Why do we use ratings from an entire YEAR. Why dont we use the 20 or 30 most recent rounds to get an accurate look at what a players rating is at the time. I just dont think the ratings are accurate when you are counting rounds that you shot 12 months ago or even 6 months ago for that matter. I understand this would prolly cause ratings to rise and fall more rapidly but it would also be a better description of how a player is playing at that time. Which in essence would eliminate alot of the sandbagging that goes on if the ratings were updated maybe once every 2 months and only using the 20-30 most recent rounds. Even if they got updated the same as they do now it would still give a better representation or a players ability AT THAT TIME.

Isnt that the reason we have ratings in the first place??? To give an accurate representation of a players ability at a specific point in time.

Any and all comments are welcome. Just wanted to see what everyone else thinks. Good bad or indifferent

Thanks guys

can someone please tell me why i have a rating of 940 when i'm always on the last card and shot over par most rounds ? :confused:

That is what i am talking about. You prolly have a bunch of HOT rounds from 12 months ago that are offsetting your rating so that it is not accurate. I think if we used the 20 most recent rounds everones rating would be A HECK OF A LOT MORE ACCURATE. Then people wouldnt be able to play Intermediate all year when they are really playing to a 935 rating but have a bunch of 800-900 rounds from 12 months ago.

Does this make sence to anyone else???

Your questions and observations are valid for new players that are on the steep part of the learning curve which I would guess is typically the first 3 years of playing regularly. Your question seems to also assume that most players get 20-30 rated rounds in every 2-6 months. This is not an accurate assumption.

There have been several posts recently that have acknowledged that you get the least benefit ratings-wise from playing in Adv where the players (on average) are rapidly improving and their ratings lag their actual ability. Conversely, you get the most benefit playing Masters since, on average, their skills/abilities are on a downward trend. So, theoretically their rating is higher than their actual ability.

Moral of the story: If you care about ratings, start playing at 40. :D

yeah when i used to play adv.am i used to shot really well,but when i turned pro master my game went south for the winter and just keep going so i can see your point pimp.

I am saying 20 most recent rounds. Whether that be over 6 months or 12 months. So if you only get 20 rounds in in 6 months then you only have 20 recent rounds to go by. The amount of rated rounds played would all depend on the player him/herself. On the other hand for the touring pros and the highly competitive Ams with a packed schedule who play 20 or more tournaments a season and play 20-30 rated rounds every 2 or 3 months. It would give a more accurate ratings for those people so they know who and what they are up against. Even if you get 20 rated rounds in in 6 months then your rating will still be more accurate then if they based it off your last 40 rounds over 12 months.

I just think we need to use more RECENT rounds in the rating so that it is an accurate representation of a players ability.

I think 20-30 is a great numberbut i would also say the ratings would need to be updated about ever 2 months during the TOURING season and maybe once during the winter months.

yeah when i used to play adv.am i used to shot really well,but when i turned pro master my game went south for the winter and just keep going so i can see your point pimp.



I just think during the heavy competition season it would give you a better view on who you are going to be playing against at your tournaments and it would let the lesser players move back down to where they should be on the AM level and it would also FORCE the SANDBAGGERS to move up out of Intermediate and Recreational.

Everyone complains about SANDBAGGING and this would be the best way i can think of to solve that. SANDBAGGING happens when someone gets good fast and then stays down in the lower division so they can dominate untill they are FORCED to move up to a high AM division. This would help to force them up faster then the system does at the moment.

cb, I would not be surprised at all if the Ratings System is changed such that recent rounds are given more weight when calculating the overall player rating.

I mean sure its not going to affect the people who play 5 tournaments a year, but i do think it will help the touring pros and the highly competitive AMS gain a better perspective on where they are in relationship to their peers and it will also show improvement at a faster rate then the system does now.

I just want to get everyones opinion on this and see where everybody stands as far as how often and what they use to update the ratings.

I played 20 rated tournament rounds in August alone, that means my rating would only account for how I played that month!? The PDGA doesn't even update ratings that offen, I don't think that would work very well for people who play a lot of tournaments.

Im saying 20-30 rounds. I also say they should be updating the ratings alot more often. As in once a months for the heavy competition months. Then maybe NOT AT ALL during the winter and low competition months.

Maybe it should be the top 85% of rounds since the last ratings update. So that would be your top 85% of rounds for the past three months if they dont want to update it on a regular basis. In my opinion the more you update the ratings the less number you have to deal with and the easier it would be but i know thats a problem becuase TDs dont always get their TD reports in very quickly but thats another subject.

Maybe it shouldnt be 20 rounds. Maybe it should be top 85% like it is now but only for the past 3 months not the past 12 months. Dont you agree that the ratings should be based on recent rounds and not rounds shot any more then 3 to 4 months ago???

The average number of rounds in a year for all PDGA members, who have played at least once in a year, is around 13. That's about 1 round per month. So doing the ratings every 2 months like we already do from May thru September means we process about one event per person every two months.

One change we are looking at is to not go back and get any more info more than 12 months earlier if a player has at least 8 rounds in their current 12 months. (Right now we go back more than 12 months if they have less than 20 rounds.) This will increase the newness of ratings for those faster rising Ams where the current process delays their ratings advancement.

If you guys are going to change wouldnt it be better to only take the past 3 or 4 months rounds and do a ratings as long as there are 20 rounds in those 4 months and then go further back if needed??? Because arent the Ratings mostly in place for the highly competive players in the sport??? That way the ratings for the fast rising and highly competitive AMs and touring pros would be more up to date and a better representation of their skill level at that point in time.

One reason we do 12 months is that matches the tour season when we do the prior yearend final ratings in Feb. There are also more members with fewer than 12 rounds in a year than those with more than that so 12 months of data is needed to get them a decent average. We also want to avoid different policies for players with different event frequencies if possible.

We are looking into providing a 'hot shot' index for players who have a lot of events and their rating is rising quickly. Look for this next year. This will involve weighting more recent rounds more heavily in the rating average.

