

It's like we're looking through a fish-eye lens to calculate scores.

But its the same in most sports. Just because the progression along the linear scoring curve does not require linear progression in skill level, does not mean that the system is defective.

In ball golf, right now I'd be about a 26 handicap. I've been a 16 handicap. To go from a 26 to a 16 isn't that much work, just practice and playing a couple of times per week. To go from a 5 handicap to a +5 handicap takes a monumental leap in ability.

In bowling, to go from an Obama level, 37 bowler to a 100 bowler... 1 or 2 lessons and a little practice, from a 200 average to 263 average... a little tougher.

From a Shaq-like 52% free throw shooter to a... well ok, he still stinks... I could make more than 52%... but I'm not 7-1 325...

The "perfect round" scenario outlines the ratings discrepancy quite well, similarly to what Tim referred to earlier. The examples below use Chuck's (length/foliage factor + 30) equation to calculate the approximate SSAs. Chuck's foliage factor range for "none" is 355-385. I will use 385 (absolutely no foliage) for these calculations.

"Perfect round" is used to describe a round in which the lowest reasonable score is achieved on each hole, discounting aces and throw-ins from the fairway. They happen occasionally, but are statistically insignificant since they would not change the SSA enough to change the points per throw.


Course A has 18 250' wide open flat holes. It measures 4500' and has a SSA of 41.69. Each stroke at this SSA would be roughly 14.5 points. A perfect round would be 36 and would be rated around 1084.

Course B has 18 625' wide open flat holes. It measures 11,250' and has an SSA of 59.22, with each stroke worth roughly 7.5 points. A perfect round of 54 would be rated around 1040.

The round on the lower SSA course is rewarded with a much higher rating (44 points!)

Did the lower SSA course play tougher?

But in my thinking, if we admit the system is flawed -- that the math ceases to be useful at some point -- that is an indictment that the system isn't accurate.



We know that players with ratings that are 50 points apart shoot about 7 throws different in score at SSAs around 68 like the USDGC. If a 975 player beats a 925 player by 7 on average, a 1025 player beats a 975 player by 7 on average, why would a player who beats a 1025 player by 7 on average not have a 1075 rating? Just because we don't have a 1075 player doesn't mean the throws should be worth any more the lower the score gets on any course regardless whether it's high SSA. There's no proof or even hint from actual results that there's anything wrong with the math "at the fringes" to indicate some adjustment is required. The only problem with the numbers on 68 SSA courses versus 54 SSA is perception, not the actual math nor the actual results.

Tim,

"As a very basic example, Imagine an open 18-hole course, total length of ~4500' -- every hole is less than 300'. Let's say SSA for this course is ~40. Someone could shoot a "perfect" 36. And get a rating of ~1056 (4 strokes x 14 pts). However, if someone takes an ace and shoots a 35, their rating would only be 14 points higher (1070). Even three aces in a sick-round that defies all probability wouldn't quite hit that magic 1100 number (7x14=98). I know ratings get more exaggerated as SSA goes up, but the example is still relevant. Getting two aces in an an otherwise "perfect" all-birdie round is statistically FAR LESS LIKELY than one ace (or zero aces), yet those extra strokes are worth the same amount as all the other strokes."

Wonderfully said.

You've "driven it down" (the SSA thing) to the exact point where each stroke is NOT equal (maybe in CK's rating world but not in the "real" world)! At that exact point, the difference between a 36 (totally doable) and a 35 (would need at least 1 ace) is huge - and certainly more than the difference between a 36 and a 37. In one case you're allowed to miss one putt; in the other you have to make a tee shot. CK, do you honestly believe that the chance of making 18 of 18 putts instead of 17 of 18 putts is equal to making 17 of 17 putts and 1 ace? I have scored an 18 (on a 9-hole course where the whole distances vary from about 140 to 250 feet - moderate to sparse foliage). There will be a crap-load of 19's on that course (and a lot of 18's too) before anyone beats my 18 with a 17.

With the current system, I believe that the difference between a 54 and a 36 (on such a course) would be the same as the difference between a 36 and an 18 (14 points each stroke). If I have this right, then a 54 is the average for an 804 player. I sincerely believe that the difference between an 804 player and a 1056 player is less than the difference between a 1056 player and god. And if the rating system doesn't reflect this, then there IS something "not quite right" at one end of the ratings.

Karl

Why does the difficulty in getting a higher rating actually need to be linear for the methodology to be valid?

Just because there are "beyond perfect" performances that are possible. Besides, if the SSA is 40 on the course that is described, I would expect that in any suitable sample size field of 1000 rated golfers, there would be multiple aces within the field.

Just because it is easier to go from an 850 golfer to an 875 golfer than it is to go from 1015 to #1 in the world doesn't mean the math is broken, just that the math is not linear as a function of ability.

Is there any reason it needs to be linear?

We don't consider ratings for courses less than 41.4 SSA to be valid due to the limiting factor of 36. It's possible the minimum should even be higher than that. And because 30 shots per 18 holes appears to be a constant on actual results, the math appears silly at that level but still may be accurate. However, the apparent flukiness of aces seesm to make not seem right.

The thing that all of you are overlooking in your attempts to force some sort of curve into the numbers is that we have actually tested a variety of functions that might better represent what's going on and linear was the best fit. Not only that, the verification of the function is whether you can predict what the ratings will be before the rounds are thrown and that has proven to work for validation. So the linearity isn't hypothetical, it's actual. Had it not been linear, we would have been using another equation long ago that actually matched the results.

The function probably shouldn't have been strictly linear to prevent the theoretical possibility of a negative rating. The limit of the R(t) should be 0 as t approaches infinity. But then try to explain that equation to the masses.

The ratings calculation itself is not linear. However, the value of one throw has the same number of ratings points at a given SSA regardless how many throws or how few throws are thrown.

Correct, not strictly linear. Where the 'majority of cases fall', yes - but at the lower end, having it linear is ridiculous (as my scenario a few posts back alluded to).

There is a VERY precise point - due to the difference between putting a disc into a basket and driving a disc (from the tee) into a basket - at which the "line" go exponential / logistic. It occurs (on an 18-hole course (of all par-3s) that is short) at 36.

Since the difference is SO different (between holing out a putt or holing out a tee shot) AND the chance of it happening is SO infrequent, some sort of formula / odds / differential should be able to be made up to account for this (...what are the chances of making a hole-in-one and the rest birdies vs. making all birdies; making 2 hole-in-ones and the rest birdies vs. a hole-in-one and the rest birdies; etc.). This would be better than a 'totally linear' model which we have now.

I'm not saying "scrap the model", just make one end of the model curve proportionately to the odds of the event(s) happening.

Karl

Chuck,

Your,
"the value of one throw has the same number of ratings points at a given SSA regardless how many throws or how few throws are thrown"

How can you accept this statement (as being worth anything) when considering the "low end" of it? Again, I ask, do you believe the difference between a 37 and a 36 is equal to the difference between a 36 and a 35?

Karl

The function probably shouldn't have been strictly linear to prevent the theoretical possibility of a negative rating. The limit of the R(t) should be 0 as t approaches infinity. But then try to explain that equation to the masses.



krupicka makes a good point about the other end of the linear system. If I hand a disc to a 5-year-old and sign him up for a tourney, let's say he finishes the round, but it takes a LOT of strokes to do it. So on an SSA 42 course, with a stroke value of 14 points per stroke, this kid shoots a 114... so he's 72 strokes above the SSA (72x14=1008), so his round rating is a NEGATIVE 8. (1000-1008) This also shouldn't be possible. Admittedly, this scenario is highly, highly unlikely, but the possibility helps show the limitation of the current, linear formula.

Chuck, I don't think anyone is trying to say that the current system isn't useful. It works as an excellent predictor of who is likely to score well, and how players are likely to place relative to one another, HOWEVER, when there are visible problems with both ends of the ratings formula, it demonstrates that there is room for improvement. So far, all examples given suggest that the linear system is distorted.... What you need to remember is that even in a fish-eye kind of distortion, objects maintain their relative order with relationship to one another. In the standard Mercator-projection maps, where Greenland looks like its the size of South America -- the distortion doesn't rearrange the globe. Countries still maintain their positions relative to one another -- Canada is still north of the U.S., Spain is still east of Portugal, etc... but the sizes and shapes of each country become distorted. That's what we're saying happens with the current, linear rating system. Obviously, the Mercator-maps are helpful -- we can tack them up on the wall and learn where countries are in relationship to one another -- and the current PDGA ratings system is helpful -- we can tack up a list of players and see who is better in relationship to one another -- however, the amount of information we can take away is limited by the distortion.

We don't consider ratings for courses less than 41.4 SSA to be valid due to the limiting factor of 36.



Chuck, how did you all arrive at 41.4 as the cutoff SSA? That information seems like it would be helpful to this discussion.

Karl

It has to be, otherwise how do you distinguish between a 35 with one ace, and a 35 with 2 aces and a missed putt?

What's wrong with a negative rating?

960 tells me I'm about 4 shots off world class, -10 tells me I'm about 101 shots off, the maths works entirely correctly.

One last set of thoughts for now:

In the example we discussed on the previous page of a short course with a theoretical "best" score (barring aces) of 36, I think it is worth mentioning the fact that on EVERY course -- no matter how difficult -- there is some theoretical lower limit, barring aces (or field aces), past which gaining each extra stroke becomes exponentially less likely. So this means that the discussion of ratings distortion for great rounds becomes valid for courses with ANY SSA -- not just for courses with low SSAs. On every course, there is some reasonable point where attaining each successive stroke gets harder and harder to do because of all the variables involved.

The example I'm playing through in my head is Hawk Hollow, John Biscoe's private course in VA. I'm using it since, at an SSA of between 59-60, it is a perfect example, and a course a number of people here are familiar with the course because of the VTI and the HHO. I think the tournament course record is 52 (Biscoe can correct me if I'm wrong) -- which is a 1058 round on a day with a round SSA of ~59.25, with every stroke work ~8 points.

I've played at Hawk Hollow several times, and I know that if I took my best score from each hole, I'd have a 46 (1106 round, based on the tourney scores discussed above... though I think my personal best score there is a 62), however, the course has so many challenges, chances for roll-aways, trees, bad-kicks, etc... that the odds of any person getting that many perfect shots in a row is just ludicrous -- it defies all probability. I'd say a theoretical "perfect round" out there - barring field-aces -- is probably a 45 -- and the odds of that are so slim its staggering -- far, far more difficult than a 36 on the theoretical short, open course talked about above. Past that 45, every stroke would require a throw in or an ace, or something else equally miraculous... and even getting that 45 would be almost inconceivable -- a day with a combination of skill, good fortune, and 400' park-jobs past trees, OB and elevation.