We are looking into providing a 'hot shot' index for players who have a lot of events and their rating is rising quickly. Look for this next year. This will involve weighting more recent rounds more heavily in the rating average.



That sounds like a pretty sweet idea. Would this be like what i was suggesting but only for people who a high volume or tournaments during the season???

Would this be like what i was suggesting but only for people who a high volume or tournaments during the season???



Yep. It would kind of be a 'buzz' list (i.e. who are the hot shots people are talking about). You would have to have a certain number of rounds for this calculation to be made. We would only publish the names of the players who were fast rising (not falling) so they might get some advance pub when traveling to events. If they played in the division where their current official rating was and not where their hot shot rating put them, you could give them the sandbagger raspberries :D

Do you guys have any idea what that magic number would be??? Not that it matters because im planning on playing in 30+ Sanctioned events next year so i would think that if i play well i will be included in this calculation.

Sounds like a great idea to me. :D

I think the hot shot rating will be a great addition to the ratings system and should help keep golfer from sandbagging in their divisions. This would also help the pros that play and the fans that watch them so they know how are hot at the time.

Way to go PDGA!!!! :D!!

It's too bad for me a Millz that those 800 rated rounds sound like they'll be around for a while.

Thats why i am saying like the past 6 months would be better then 12.

Not everyone plays 30 tournaments a year. Some people only play 5 or 6 (or even less). Only keeping 6 months worth of ratings might be as few as 4-6 rounds for those players.

Then you do what they do now and go back till you have enough rounds to do a rating. Those people must still have rounds from 6 years ago in there rating then :mad:

I think they should just keep it as 1 year

I think they should just keep it as 1 year

I hope we'll just do whatever the ratings people want us to do. They have done a great job this year with getting new ratings out close to schedule, and dealing with the occasional special situation. They should get whatever they want if it isn't a cash bonus! :D

I was going to post this on another thread but it would have been thread drift, so I'll toss this out on a dead ratings Q&A thread.

Here is one that occured to me today while looking at some stats: A player who plays a lot of events and sometimes shoots 980 one round and 860 the other at the same 2-round tournament will have a higher rating than a player who plays a lot of the same events and shoots the same tournament scores by shooting 930 one round and 910 the next, even though they tie on score that way. In other words our current system underrates the more consistant player and overrates the streaky player, even though at the end of the day they are tied.

If you understand why this is, you have a decent handle on how ratings work. If not, I won't tell you why until Rodney or Chuck or someone who should know better says it is not so!

:D

This is true for some players under the current system, but will be resolved with the modified process being developed for 2005 starting with the mid-May update.

A simple answer in the long run is that the inconsisent player is dropping scores that would have a greater dampening effect on their rating than the consistent player.

The wider the seperation of the top scores and bottom scores in an averaging scheme that drops the bottom scores, the greater the effect dropping the lower scores has on the total.

i.e. if you have 20 rounds scattered evenly between 930 and 911 (about 1 point seperates each rated round) and drop 3 (15%) you effect the average by 1.6 points (920 to 921.6).

If the same spread of scores were produced by the inconsistent player between 980 and 860 (about 6 pointss seperate each rated round) and drop 3 (15%) you effect the average by 9.5 points (920 to 929.5).

The bad news is that by this explanation, my lowly 914 rating is inflated. From all the rounds in my current rating, it should be 905. My fluctuations are worse than his though, from 814 to 977.

That is it.

The ratings are awesome.

Yea, I always thought that was kinda dumb. If you drop the bottom 15% then you should also drop the top 15%, but then again, no one would want to have their best rounds dropped. So I guess it works out the way they are being calculated right now. However, I did just shoot 2 rounds rated about 90 apart yesterday, so actually this is a good thing, it might actually help my rating since the bad round is getting dropped :)

Yea, I always thought that was kinda dumb. If you drop the bottom 15% then you should also drop the top 15%, but then again, no one would want to have their best rounds dropped.



If you think a little more on this, you'll understand why only the bottom is dropped. Many seem familiar with the bell curve for stats distributions. However, that's not true for players real scores and for the scores as received on TD reports. The 15% drop does two things. One, when the ratings first started in 1998, TD reports and the process for handling them was much less automated and accurate than today. We were concerned that a certain percentage of reports would be inaccurate, especially in providing the proper course configuration for the divisions. The 15% drop was partially a 'fudge factor' to automatically deal with this issue and not have to constantly track down TDs for corrections.

The second reason is the player score distribution issue. Players can only influence the bottom of their curve, not the top of their curve. At any point in time, a player can tank their round for a variety of reasons such as going for ace runs or 'tin cupping.' On the other hand, a player cannot go out and say, 'I think it's time for a Climo round' and actually deliver it. So, it makes no sense to drop any top rounds just the bottom ones.

However, starting with the mid-May update this year, no rounds will be dropped unless they are below a certain statistical value relative to your individual rating. We can do this now because TD reports and the process itself have gotten much better. In addition, we now have stats on what constitutes a 'normal' round in a typical player's distribution versus an 'abnormal' round. This allows us to just exclude those abnormal rounds where needed.

The mathematically inclined might look at this new process in light of Bruce's post above and think most active player's ratings will go down if rounds are not dropped. This would be true. However, we plan to tweak an internal factor in the calcs so that most player's ratings will stay close to their current number. Those who have been steadily improving will perhaps see a bigger jump up than they would have under the current system.

So my 160+ point spread between good and bad rounds will be considered normal for me ?

Your 814 will be abnormal and dropped for the mid-May update and the 837 will be too old to be included along with some other rounds that will be too old by then.

Isn't it also the case though that the player with rounds less streaky may also simply have plateaued and become anchored to their skill level (not that a new factor might not increase or decrease it) and the streaky player is more likely to be still learning and on an upcurve?

In that sense, i think the player with more consistent rounds should have a more difficult time improving to a higher level since they have demonstrated more soundly where exactly their game is at. Does that sound right Chuck?

With the new process, players who have maybe been consistent for a while will at least have the chance to move their rating upward quicker should they have the opportunity to practice more and enter more events.