I'd say that anything past about a 58, every successive stroke shaved off a score becomes more and more difficult. I think anyone who's played there would agree that the level-of-play difference between a 52 and a 58 is far greater than between a 58 and a 64, because there is so little opportunity for mistakes as the scores creep lower.

Think of it as having a probability similar to rolling two dice. Your possible combinations of numbers are between 2 and 12. However, the two least likely numbers are a 2 or a 12, because in each case, only one combination of both dice achieves that score(1,1 and 6,6). Getting a 3 or an 11 is easier, because you have two combinations of disc that can reach each total (1,2;2,1 or 5,6;6,5)... obviously 7 is the easiest number to roll, because there are the most combinations of dice that total 7.

The odds of getting a perfect score on a course are similar, but even more exaggerated. A perfect score on a course (on HH, let's say that is 45), requires zero mistakes (or one "miracle" for every mistake -- but as we've discussed, miracles (aces) are statistically insignificant)... a round with one mistake is much, much easier, because there are 45 chances to make one mistake. A round with two mistakes is easier again, because there are even more possible ways to reach your total score. A round with 3 mistakes even easier, etc., etc., etc. The bottom line being, that rolling a "perfect" round is obviously exponentially harder than an "almost prefect" round.

Please don't misunderstand that I'm saying that disc golf is random -- disc golf is a game of skill. But I'm using dice as a way of helping people think through the probabilities of finite numbers that add up to reach a total.

What I'm learning from this thread is that there are some people here I would LOVE to play dice with. You can have numbers 2 and 12 and I'll be generous and take just the number 7 -- because all numbers are equal and the distribution between numbers is linear, right? You should win because you have two possible totals and I only have one. Little does you know what you're in for... How about it anyone, dice? :)

Neil,

I agree with you - and disagree with you.

Your...
"What's wrong with a negative rating?"
...I agree with you; their is "nothing wrong" (technically) with having a negative rating as, and you correctly point out, the math DOES work. My only comment would be that the range of numbers (which normally contain our scores) is shifted too much toward 0 (and thus allowing a rating 'to go negative' IF the score were that bad). Had the SSA # for a course been set to "equal" that of a "good Pro" - and that # been 10,000 (instead of the 1000-rated person it is now) - the chance that anyone would get a negative round would be almost nil. Shift the scale where each stroke equals 100 points (on average) and there's no problem - mathematically - although it's akin to hyper-inflation in the currency community!

Your 35 w/1 ace vs 35 w/2aces and a missed putt is interesting but I don't think "bursts the bubble" on my "theory". That would just lend itself toward that missed putt being VERY expensive...the price you have to pay when approaching perfection (ace-ing every hole). You get a wicked big increase in your round rating when you do get 17 birdies and 1 ace, why shouldn't you drop wickedly big for throwing one ace "away"?

Karl

I believe this is the 3rd or 4th time I've posted these exact sentences, but why not throw them out there again:

Per-stroke point ratings based on the significance of each throw are 100% valid &amp; necessary for a decent round rating system. On a "humpback" curve of score distributions (with round score on the x axis &amp; number of players achieving said score on the y axis), the steepness of the curve (i.e., the significance of a particular difference in scores) should be equal to the points per throw (at each point on the curve), as opposed to the number of points per throw being constant. To argue otherwise is just plain wrong, as the significance (and desired round rating difference) of the difference between the 499th &amp; 500th throws or between the 34th &amp; 35th throws (on an SSA54 course, for the sake of example) is nowhere near the significance of the difference between the 50th &amp; the 51st throws.

I really don't understand why we keep harping on the issue of scores of 37,36,35 with an ace, etc. because this has already been addressed by Chuck:


We don't consider ratings for courses less than 41.4 SSA to be valid due to the limiting factor of 36. It's possible the minimum should even be higher than that

Jeff,

Oooh, it's been "addressed by Chuck"...therefore it's gospel. Yeah, right.

Think for yourself Jeff.

Let's talk logic and the real world here in a language that all participants can understand, not statistics-ese where someone can "baffle the masses with BS".

Do YOU Jeff believe that the difference (in one's ability to shoot) between a 35 and a 36 is equal to the difference between a 36 and a 37?

And any system (argument) which cannot support itself at both ends (the extremes) is one with "holes in it". We're trying to point out the holes - so they can be fixed for the future. Then we WILL have a rating system which will work for all situations...one which we don't have now if "...don't consider ratings for courses less than 41.4 SSA to be valid due to the limiting factor of 36." Gee, so we have a rating system that is "valid" only on SOME courses. Wonderful. I'm glad that these MB discussion exist; they bring out interesting information....

Karl

Do YOU Jeff believe that the difference (in one's ability to shoot) between a 35 and a 36 is equal to the difference between a 36 and a 37?



I could care less because there are very very few disc golf courses with SSAs that allow such scores and the ratings produced by such are not valid to begin with. Personally, I would never even want to play such a course.


And any system (argument) which cannot support itself at both ends (the extremes) is one with "holes in it". We're trying to point out the holes - so they can be fixed for the future.



There are many many mathematical &amp; scientific equations which fail at the extremes, especially the low end, but are still considered perfectly valid. Considering the increasing trend is to design disc golf courses with higher SSA values, this will become increasingly less of a "problem" in the future.