My query was more along the lines of individual determination than my specific case. Meaning will each player based upon his/her past rating swings have different lower bounds, or will there be a standard formula ?

i applaud the pdga for continuing to sit down and try to improve the ratings system to be a more accurate reflection of players' abilities. It's not an easy task by any means.

But I definitely think adding more weight to more recent rounds is a good idea. I just crunched my own numbers, and while my rating is currently a 920, I discovered that if I shoot two 920 rounds at my next event, my rating will go UP three points because of dropping an additional sub-900 round from the calculation. It seems odd, but I suppose that's how the system currently works: and truthfully in the long run, would probably work out to be the most accurate but would also temporarily give my rating a swing. While, if instead, adding weight to my recent rounds would give me a more accurate depiction of how i'm playing *right now*.

Hey Chuck....

How many years can a person go without playing and not suffer a ratings penalty?

A person's rating will not change if they haven't played except when we change the calc process like we'll be doing with the mid-May update. But even then, ratings of inactive players should remain about the same. No one retains any rounds more than 24 months older than their most currently rated round. So, if a player like Stokely starts playing tournament rounds again, his 2001 and older rounds for his outdated rating will disappear in the first ratings update after the new rounds, and his new rating will only be based on those new rounds.

One, when the ratings first started in 1998, TD reports and the process for handling them was much less automated and accurate than today.



This reasoning is agreed to be no longer valid thanks to the great efforts of many people.


The second reason ... At any point in time, a player can tank their round for a variety of reasons such as going for ace runs or 'tin cupping.' On the other hand, a player cannot go out and say, 'I think it's time for a Climo round' and actually deliver it. So, it makes no sense to drop any top rounds just the bottom ones.



I missed this jump in logic. I agree that dropping bad rounds is a good idea, but I do not see how outrageously good rounds should then be counted. Is it because it is harder? This doens't make sense to me. The rating should represent an average, and to do this - bearing in mind that we all agree that droppng bad rounds should be done - you would need to drop bad rounds and good rounds equally in order to find an average rating. The idea of dropping the top 15% and bottom 15% seems like a good solution, so over your last 20 rounds, you would drop your best and worst 3, as they are probably statistically abnormal. Only dropping the bottom gives a false impression of a player's rating, and potentially an unequally false impression. imo.


This allows us to just exclude those abnormal rounds where needed .



Abnormal rounds can be good and bad.


The mathematically inclined might look at this new process in light of Bruce's post above and think most active player's ratings will go down if rounds are not dropped. This would be true. However, we plan to tweak an internal factor in the calcs so that most player's ratings will stay close to their current number.



I am definitely jumping into this thread mid-stream and understand that I may view player ratings incorrectly, however, to me, a 1000 rated player will, on average, shoot world class par. (please correct me if this is wrong)

By not subtracting the best scores and by adjusting an internal tweak so that adjusted ratings do not drop seems to make my above presumption about 1000 rated players even more untrue.

I wish my bowling league would drop just my bad scores!

In my opinion, we should drop top and bottom scores and determine player ratings as accurately as possible.

Also, imo, the player ratings should drop so that they more accurately represent how a player shoots. It is better to make this adjustment sooner rather than later. Let the ratings drop, and let them drop fairly.

Maybe the rating should be calculated as a median of the last 20 rounds & not a mean. That'd discount statistically insignificant rounds (high or low) without introducing the conundrum of deciding what number (if any at all) high and/or low rounds to leave out of the calculation.

steve, i think the crux of the argument is that it is easier to have a terrible round than it is to have a brilliant round.

further, terrible rounds can exist for a number of reasons, including deliberate/pre-meditated, makeup of field/propogators, and (of course) simply having a terrible day.

however, there is only one way to have a brilliant round - and that is to shoot at the very top of your skills. its not possible to force yourself to shoot 5 strokes better, to make all the 35 footers, or to park all the upshots. when it happens it means the player executed at the top level available to him, without exception.

a brilliant round is truly representative of a player's skill, whereas a terible round is not.

ray, as most tournaments dont have an odd number of rounds finding a true median would be near impossible since you would always be left with TWO rounds in the middle of your numberset. Which even then would require taking the mean of those rounds.

Also, say someone is a wildly inconsistent player who throws 1000 rated rounds half the time, and 900 rounds the rest of the time. Assuming we DO have an odd number of rounds to work with, if that player had one more 1000 rated round than 900-rated rounds you're saying that player should have a 1000 rating? That doesn't make sense by any stretch of the imagination, as we are obviously not looking at a true 1000-rated player.

Cromwell,
The statistical standard for calculating the median of a set containing an even number of elements is to take the mean of the median two elements of the set, which would still, of course, eliminate statistically insignificant (high or low) rounds, as well as yield a rating of 950 for your wildly erratic imaginary golfer, which is probably exactly what he deserves. :D

steve, i think the crux of the argument is that it is easier to have a terrible round than it is to have a brilliant round. a brilliant round is truly representative of a player's skill, whereas a terible round is not.



That is a valid point, but I wonder if this hits on what the ratings are trying to measure. Are the ratings a measure of how well a player CAN shoot, or a measure of how well a player DOES shoot.

If the ratings are a measure of how well a player CAN shoot, yes, we should only include the player's best rated round.

However, if we are trying to truly represent how a player DOES shoot, it would seem to me that we would need to include an even range of scores (if you remove the bottom scores, you should remove the top scores too).

There are currently 4 players rated at 1000. Their average rating over their past 20 rounds is: 995, 987, 988 & 994. On average, none of the players rated at 1000 shoot World Class Par.

Again I ask the definition of a 1000 rated player. My guess would be that on average a 1000 rated player would shoot World Class Par. But this is currently not the case.

The goal in generating our player ratings is to determine the average of all rounds defined as normal for a player. A 'normal' round is one where a player is presumed to be attempting to shoot the best score they are capable of shooting. An abnormal round would be one in which a player has given up shooting for a good score and is perhaps running for aces or taking abnormal risks like tin cupping. In addition, 'abnormal' penalties such as those for starting a round late with par+4s or even for misadded or late scorecards might be involved it the round was already on the low end of 'normal' for this person.