Jeff,

Your...
"Personally, I would never even want to play such a course."
...is your opinion which, while I don't agree with it, I can accept.

Your...
" and the ratings produced by such are not valid to begin with."
...just shows me that you agree that the ratings system can't handle ALL scenarios possible.

Your...
"There are many many mathematical &amp; scientific equations which fail at the extremes, especially the low end, but are still considered perfectly valid."
...is a very dangerous statement to make, especially when talking about the 'mathematical' part. I'll give you that "scientific equations" can (in rare cases) be "valid" and still fail at the extremes, but your paraphrased -

""mathematical...equations which fail at the extremes...are still considered perfectly valid""

- is VERY misleading...almost like you're trying to get the reader to believe proven mathematical theorms which fail at the extremes can be valid. Don't go there. Science CAN fail at the extremes; math, if proven valid, by its nature CAN'T fail at the extremes - otherwise it wouldn't be "proven valid".

Karl

An SSA 41.4 or whatever it is still challenging.

It is a challenge to birdie every hole.

math, if proven valid, by its nature CAN'T fail at the extremes - otherwise it wouldn't be "proven valid".


Though I agree that the round rating system is (at best) weak math, especially at its edges, your statement above is overly broad &amp; is easily refuted by Godel's First Incompleteness Theorem:

Any effectively generated, sufficiently powerful (i.e., at least capable of expressing elementary arithmetic) theory cannot be both consistent and complete. In particular, for any consistent, effectively generated formal theory that at least proves certain basic arithmetic truths, there is a statement that is true, but not provable in the theory.


Therefore, all math MUST fail at at least one of its extremes if it is sufficiently powerful to even express basic arithmetical truths.

An SSA 41.4 or whatever it is still challenging.

It is a challenge to birdie every hole.



Agreed, and if that's what floats your boat, then more power to you. Personally, I lost interest in that years ago. For local play I typically only go to courses with a minimum SSA of about 54 (Little Lehigh, Jordan Creek) and when I road trip, it's only to world class high SSA courses like those found in Pittsburgh.

I personally enjoy courses with an SSA around 49 - 54 for casual play b/c its not pitch and putt but its not whip my butt hard.

Ray,

I could muddy the waters by stating statistical examples of the liar's paradox too, but let's get back on subject.

I think that most of us agree that the present rating's system serve SOME purpose (different purpose(s) to different people); my point is that at the low end this system needs work. CK has even mentioned that at some level (41.something, I forget) the SSA stuff is NG. Why that point? If there is ANY point at which the system "can't handle it", then the system has a fault. Either change to another system (which CAN handle it) or "tweak" the system to handle it. Either works in my opinion. We've now "uncovered" that the low end (really easy courses) can't handle it - and god knows if the "SSA" was REALLY low what would happen. And while there may not be too many such courses, ANY system which can't handle ALL scenarios needs work. No, you just can't ignore this deficiency by saying 'you can't play on this course because our rating system can't handle it'...that's a cop out. Either the system has to handle all situations (stressed by testing it at its limits / ends) or you get a new / revised system.

Karl

There are many many mathematical &amp; scientific equations which fail at the extremes, especially the low end, but are still considered perfectly valid. Considering the increasing trend is to design disc golf courses with higher SSA values, this will become increasingly less of a "problem" in the future.



yes, there are some scientific models which start to breakdown at the extremes, but I don't think any scientist is opposed to trying to find a better system -- an equation that fits all the variables and remains consistent at all levels. Just because a system is useful doesn't mean it is ideal.

I don't think anyone is saying that the current ratings system isn't useful -- it is incredibly helpful. But just because something is working okay is no reason not to look for ways to improve it. Especially when the existing model has obvious places where it's usefulness starts to fail.

And as I was trying to point out in the Hawk Hollow example earlier, the problem of low-end distortion is NOT solved by higher-SSA courses. With increasingly challenging courses, the low-end score just shifts upwards and how you define a "perfect" round score becomes a little blurrier.

Higher SSA courses increase rather than decrease the need for a more consistent rating system. As courses become more difficult -- and there are more places for mistakes -- the odds of a "perfect" round become lower and lower. And therefore the difference in overall skill required to shoot a perfect round as opposed to an almost-perfect round becomes greater and greater.

Going back to the "perfect round" of 36 -- (barring aces) -- there are ~289 ways to score a 38, there are 18 different ways to score a 37. There is only ONE way to score a 36.

Going back to the Hawk Hollow example, getting that magic perfect round of "45" that I suggested requires a lot more skill than getting the 36 above, because some of the deuces required to get that 45 are ~450' shots with trees, elevation and OB in the way; and some of the 3s required to get the 45 require multiple drives with trees, elevation and OB in the way -- much harder to string that many perfect shots together than to get the perfect 36 on the theoretical open course. While the best players in the world might make a couple of shots like that in a round, none of them expect to make 100% of those shots every time. So the odds of "perfect rounds" get lower and lower...and therefore each successive shot should be given greater weight.

Ray,

I could muddy the waters by stating statistical examples of the liar's paradox too, but let's get back on subject.


If you don't want the waters muddied by correct conclusions, do not stir the pond with incorrect assertions.

Maybe its easier to see the challenges when we look at a two-round, tourney-day scenario.

It has always amused me is that two players could theoretically tie in placement for a tournament, with one player shooting a much higher average rating than the other for the tournament (lots of variations of this).