Since we don't want to get into mind reading, we need a mathematical way to include normal rounds and exclude abnormal rounds to the best of our ability. In 2005, we're tentativley looking at excluding only rounds that are more than about 2.5 standard deviations below a player's rating. That would mean only the worst normal round out of about every 100 'normal' rounds would be dropped.

I understand your point. You are looking for a "normal average" as opposed to a real average. If someone tin cups a round, and shoots a "non-normal" round, the ratings want to discount this because the player wasn't really trying. However, could it also be argued that a player that shoots a tin cup round is a tin cup type of player - ie erratic - and not including these rounds is a disservice to those who play good consistent golf and would not earn any drops?

If someone shoots an ace-run round, or a tin-cup round, or a round hungover, that was their choice. They shot the round, the round should be included, becuase by shooting the round they have demonstrated that they are the type of player who will, at times, do just this type of "non-normal" shooting.

(and I presume by tin cupping you are talking about going for it on every hole as opposed to playing hungover, alhtough I guess it doesn't matter, because we are all gonna remember that 12 - which I would argue is "normal" for tin cup.

friZZak

TTrim

steve, i see where you're coming from, but consider this: with ratings increasingly used for serious large scale events (ie Mid-Nationals) including a beyond crappy round leaves the system open to manipulation. a 920 player rated player on the uptick could enter a tourney and blow four rounds out at 850 in order to keep his may rating just below 925. that makes him a top seed for a national title instead of scrapping for dfl in the top division.

that scenario alone makes a good case for the elimination of truly terrible rounds.

A 'normal' round is one where a player is presumed to be attempting to shoot the best score they are capable of shooting. An abnormal round would be one in which a player has given up shooting for a good score and is perhaps running for aces or taking abnormal risks like tin cupping.



Steve, 'normal' is our chosen definition as specified above which is when a player is shooting for best score. Even if a player 'normally' is prone to ace running or tin cupping escapades, that falls outside 'normal' as we've chosen to define it. And, Pat indicates the primary reason for dropping those rounds, which is manipulation. Other PDGA members have more at stake in knowing your normal rating, as defined by our system, than you do in the sense that the purpose is to do a decent job determining a player's appropriate division for fairness to other members who play in divisions below the top division for their age and gender.

I'd actually like to see more of the bad rounds dropped. Granted I am a long time ball golfer and the USGA uses the best 10 of your last 20 rounds to calculate your handicap. It is a measure of potential, not normal. If we used a system similar there wouldn't be any need for bagger calls anymore.

I just don't think that the new way would be fair to people like me. For example I have a 956 rating and rarely shoot below my rating. I am consistantly in the 960-975 range with a few good rounds. So would this system drop my high rounds even though I am consistantly above my rating?

actually using the best 10 of your last 20 rounds would make your rating go up. Think of it as dropping the worst half of your scores.

Hey chuck... how hard would it be to tweek the ratings software to give someones ratings at Ctiers,Btiers,Atiers,NT's, etc.....? Say ... somebody like.... uh... me? ;)

just curious.... plus I know how much fun math is to you.

-tB

Yeah I know your way would make me like a 980 player (that would ne sweet)but I'm talking about the other thingy. :D:D

Todd, are you saying calculate a person's rating using only their C-tier results and another rating using only B-tier results? Players could do that for themselves right now by looking up their results and doing several tier averages. I think with Dave Gentry coming on the PDGA staff, we'll be able to do some interesting reports in the future on a more regular basis, similar to the ones Rodney used to do occasionally.

I talked with Dave yesterday and we'll be looking at using 2005 as a base year for future comparisons. We may eventually go back and recast old data using the new process so our records can be valid back to the beginning in 1998.

The new calculation process will not throw out any top rounds for a player. In fact, I don't know of any sports rating system that ever removes a player or team's best rounds. If top rounds are removed, it will be due to an error in the reporting process that caused the whole junior under 10 division to have rounds over 1000 for example.

just for fun i ran some samples using the formula chuck described - dropping rounds that are lower than the round average minus 2.5 times the standard deviation of the rounds.

here's some examples of the results:

Barry Schultz was 1036, now 1031
Nolan Grider was 1006, now 998
John Maiuro was 979, now 974
Claydog Brunit was 922, now 914
Pat Brenner was 920, now 913
Gary Duke was 914, now 907
Dan McGough was 888, now 880

its a small sample using the 12/22/04 ratings detail. looks like the effect is roughly the same across the various ranges. of course, the fewer rounds outside the SD range, the less effect the calculation change will have.

i wonder if there will be a downward amplifier effect due to propagators having lower ratings to begin with... but i'll leave that to chuck to worry about :)

That looks about right. Rodney and I figured that the adjustment factor in the formula would need to add back the equivalent of about 7-8 points across the board. Do that to Pat's calcs and the ratings remain almost the same. BTW, that 2.5sd is a sliding value that changes at different rating levels, which we've determined, but won't publish. It's not based on a specific player's numbers. Number sleuths will be able to figure it out pretty well once the player stats are available online and the rejected rounds are indicated.

who's gonna tell nolan to mail back his tour card ??? :D

Okey doke. I guess I did not understand that the goal of the system is to set a rating system so that it is harder to manipulate, rather than set a rating system which accurately represents how a player shoots relative to World Class Par.

I was under the misconception that a 1000 rated player shoots World Class Par on average - when in fact they tend to shoot about 1 stroke worse. I can live with this, as long as we all understand that this is the case.

To be honest, it never occured to me that a player would pay money to shoot poorly so that they can win a World Championship of mediocrity. But if this is what we are afraid will happen, then by gum, make the guy play bad for 20 rounds in a row to qualify for that mediocre division.

Wow! I missed a whole page of conversation. My above post was replying to the bottom of page 5.

After reading the top of page 6, I can absolutely understand the system (although the 1000 rated player not shooting WCP seems non-logical). Especially with the ball golf example of dropping half the rounds (the bad ones, that is) to determine a handicap. Potentail vs. average.