Simplest example: a two round tournament where you play a really hard first round (SSA 69), and a really easy second round (SSA 42). On the hard courses, each stroke is worth ~7 points. on the easy course, each stroke is worth ~14 points.

So player A shoots a smokin' first round with a 59 for some 1070-rated golf, then shoots a crappy 2nd round on the short course, with a 47 (930-rated golf).

Player B shoots a solid first round with a 69 (1000-rated golf), but then comes back to shoot a smokin' 2nd round with a 37 on the short course (1070-rated golf).

At the end of the day, players A and B are tied with a total score of 106 for the tournament, but player A averaged 1000-rated play for the day; player B averaged 1035-rated play for the day. They both walk home with the same cash in their pocket; who is the better golfer?

(you can run the numbers a little different and easily come up with a scenario of two golfers who shoot the same average-rating, but whose scores are very different -- just take golfer B and have him shoot a 79 the first round (930) and the same 37 the second round (1070), so that both players average 1000-rated golf, but one of them is 10 strokes behind the other in a one-day tourney. And as we all know, 10 strokes in a one day pro-tournament is HUGE. It's the difference between hundreds of dollars and taking home nothing. Don't try to console the player who is 10-strokes behind, and still shooting 1000-rated golf, that the ratings system always works.

A weighted distribution of stroke value instead of a linear stroke value could eliminate this scenario. If the players shoot the same total score, they should have basically the same average rating for the day. If they players shoot very different scores, they should have different average ratings at the end of the day.

Thoughts?

Very simple:

Per-stroke point ratings based on the significance of each throw are 100% valid &amp; necessary for a decent round rating system. On a "humpback" curve of score distributions (with round score on the x axis &amp; number of players achieving said score on the y axis), the steepness of the curve (i.e., the significance of a particular difference in scores) should be equal to the points per throw (at each point on the curve), as opposed to the number of points per throw being constant. To argue otherwise is just plain wrong, as the significance (and desired round rating difference) of the difference between the 499th &amp; 500th throws or between the 34th &amp; 35th throws (on an SSA54 course, for the sake of example) is nowhere near the significance of the difference between the 50th &amp; the 51st throws.



Or did I already say that? :cool:

Ray,

Not incorrect...and practical ! Something, which you - by stating Godel's First Incompleteness Theorem as being relevant in this case - aren't being here. By stating such a theorem as "part of the discussion", you are stating things tantamount to saying that 'virtually no matter what you say mathematically it can be proved to be untrue'. Unless you wish to give everyone a serious boost in mathematical theory knowledge and have many hours to discuss the merits of such, you'd do better to look at the situation at hand - a ratings system which has (at least) a hole which can't be handled.

You're obviously intelligent (on this subject) enough to know about a very archane theorem, I will take it that you're also intelligent enough to know when it is practical to bring it up (when in the midst of mathematicians talking about some "heavy stuff") and when it's relevance is virtually non-existent (as it is here when a bunch of disc golfers are talking about how to make a ratings system better).

Karl

As you can see by my most-recently-previous post, I almost agree with the crux of your statistical argument. (Both you &amp; Tim seem to be dancing around the fundamental unsoundness of the round rating system, almost but not quite hitting it head on.)

However, no matter the context, when one makes the overly-broad statement:

math, if proven valid, by its nature CAN'T fail at the extremes - otherwise it wouldn't be "proven valid".


particularly considering the USE OF CAPS FOR EMPHASIS, one must be prepared to be refuted.

Ray,

Your assumption that the scoring spread would be distributed along a Gausian curve MAY be corrrect for a higher SSA course (I don't know, I haven't looked at the numbers), but I doubt very much that it would be if done on a low SSA course. The curve would be a long cog railway up the steep high-rated player side (as they approached "perfection") and a long run out as the rating got lower - as someone / more than one person could "blow up" and do really poorly but even the best could only get so low.

And if I'm reading you right, your contention that the significance of points around the 50, 51 range should be more than that of points in the 35, 36 range or 499, 500 range would lead - if implemented - to a rating distribution of ALL players to be linear...and there would go your typical player distribution rating curve. I think it's an easier fix to shore up one end of the existing system than to weight each point-to-next-point spread differently than the ones next to it.

Karl

Ray,

Yeah, I do use caps maybe too often. And while it may be a mathematically true refutation, it is not a practical refutation (which, to me in this case, is more important). My intent is to try to get people to think logically, practically, and clearly on the subject at hand. As long as we're doing that, different ideas are good.

Karl

Tim,

Wow. I never "did the calcs" to see that such happens. I guess I played very few tourneys that played on 2 courses (as such). Yup, a weighted distribution of stroke value would "solve the problem" but I wouldn't want to be Chuck / others who play with the ratings system to keep it together if this was the case. I would believe it would be very hard to do such. Would it have to be based on concrete SSAs (something we don't really have)? Would it be done only intra-tournament? Good find / realization...but one which I don't have a clue how to fix (at this time).

Karl

Ray,

Your assumption that the scoring spread would be distributed along a Gausian curve MAY be corrrect for a higher SSA course (I don't know, I haven't looked at the numbers), but I doubt very much that it would be if done on a low SSA course. The curve would be a long cog railway up the steep high-rated player side (as they approached "perfection") and a long run out as the rating got lower - as someone / more than one person could "blow up" and do really poorly but even the best could only get so low.