My bowling league uses our averages, and then the team with the lower average gets 80% of the pin difference. Again, this is because it is easier to shoot worse than better, and it accounts for potential vs. average.

I am happy dappy on this subject, and for some reason felt the urge to type this all up so everyone else understood what they apparently already got.

I'm going to go back to organizing the MSDGC now. You guys keep up the good work.

I was under the misconception that a 1000 rated player shoots World Class Par on average - when in fact they tend to shoot about 1 stroke worse. I can live with this, as long as we all understand that this is the case.




They actually do shoot SSA (it's not WCP any more) the way the numbers are calculated. If ten 1000 rated players go out and all shoot 55, then the SSA is 55 and their round ratings will be 1000. Now that we're hardly dropping any rounds in the new process, 1000 will mean 1000 slightly more than it did under the current process.

Rodney has been one of the driving forces toward getting this correction implemented in our process. It's possible that this few point differential has cumulatively lead to a slight inflation in the ratings over time and why players believe they get better round ratings when playing with higher rated players. The new process should adjust these issues to the extent they are indiscernable

chuck, i'm a bit unclear about when/how the new calcs will be rolled out. i understand the equations, just not when it will start to be used and if we're switching over completely, running parallel, or what. could you clarify?

The new calc process will start with the mid-May update. No need for parallel systems. We're working on the formulas now to synchronize the process so there will be little change for most players. The mid-Feb update planned for posting between Sunday, Feb 20-Tuesday, Feb 22 will be under the current process.

Chuck,

I have 19 rated rounds with 4 more being added this upcoming update. Since you keep the best 85% of your rounds that leaves me with 19.55...Will they round up to 20 rounds or round down to 19?

I have my rating at 946 with 19 rounds, and 944.45 at 20 rounds...I'm not sure if I did it right, but the 4 new rounds are from the Hawaii State Championship: 907, 946, 907, 1006...Do you know which rating will stick?

Thanks,

Dave

Those Hawaiian values are unofficial round ratings that will likely be slightly different when processed and remember that no 2005 events are included in this next update. The mid-Feb update is always the official yearend ratings for the prior year so no 2005 events will be included. The current formula excludes a minimum of 15% so that's how to determine the number of rounds when more than 20.

The current formula excludes a minimum of 15% so that's how to determine the number of rounds when more than 20.



I think I understand, but would the formula round up to 20 or down to 19 in my case?

1000 will mean 1000 slightly more than it did under the current process.

No need for parallel systems. We're working on the formulas now to synchronize the process so there will be little change for most players.



Hello Chuck,

Confused again. in response to me, you stated that since fewer rounds are being dropped, 1000 will mean 1000 more than it currently does.

And then in your next post, you talk about the new formula and a potential tweak so that current ratings will not move much.

These two statements seem contradictory.

DD, 19.

SD, under the new calcs, a player with a 'stable' rating of 1000 will shoot SSA and their average rating should actually average 1000 since almost no scores will be dropped. You were correct that under the current system, a stable player actually averages slightly below their rating but ends up with the same rating because 15% are dropped.

Then why are you "working on the formulas now to synchronize the process so there will be little change for most players."

This seems like it can't be true. Most 1000 rated players would lose 7 to 10 points, unless a correcting value (aka. additive fudge factor) is introduced.

I understand the idea of dropping bad scores, even the idea of dropping the worst half. I just want to be clear about what it means to be a 1000 rated player, and it seems like you are trying to say that the ratings won't change AND that a 1000 rated player will shoot SSA. These seem contradictory.

Sincerely,

A dense stick.

We're not getting into it until we have it tested and analyzed. So no details now, but it will work. How it works will be explained when it's ready. It's not a fudge factor, more like strawberry.

looks like the radar is being jammed sir...
strawberry?! i hate strawberry!!
there's only one person....

How are you guys addressing the various levels of tournament players out there?

Player 1 plays 4 tournaments a year on his local courses:
1000 Player Rating

Player 2 plays 30 tournaments a year, throughout the country on familiar and very strange courses.
1000 Player Rating

Who is really the 1000 rated player here?

With the current method, players that have 75+ rounds being rated in the current year, how does that rating even move up or down compared to Joe 4 rounds? The 75+ rounds might show more consistancy, but it sure seems that Joe 4 rounds can jump his ratings up or down just by playing a couple more tournaments, the 75+ person has to wait an entire year to replace those rounds with better rounds. Even then, with that many rounds the rating may bump just slightly.
Once you have 150+ college credits, your GPA is very lathargic.
Isn't the idea behind player ratings to show how that player is CURRENTLY rated per quarterly update?

How are you guys addressing the various levels of tournament players out there?

Player 1 plays 4 tournaments a year on his local courses:
1000 Player Rating

Player 2 plays 30 tournaments a year, throughout the country on familiar and very strange courses.
1000 Player Rating

Who is really the 1000 rated player here?





I don't think that they figure in the home course factor between player #1 and .. uh.. you. ;)

-tB

Only rounds 12 months prior to each player's most recent rated round are included their rating. So, older rounds are being dropped in each of the five updates. Also, with the new process we plan to double the weighting on the most recent eight rounds.

The local courses were said to show that it is easier for someone to score better on courses they are familiar with.

The point is more about both guys being considered world class players with one guy only playing 4 tournaments vs the guy playing 30 tournaments.

And about having Ratings be an actual current reflection of play.

8 rounds? That's only two 2-day tournaments. If you guys are updating every 4 months. the heavy tournament players are sometimes playing 10-16 tournaments or 40-60 rounds.

Jay - I think I read somewhere on this thread that the new ratings would only use the last 20 rounds, and that they perhaps have to be in the last year. And they are apparently going to weight the most recent 8. But I might have read this wrong because I only skim posts, unless I wrote them.

Besides, I am still trying to get the strawberry smell off my radar. :cool:

We're updating every two months during peak season and the average number of new rated rounds per active player across the PDGA in each update is TWO (13 rounds per year). Looks like fewer than 1 in 40 players will have more than 8 new rounds per update.

Is Scott Stokely still a World Class player with a rating over 1000 last established in 2001? Frankly, I don't know. I'm not sure I'd bet against him if he returned. But his rating right after the event would only include those rounds and none of his old ones. So, we would know right away his current level of play.

As far as locals being World Class players, how many people had not heard of Darrell Nodland, 'hiding' on his local courses in ND with around a 1000 rating, when he burst on the USDGC scene with a top five finish. There may be a home course effect but it's mostly meaningless unless that player is lucky enough to have an 8-round World Championship played on it. We don't send checks (yet) to players just because they have a 1000 rating, so why does it matter? They still have to earn the cash by playing.

Stokley played in 2004 and is rated 992. So I guess I'd bet against it. :D

However, it made my point that his rating immediately reflected his current performance since it's based only on his three rounds in 2004 and none from 2001 or earlier.

Chuck:

Do you know if the New Years Throwdown in Oskaloosa (January 2nd) has been reported yet? I would love for that to be in the update. Any chances??

(BTW, I got the first PDGA tourney ace of 2005!!)

No 2005 events will be in the Feb update by design. The mid-Feb update is always the final results to complete the prior year.

...Also, with the new process we plan to double the weighting on the most recent eight rounds.

ok, so the new calc drops rounds outside of ~2.5sd of average, AND weights last eight rounds. whats it mean to "double" the weighting? count them twice? if so, is that before or after the average rating for sd purposes is calc'd?

while i understand where you are coming from with the recent round weighting, eight rounds feels like not very many.

if eight rounds extends back into the third most recent event, will all rounds from that event be used?

i believe that players do better or worse on particular courses, for a variety of reasons. weighting certain rounds could easily skew a rating up or down depending on the course.

it could also magnify the skewing that even now occurs as a result of the makeup of the field (ie, a pool of ten 1000 rated players each shoot 54 on a course with a 54 SSA. they each get 1000 ratings for the round. the next day a pool of ten 920 rated players shoot 54 on the same course, but they do not get a 1000 rating for the round.)

If all the details were confirmed, I would tell them, but they aren't. All of this is still subject to testing. However, eight rounds is a half to full year of rounds for almost 2/3 of PDGA members who play at least one round in a year. So the impact will primarily be seen by boosting the ratings of the fast improving players which have been subject to a little more lag in the current system. I fully expect that not dropping the 15% of rounds in the new system will resolve the sometimes true 'urban myth' that playing in a pool with higher rated players provides a better round rating.

one man's "urban myth" is another man's dirty secret :D...especially when its "sometimes true" ;)

We really don't know for sure one way or the other. It looks like there's a bias in the direction of higher rated players generating higher rated rounds but not to any level of statistical certainty. If it were true we'd see the effect almost always. But for the few times it's been checked, it may be 60/40. There are several sub variables like age and gender of players in the mix that makes sorting it out a challenge. It's one of those rainy day projects which may be even less useful after the process change.

agreed.

is the following scenario true?

a pool of ten 1000 rated players each shoot 54 on a course with a 54 SSA. they each get 1000 ratings for the round. the next day a pool of ten 920 rated players shoot 54 on the same course, but they do not get a 1000 rating for the round

Not possible. If ten 920 players all shoot the same score, that score will be rated 920. More likely scenario, if the 920 players all shoot 62, they will get a 920 round rating and one 920 player shooting a 62 while playing with the ten 1000 rated players might get a 925-930 round rating. But the mix of women pros and Master+ aged players in those groups can flip it the other way.

Chuck:
Have the results for that tourney even been submitted yet? If so, can you tell me ratings for both rounds?
(I'm Chris Boro PDGA#23490)

I won't know the ratings until the 2005 events start getting processed in early May. The PDGA does not do the unofficial ratings you see online right after each event. Those are posted by the TD so players can see results and unofficial round ratings. You need to contact the TD to have this done if you want to see preliminary ratings.

I won't know the ratings until the 2005 events start getting processed in early May. The PDGA does not do the unofficial ratings you see online right after each event. Those are posted by the TD so players can see results and unofficial round ratings. You need to contact the TD to have this done if you want to see preliminary ratings.



Thanks!!

chuck, you missed my point on that last scenario, but thats ok. here's another one for you:

intial scenario: a player has 21 rated rounds. the lowest round is an 880, the highest is 920. from lowest to highest, each round is 2 rating points higher than the previous. (880,882,884...918,920)

new calc rating: 900 with no rounds dropped.

now replace the 890 round with an 850 round.

new calc rating: 901 with the 850 round dropped! so the rating went up by shooting a crappy round!

now change the 912 rated round to a 960.

new calc rating: 900, ie by shooting better the rating went down (because the 850 round is now included)

hmmm... whatever deficiencies there might be in the current calculation, there is no way a rating can go up by shooting a worse round, or vice versa. while its great to seek to improve the formulas, before new calcs are used for official ratings any possibility of this happening must be eliminated.

Your example is the reason we don't want to update ratings after every round because it introduces more chatter in the numbers by doing an update one round at a time. First of all, it's possible to just make up unlikely distributions that will have odd results like 10 rounds at 1000 and 10 rounds at 900. However, most distributions are not uniformly linear like your example but are more bunched like the conventional bell curve. You've also not indicated the time sequence of when rounds were thrown in your example. Since we update every few months, roughly five rounds in your example would be dropped with five new ones added before a new calc was done. Which rounds were they? If the player was improving, perhaps five from the bottom half were dropped and five somewhat higher rated rounds were added. If it was a stable player like myself, the five rounds dropped and added probably were about the same.