I never stated nor implied a Gaussian distribution, merely a "humpback curve". No matter the curve, though, it seems obscenely obvious to me that the slope of the best-fit curve should be directly proportional to the ratings difference between two scores.


a rating distribution of ALL players to be linear...and there would go your typical player distribution rating curve


I'm not sure what you're trying to get at here. Please clarify.

Ray,
You sound like you've done a lot more high-level math than I have -- my background is in some math, but my strongest skill in regards to math is common sense and seeing where things don't seem to line up. Any chance you've got the background to come up with a formula that would allow the theoretical players mentioned above to come up with the same average rating at the end of the day -- the two who shoot the same two-round total score, but have average round-ratings for the day that are 35 points apart under the current system.

I have to believe there is some sort of "humpback" formula, as you suggest, that would allow players shooting the same total score to have approximately the same average rating at the end of the day. I understand it won't be perfect -- too many other variables in the system for any formula to be perfect -- but I would be curious to see what that formula would look like. I just don't have the background to know that whatever crack-monkey calculation I might came up with is being done correctly.

Thanks to you and Karl for the entertaining discussion. I don't expect the system to change -- Chuck seems pretty grounded in his current thinking (and I certainly don't want to volunteer for his job) -- but it's certainly worth trying to explore ways the system could be improved... someday. And it helps keep the brain awake as I'm trying to wrap things up here at the office.

Ray,

Earlier you stated:

"(Both you &amp; Tim seem to be dancing around the fundamental unsoundness of the round rating system, almost but not quite hitting it head on.)"

So grab a hammer and help us belt it! ;)

As for the responding to your other questions, my (this) computer is stupid slow tonight (I'm still on dial-up...duh); I'll have to answer it tomorrow.

Karl
Ps: It's taken me over 8 minutes to just type this!

Karl

I don't consider Chuck's word to be gospel, but I also trust him not to come on here and lie. So with that in mind, read this again, rather than the bit that suits your argument:


The thing that all of you are overlooking in your attempts to force some sort of curve into the numbers is that we have actually tested a variety of functions that might better represent what's going on and linear was the best fit. Not only that, the verification of the function is whether you can predict what the ratings will be before the rounds are thrown and that has proven to work for validation. So the linearity isn't hypothetical, it's actual. Had it not been linear, we would have been using another equation long ago that actually matched the results.



On top of this, any and all of the variations you point out at the extreme ends of scoring are so far in the statistical noise as to be irrelevant, compared to the variance you can get from a single gust of wind/roll away/cut through.

I'm going to ignore the 35/36 - 36/37 bit of your argument because it muddies the water with the ace issue. Instead it's more useful to talk about the linearity nature at higher numbers.
So, why is the difference between a 53 and a 54 the same as between a 54 and a 55? The answer is because it's about measuring scoring spread, not the absolute value of the shot.

You state that it is harder to save each progressive shot, therefore each should be valued higher than the last. But in reality, it's also easier for someone else to drop that shot and maintain the scoring spread that is reflected in their ratings.

It is just as difficult for a 990 rated player to turn their 55 into a 54 as it is for a 1000 rated player to turn their 54 into a 53. Equally it's just as hard for the 900 rated to turn the 64 into a 63.

Neil,

Not only have you asked me to re-read CK's statement - which implies I'll have to address every part of that statement, but then you add a whole lot of your own stuff. That's a LOT of writing...but I'll try. My comments [in brackets].

I.
CK's 1st sentence:
"The thing that all of you are overlooking in your attempts to force some sort of curve into the numbers is that we have actually tested a variety of functions that might better represent what's going on and linear was the best fit."
[His "to force some sort of curve into the numbers" is incorrect; at least when refering to my point. I'm just "forcing" a different curve ONLY at one end (the low end). I haven't looked deeply enough (maybe Ray or Tim has) into the "middle" or the "high end".]

II.
His
"...the verification of the function is whether you can predict what the ratings will be before the rounds are thrown"
I don't necessarily believe is true.

III.
His
"So the linearity isn't hypothetical, it's actual."
I don't doubt it a bit...he just needs to "attach" a non-linear / more exponential tail at the low end.

IV.
His last statement doesn't really state much.

V.
Your
"all of the variations you point out at the extreme ends of scoring are so far in the statistical noise as to be irrelevant,
[Not irrelevant at all if you're trying to hold a tournament on a very low SSA course.]
compared to the variance you can get from a single gust of wind/roll away/cut through."
[I believe you know that any mathematical model should be set up on its own merits...and THEN take under consideration other factors which might affect that model. If I were to use your thought process, then I could introduce the factor that I have a wicked bad cold now, can't see straight, my equilibrium is all goofy, and THAT would affect scores even more than an occasional roll-away. You can't use outside factors to set up the system.]

VI.
Your
"I'm going to ignore the 35/36 - 36/37 bit of your argument because it muddies the water with the ace issue. Instead it's more useful to talk about the linearity nature at higher numbers."
[Again, I have no problem (although with Tim's / Ray's revelation, I'm getting more inquisitive about such) with the linearity nature of the higher numbers...it's the low end I have a problem with.]

VII
You say
"You state that it is harder to save each progressive shot"
[I did? Where? I don't recall such!]


Of course - as is VERY typical in MB "discussions" - one will pick out the 1 point that someone wants to attack and focus on that. And if person A states 10 things, and person B "refutes" (and I use this term very lightly) 1 - but not the other 9, then person claims "victory". Silly. All ya'll are missing the big picture (while trying for your little battle wins)...and that is we have a decent rating system that needs a little tweaking (at least at one end) but too proud people refuse to acknowledge it. But (everyone) go ahead, take your shots.