The one thing we haven't figured out a way to handle (but it's not clear we need to) is deliberately tanked rounds so they drop from a player's calc. It doesn't hurt anyone if a player has an artificially inflated rating. It only hurts if someone can tank several rounds just enough that none are dropped and their rating is pulled down into a lower bracket. However, this would be difficult for anyone to calculate live on the course. And, the entry fee cost to precisely tank enough rounds would not be regained from the value of winning a few prizes in the lower division before your rating bounced back up again.

coupla things... the time sequence doesnt really matter, since as you point out, we only update a few times each year. a better phrasing of the scenario would have been two players each turning in 21 rated rounds. the one with the lower average has the higher rating. i'm not convinced that this cannot happen in the real world, or even that it is unlikely.

besides, wouldnt you want a calculation that could gaurantee it couldnt happen?

now about the tanking thing... its pretty hard to shoot bad enough to drop a round using the standard deviation calc. last wekeend my first round was ranked 91st out of lifetime 96 rated rounds. (not deliberate i assure you) it is within 2.5 sd's and wont be dropped. it looks like if it was one stroke worse it might have dropped. regardless, it seems fairly easy to shoot bad enough to have it affect your ratings by not having it dropped. and weighting recent rounds more could make the problem worse, not better. i agree tho that the value of such a tactic is probably low, but then again other players may not universally share my sense of "value".


if this is true: "It doesn't hurt anyone if a player has an artificially inflated rating. It only hurts if someone can tank several rounds just enough that none are dropped and their rating is pulled down into a lower bracket. " then the solution is really simple - only include the top N rounds where N is a reasonably smallish number.

besides, wouldn't you want a calculation that could guarantee it couldn't happen?




We're not using ratings for laser guided missiles. It's to group players in fair divisions. I would hate for the aim on a missile to be plus or minus 6 degrees. But we can live with the probability that a human being is likely to throw a round within plus/minus 6 throws of their rating. Now precision remote control players with specific skill sets would be interesting...

Here is an example (http://www.pdga.com/tournament/tournament_results.php?TournID=5008&year=2004&includeRatings=1#Advanced) of one of the issues I see with the ratings system. I shot a 60 and a 60 on a 21 hole course. Yet, I got rated 18 points lower for my second round! The conditions were very similar both rounds - but if anything, the second round was tougher due to a swirling 5-10mph breeze compared to a calm morning round.

Anomaly #1: Why the difference in ratings for the same scores under the same conditions? I suspect that this was because this is a new course and it was the first time many/most people played it. So people were more familiar with it round 2 and scored better. Conclusion: ratings do not objectively measure real skill.

Anomaly #2: Scores of 51 were rated 1030 in the first round and 1017 in the 2nd. That is 13 points different. But, scores of 70 were rated 833 and 811 � 22 points different. Why the difference? Conclusion: lower scoring players get penalized less for missing a putt (or adding any stroke to their score) than higher scoring players.

I love ratings and think they are by-in-large pretty darn accurate. Since I play tournaments infrequently, my rating is what I play for. Ratings are a great way to measure my relative �score� as I compete against myself.

I find these observations more interesting than troubling, but I am nonetheless a little confused by them.

Anomaly #1 - You admitted that players maybe knew the course better the second round so your 60 was not as 'good' the second round. That's a pretty objective measure of skill presuming that's the reason. And frankly, it doesn't make any difference. You assume the conditions are the same in both rounds but that isn't the case, even if the weather is identical. Players are more tired, maybe lost some discs and/or chose different ones, tried different routes after watching others the first round. The angle of the sun changed so perceptions are slightly different. But the largest factor is normal human variance. When you consider 20 rating points is 2 throws out of 50, that's ony 4% difference and about the worst range we see. We have some courses like Knob Hill that have been within 0.1 throw of the same SSA for four years of events on that layout.

You can try this experiment which is actually pretty close to a typical player's score distribution. Roll two standard dice ten times and see how close the average is to seven for the ten rolls. Every 0.1 away from the 7.0 expected average is equal to about 1 rating point. If you have 8-sided dice, it's even closer to our real world. Then, do it another ten times. Then another ten times. Compare the range of numbers and see what you get.

We currently require at least 10 propagators to generate round ratings but that will likely drop to just 5 this year because several locations don't have 10 propagators and sometimes lower divisions don't have 10. Now try the experiment again with just 5 rolls and you'll see we've added a little more fluctuation but more players will at least get their PDGA benefit of round ratings.

Anomaly #2 doesn't happen. Someone played a different course with that 70 in that second round. There's a little ratings spread as the numbers move away from 1000 rating but not that much with an SSA difference that close.

Question: If a person has played a number of PDGA events before he joins the PDGA, do those early events count toward their initial player rating or only the events played after joining?

Question: If a person has played a number of PDGA events before he joins the PDGA, do those early events count toward their initial player rating


Nope.

The answer is normally NO but sometimes there's enough info to link up the name and PDGA number so the earlier events can be rated. But that's more of an accident than policy. We don't go back and try to link up the info because it's surprising how many people in the PDGA have the same names and even live in the same state. The PDGA number is the primary link for data. So a new player without a PDGA number would have no link back to rounds before they joined.

Chuck, is the inability to rate the Junior girls at Worlds off the am grand and am seniors who played in the same pool insurmountable or merely deep low priority? What should I tell my club members?

I have look at several peoples names and only noticed that Chuck's rating was updated. Is there an error or is my computer just not showing the update.

mine was updated

Only players who had new rounds added to their rating will show a new date for mid-Feb by their rating. Otherwise, your rating will indicate the last time it was updated which for many will still be mid-December or some earlier update.

I have look at several peoples names and only noticed that Chuck's rating was updated. Is there an error or is my computer just not showing the update.

The update was mainly to correct errors and such. The IOS Saturday rounds were recalculated changing someone's rating by a few, I read at www.discontinuum.org/forums. (http://www.discontinuum.org/forums.) Another of our IOS players lost a couple of points on a ratings correction from an Indiana tournament.

So tournaments from September 2003 wont be taken off my rating?

They don't drop until you have newer rounds for an update. If you have no new rounds, they stay there indefinitely.

Bruce, check your Rollerdoos email for an answer, sent to you more as a Board member than a Dad.

where's VPO Pro weekend ? it was posted unofficially after the event, then disappear just before the december ratings update, never to be seen again.

I never saw the report and I processed all events I received from PDGA HQ/Roger for this update including VPO-Am. Contact PDGA HQ.

cool, thanks.

btw, awesome job as usual with the updates! the ratings are one of the very best things about the pdga. some people may not agree, but the funny thing is those people can still tell ya their rating in a split second :cool:

It's what you love to hate, like looking in the mirror without makeup (not that we do or that there's anything worng with that - Seinfeld)

woo hoo...my rating went up a whopping 1 point!!!

Chuck, how far back did you go, or rather, what determined which events you re-did.

March Madness (3/04) would have given me another rated round. Kentucky State Championships (10/04) would have as well. In fact, the round that was included in the dec update (from KSC) is now missing.

Chuck
why does my player rating still say Dec 15 2004 by it.

I'm not positive but I believe Roger only went back to events that were part of the July update and moved forward from there. The purpose was primarily to include some events that had been promised for using only 5 propagators that were missed before. However, a few other events got a bonus with a few more rounds in the process. For 2005, the baseline for all events will move to a minimum for 5 propagators. So several more events and divisions will get rated rounds this year.

Chuck why does my player rating still say Dec 15 2004 by it.



Already answered up thread.

Thanx

I was wondering how my rating went down 2 points, when there are no new tournaments in my details?

:confused:

Chuck,
Last September I played in te Lumberjack Open, and was apart of a division that didn't get a rated round for one of the three rounds we played. There are 5 rated players in the division, but it did not get rated. Did this event get overlooked, or is this round not able to be processed?
Thanks

Even if you go up again I will pass you in May. :D:D

Not even if you start to play in a real division.

:p

I have 10 rounds going in that average 986.5 and hopefully a few more as well. :D:DSee you this weekend

I have 10 rounds going in that average 986.5 and hopefully a few more as well. :D:DSee you this weekend



bagger!!!!!j/k :eek: :eek:

But i'm still rated low enough for me to be ligit. :D:D

Yeah don't worry, the last update really did not change my rating at all. Even though I have averaged a 920 in the last two tournaments I played in, my rating is good enough for me to play recreational. Do they have a recreational division at Worlds? Perhaps I will enter that.

But no one has wanted to take me up on my ratings challenge. It would be them giving me points based on our player ratings. As long as they beat the spread they would be fine, but no one has wanted to do that yet. I cannot understand why, I am only a recreational player!

Yeah I had a tournament go in on the end of year update with ratings of 997 and 992 but my rating only went up 2. However when you get more and more rounds above your rating the lower rounds drop and hence you make a nice jump. :D:D

I'm 937 rated based on a gazillion rounds played in 2004. Throwing with my uninjured left arm on Saturday I threw an 840ish round and a 780ish round according to unofficial ratings generated by the on-line score posting system.

Assuming that this injury is going to be longlasting, can you tell me how well I have to shoot for my new rounds to count? Do I have to play 20 rounds or a year to completely dump my old good rounds?

I'm feeling good about the 840 round. The 790 round was not as special. I'd rather play than just hang out in the pavillion. But it would be nice to shoot my way down to a division where I could be competitive. So long as I keep improving, I could be competitive in Men's Rec by August.

Your most recent round before this weekend was Oct 2004 but you have a bunch just before then with Worlds in August. If you can play an event this year in late August with a date more than a year after last year's Worlds and it gets reported in time for the Sept update, then the Worlds and events before then will drop out. We're now only going back more than 12 months before your most recent rated round if you have fewer than 8. Your most recent round ratings will be doubled so you have a chance to get down to Rec level rating by Sept and a real good chance by the Dec update.

O.k., after reading my question I see I was sort of incoherent.

You are right I played a lot of rounds from April of 2004 through October of 2004. Let's assume that starting this weekend I resumed playing tournaments [because I did] and I'm playing left handed shooting in the 840-780 range [because I am] and that I'll be playing left handed for a year or more [because I am afraid of hospitals and people who wear masks].

Question #1: Are all these 840-780 rated rounds thrown out because they are too low compared to 937?

Question #2: If these 840-780 rated rounds count, how many of them do I need to totally replace the rounds which are currently giving me a 937 rating? 20, or a year's worth, or one after a certain date?

I think when my left arm gets a little more endurance I'll be shooting decent recreational golf and I will be able to have fun competing against rec rated players. I already have good distance at random intervals though in random directions. I just need to spend some lefty time at the driving range.

The low rounds will be suppressed for the calculation if they're too low but not deleted. At some point in the time line, your rating will be low enough that they'll be used and help pull your rating down assuming you also get more low rounds. You'll have to do the math based on when you're playing other events. Just remember that you count back no more than 365 days from your most recent round and the most recent 8 are double weighted. In theory, you could shoot rounds more than 100 pts below your current rating and it wouldn't go down because those new low rounds aren't used. However, I believe the date of your most recent round will be used as the time reference point (even if the rating itself isn't used) so you will be moving the most recent 12 months forward and dropping older rounds along the way.

so who does the ratings?

My post on this thread explains the process:

http://www.pdga.com/msgboard/showflat.php?Cat=&Number=182538&page=&view=&sb=5&o=&fpart=120&vc=1

so what about lost scores?

lost rating

Chuck, the ratings for the Monkey Paw seem a little low to me. The WCP is supposed to be 58 with a course par of 61. Thanks.

If you look at the history for Tinicum, the SSAs from the event aren't out of line with some other events. The specific pin combinations aren't shown so you would know better what setups these numbers might represent. With wind being a factor there, it's also unclear whether some of these SSAs might have been inflated with wind that was tougher than what you just had.

www.pdga.com/tournament/course_ratings_by_course.php?RatingCourseID=100 (http://www.pdga.com/tournament/course_ratings_by_course.php?RatingCourseID=100)

Your right Chuck, they aren't that much off. I thought hanging with 2 1000 rated players would yield a higher rating. :D

Next weeks results from Nockamixion will be interesting. I suspect it will rate out similiar to Rock Hill or Ideywild.

Preliminary calcs on this years Monkey Paw is that the SSA is around 53.9.

Of course there was light wind and hole 7 was in the middle position.

I'll try to have full results posted by tonight.

MADC ratings don't do all the hoop jumping PDGA ones do, so go with theirs ...