Karl

Ps: Tis' a really good discussion - no matter which "side" you're on...and I think we'll all become a little bit more aware of things because of it. Thanks to all for letting me be a part of it.

There are only a handful of courses that have SSAs below 42 and have PDGA events. We don't know if the system breaks down below 41.4 because we don't have enough data to really test it. We developed the ratings sytem and still operate it on a per hole basis so 41.4 is 2.3 per hole and our original reference value is 2.8 per hole which is the 50.4 that's been discussed before as the average of the four courses in the 1998 Cincy Worlds.

The problem with looking at ratings for one round in general is the fact that the unit of scoring is one shot where scores with finer resolution can't be thrown such as 2.5 or 2.86 on a hole. here's no such thing as scoring 0.8 on a throw. So what happens as the course SSA gets closer to 36 is that the ratings value of one shot gets much higher than at say 50 or 65 SSAs. This makes luck factors such as aces much more of a factor in making bigger jumps in individual round ratings.

However, there's no indication the actual calculations would break down if all of a sudden we started playing a bunch of courses with SSAs between 32 and 40. There are courses with 60-130 ft holes (Tiki) where aces are more common. If higher rated and lower rated players play that course 25 times, their individual round ratings would be more volatile and jump around more than on a higher SSA course. But there's no reason to believe that the average of those ratings would be much different than on higher SSA courses. The only variance might be that one player happened to be a better or worse long putter/approacher which was more apparent on these really short layouts.

Karl, this is a good discussion, and I'm not attacking you. I don't agree with you, but that's ok :D

Anywho...


"...the verification of the function is whether you can predict what the ratings will be before the rounds are thrown"
I don't necessarily believe is true.


Not sure what to say about this. If the model predicts something, then actual results show the same thing, seems to me you have a good model!


Not irrelevant at all if you're trying to hold a tournament on a very low SSA course.


Personally, I don't think PDGA tournaments should be held on anything much less than SSA 50, certainly 45 as an absolute minimum, and that's for purely competitive reasons, rather than ratings, but that doesn't invalidate your point I guess.


"I'm going to ignore the 35/36 - 36/37 bit of your argument because it muddies the water with the ace issue. Instead it's more useful to talk about the linearity nature at higher numbers."
[Again, I have no problem (although with Tim's / Ray's revelation, I'm getting more inquisitive about such) with the linearity nature of the higher numbers...it's the low end I have a problem with.]


The point is that there is no difference though, the same feature occurs at range from higher SSA also. Your example is SSA 40, and the 36 as a limiting factor. But simply adding a throw onto each hole of that course makes it an SSA 58, with 54 as the limiting score (basically 18 very easy but un-eagle-able par 4s). Add/remove throws from the course as you like, and you can see that for any particular SSA, there will be a 'perfect' score. With very poorly designed courses, the perfect score will be close to the SSA, with others it may be greatly different, and this will vary massively and unquantifiably from course to course.

If you have an unquantifiable feature, I can't see how you can account for it in the maths...


"You state that it is harder to save each progressive shot"
[I did? Where? I don't recall such!]


Apologies, looking back that was Tim... But I think that is kind of what you are saying though, that the closer you get to the 'perfect' score, the harder it gets, and therefore the shots should become more valuable?

Neil,

You stated "I don't consider Chuck's word to be gospel, but I also trust him not to come on here and lie."

On 9/15 CK said:
"We don't consider ratings for courses less than 41.4 SSA to be valid due to the limiting factor of 36."

And on 9/17 he said:
"We don't know if the system breaks down below 41.4 because we don't have enough data to really test it."

This may not be lying, but is sure is confusing at best and potentially contradictory. It is statements like this that bother me - when we lay credence in what a certain person says / does (that their actions affect us - in some minor way...like ratings) yet their words mince together in such a way as to make (at least me) think there's as much BS as there is truth to it.

Karl

There's no reason to believe there's a problem with the math for courses below 41.4 SSA regardless of your ace example. It's no more lucky than a spit or a ricochet in the basket that changes a score by one throw. We can operate as if there's no problem unless enough data develops that indicates a problem. But it's a non-issue because these courses are rarely in tournament play and it's not like there's a trend to play more of them.

We've chosen to not use results from courses under 41.4 SSA for best round record purposes. That's the only place we won't use them. However, the PDGA has decided to still produce round ratings for players on courses under 41.4 SSA as a customer service. That's no more suspect than doing ratings on 50 SSA courses with only 2 propagators. Again, it's done for customer service, not necessarily for accuracy of the ratings.

Neil,

You stated "I don't consider Chuck's word to be gospel, but I also trust him not to come on here and lie."

On 9/15 CK said:
"We don't consider ratings for courses less than 41.4 SSA to be valid due to the limiting factor of 36."

And on 9/17 he said:
"We don't know if the system breaks down below 41.4 because we don't have enough data to really test it."

This may not be lying, but is sure is confusing at best and potentially contradictory. It is statements like this that bother me - when we lay credence in what a certain person says / does (that their actions affect us - in some minor way...like ratings) yet their words mince together in such a way as to make (at least me) think there's as much BS as there is truth to it.

Karl



But in the same paragraph on 15th he said "the math appears silly at that level but still may be accurate". So it's not actually contradictory. Regardless, I'm not getting into a discussion of Chuck's pros and cons, I don't know him, and I guess you don't either.

Do you want to address any of the actual content of my last post?

Neil,

God blast it! I just type something and the stupid thing dissolved into thin air. Crap. I'll try again...

Your statement...
"But I think that is kind of what you are saying though, that the closer you get to the 'perfect' score, the harder it gets, and therefore the shots should become more valuable?"
...I'm totally not sure about, but I'm pretty sure that at the 36 point, the "linearity" of the model goes exponential. I wish I could draw on a computer :( then I could show you what I mean. But try to imagine it by looking at your keyboard. Mentally draw a line from the right Ctrl key to the M key to the F key (which is my 36) and then takes a wicked upturn to the F3 key and beyond. Or, if you wish, from / key to M key to V key to E key to F2 key (if you want to keep the initial line flat). This is what I believe are the odds of such happening (in real life) when one gets a 36 (F or V) vs a 35 (F3 or E), etc.

Bad picture, but the best I can come up with now.

Karl

Chuck,

You're the "king of ratings" so you can, I guess, do whatever you wish but knowing that you have a model with a flaw at one end - one end which, although not used too often, IS used - and that you choose to do nothing about it is a cop out.

And your statement...
"It's no more lucky than a spit or a ricochet in the basket that changes a score by one throw."
...is laughable. I have 6 career aces; I have had hundreds of spit outs. I guarantee every player out there has more spits than aces.

You continually back up your "defense" with these such statements and you wonder why people "question" the validity of your work. Some of it is great; some of it is good; some of it needs tweaking...this latter one you just can't accept.

I have nothing further to say.

Karl

I understand what an exponential looks like...

My point is that this doesn't happen anywhere in particular in the scoring range. If one of the holes on your very easy course has a 60 foot hole, well there are going to be a lot of aces on it. If it's a dull par 4 course like I described, there is a point where you require a fairway ace to improve your score, and it isn't 36. This point is different for every course, and it's mathematically unquantifiable. So it's pointless to attempt to get it accounted for.

I just want to point out that weighting strokes more and more as you get lower will not fix the situation with players tying in a tournament and having different average ratings. If a player beats you by 1 stroke in the middle of the bell curve the first round and you shoot a 35 to his 36 the 2nd round you will have a better average rating than him. The only way to have the same average rating for the tournament would be to weigh each stroke the same on every course regardless of SSA, which I think we can all agree would be a step in the wrong direction.

I never notice the weird things about our ratings system until they happen to me.

I'm rated 935, the lowest rated Advanced player on the planet, give or take the decimal points we cannot see. I recently played a two-round tournament on an 18-hole course where I shot two rounds that average out to 955. If I understand the ratings system correctly, and if I play no more tournaments that count in whatever ratings update that those rounds count in, shooting two rounds at 20 points above my rating should make my rating go down. :eek:

I could be wrong because I only have a weak understanding of ratings, but it looks like those two rounds averaging 955 will be double weighted and count as 4 rounds in my average. Adding those two rounds will cause 12 older rounds to drop, but those 12 rounds are weighted as 15 or 16 rounds because all or most of them were on 24-hole courses. Those 12 rounds are rated around 945 on the average. So those 12 rounds actually do more to pull up my rating than the two higher rounds that are replacing them. By my math my rating falls by about a half of a point by adding those two rounds. Maybe my reward for playing a hot-for-me 955 rated tournament is that I'll be eligible to resume playing intermediate? :p

I was looking for a thread titled "What's Wrong About Ratings" but maybe that was on another website. To me it feels wrong that shooting over my rating should ever lower my rating. If there is any interest in the topic, especially if anyone wants to check my math, I'll follow up with a post about what I think is wrong with ratings that they produce this kind of anomaly and how this could be corrected with a slightly different ratings system.

Remember that at least for amateurs, the primary reason for ratings is to get players competing in a division at their current skill range. For those right near a ratings break such as 935, having some ratings volatility up and down is a good thing so players aren't consistently stuck either at the bottom of a higher division or at the top of a lower division.

We could almost eliminate the problem where your next ratings update drops when your oldest higher rated rounds drop off by retaining round ratings for two or even three years before dropping off. But fast improving players would get discouraged with slower increases in their rating and everyone else would be upset that they were bagging in a lower division because we didn't update their rating fast enough.

Bruce, I'm rated 935 as well. So you are not the lowest rated Adv. on the planet. Just one of a group!

My favorite ratings story is shooting a 59 (-13) at Hudson Mills during the 2003 DGLO. My round WAS rated 1001 (I think), my first 4-figure rated round. Some guy named Barry was the only other person to break 60 that weekend ;0. A few months afterwards, Mr. Kennedy "re-worked" the formula and the round dropped to 999. That one hit me below the belt! I'd call -13 a HOT round that was better then what the rating says it is. Then, in 2005 at Bowling Green, I shot a 52 (-3, I think #2 is a 4) at Hobson Grove. Not the best score, but that round gets rated a 1002! Whatever, I know how I feel about how I shot.

As it turned out, the addition of those two rounds that averaged 20 points over my rating lowered my rating by TWO points. I have an intermediate rating and I'm not afraid to use it.

Ha! My last Int. was 2002 Bowling Green. I'll never go back down. Anyway, I'm 943 now!!! No longer a benchmark. Yay!