What flaws do you think there are in the handicap system?

[DaveP043]

52 minutes ago, Yukari said:

I haven't read the entire post, but one problem I see is someone who has a wide range of scores versus someone with a narrower range of scores.  Even within high handicappers, some have a relatively very range of scores, meaning they are more consistent compared to someone whose score ranges all over the place.

In the USGA system, your handicap is based on the best 10 of your last 20 scores.  Say we have two players who both average 90 on the same golf course.  The "steady" player's best 10 may average 88, while the "erratic" player's best 10  could average 86.  In a match, erratic would give steady 2 strokes.  That may or may not be fair, but its a consequence of the USGA choice to base your handicap on you better scores, and throw out your worse scores.

20 minutes ago, dave s said:

Interesting take on course hdcp from my Thursday night league.  Our league handicaps each hole played based on actual scores.  On the back 9, hole 10 has a green that can easily be 4-putted. The hole rolls up a lot of high scores due to OB left, a bunker twice the size of a school bus (and deep!) guarding 75% of the green front and a 45 degree bank of deep rough behind the green.  It's the #1 hdcp hole in terms of stroke average on our league.  Due to its short length - par 4 at about 340 yards, it's scorecard hdcp is 12.

#11 is a 165-180 yard par 3, all carry over a ravine with large, old-growth oak trees narrowing toward the green.  It's sports the 2nd highest over par stroke average on the back 9.  The card says 14 as the hdcp because it's a par-3.

When you look at how many rounds our league has played with guys who can shoot sub-par rounds on occasion to guys who rarely break 50, what we shoot out there certainly DOES NOT match what the course indicates as proper hdcp.

We hdcp our league based on what the league shoots, not what the card says which is more fair than going by the card itself.

dave

The USGA has some specific recommendations regarding the handicap numbering of the holes.

 https://www.usga.org/HandicapFAQ/handicap_answer.asp?FAQidx=25

In general, they suggest the holes be ranked based on the difference of average scores between a group of low-handicappers and a group of higher handicappers.  The Number 1 handicap hole should be the hole with the greatest difference in scores, that's where the stroke will most likely be needed most by a higher-handicap player to halve the hole in a match.  To me, this makes sense, its the fairest way of allocating strokes.  This may not be the hardest hole on the course.  My home club has an uphill 174-yard par 3 with a severely tiered green, it may be the toughest par on the course.  It's handicap number 16, and I believe that's because its hard on good players too.  It would be interesting to take the data your league has, and see where the difference in scores falls, similar to the USGA recommendation.

[natureboy]

12 hours ago, iacas said:

Do you still think that trees 40+ yards away from the center of the fairway are not considered? Because they are. And have been. The corridor is currently, IIRC, 100 yards wide. Used to be 120.

I didn't say there were not considered at all. IMO the rating system under-weights their potential impact on scoring for higher handicap players (for whom the rating system also has an extremely optimistic accuracy profile). But it's probably not a huge effect.

2 hours ago, SG11118 said:

I'd argue that a plot of scores vs handicap on a set of tees from a wide variety of unique golfers will tell us exactly what we are looking for already.  A score of each of the 0.0 golfers who golf the course will be on this plot just the same as a score for every other handicap of golfer.  Statistically you should be able to fit a line on this plotting of scores.  Where the line is when it crosses the 0.0 handicap line will be the new Course Rating for that set of tees.  If the slope of this line is different than 1 stroke per 1 handicap unit, that difference in slope will determine the new course slope (I don't know if I advocate continuing to drop the 10 worst of the golfers 20 latest scores, but if we do, the line may need to be statistically adjusted slightly to account for this). 

I guess you could argue that when we are switching from one course rating system to another, the respective handicaps of the golfers that we'd be plotting against their scores is still based on the old system, so would be slightly in question for the first couple of years.  However, by limiting the posted scores to one per golfer per year, I've tried to over-weigh the impact of guest play and traveling golfers on the system to try and bring about uniformity from region to region as much as possible.

I've acknowledged, there may be some statistical problems with some sets of tees not getting enough unique golfers over a one year period.  If another set of tees on the same course does get enough play, we could possibly utilize the CR and slope for this set of tees along with the statistically lacking set of results from the tee in question to fit in a new CR and slope for it.  If a course can't get enough unique play over a one year period, it is acceptable to stretch the time frame of the plot to multiple years.

If anything, the current system of rating courses takes length into account too much in setting course handicaps and slopes.

I see, I didn't understand that you were suggesting going off the existing HCP system to start. So you would have an initial fixed point of reference.

Not sure that a deviation from 1 stroke to 1 HCP point would give you the same 'average' slope of 113. I think it would make sense to try to maintain the familiar 'scale'.

I disagree with limiting to one score per player from a set of tees. The accuracy of the central limit theorem is dependent on large 'N'. If you took an average for an individual player it's essentially like taking all their scores anyway. This is also why I would think multi-year averages (like for Broadie's SG baseline) for the rating would be more accurate (and less volatile) than single-year numbers that might reflect unusual weather or course maintenance.

Basically, the idea seems feasible to me. But the drawback of rating a brand new course would put you back to reliance on a system based on the physical characteristics of the course so you still need something like the current rating system. To me that makes using the 'real data' as an occasional accuracy tweak / correction to the existing system preferable and less disruptive.

Edited July 22, 2016 by natureboy

[iacas]

Just now, natureboy said:

I didn't say there were not considered at all. IMO the rating system under-weights their potential impact on scoring for higher handicap players (for whom the rating system also has an extremely optimistic accuracy profile). But it's probably not a huge effect.

You continue to be wrong about how "optimistic" the ratings system is for bogey golfer accuracy.

You think it's the diameter, but it's the radius.

[natureboy]

21 minutes ago, iacas said:

You continue to be wrong about how "optimistic" the ratings system is for bogey golfer accuracy.

You think it's the diameter, but it's the radius.

I don't think so. The accuracy table for male bogey golfers indicates the width of the area into which a bogey golfer is expected to hit their 200 yard tee shot 2/3 of the time as 33 yards wide. It doesn't say half-width of the area.

Broadie's numbers for a bogey golfer's 1 sigma degrees offline (which if you look at + & - 1 sigma is equivalent to the USGA dispersion width for 2/3 of the shots) gives an expected dispersion width of ~ 49 yards. That seems like a significant divergence. The USGA's number equates to an accuracy off the tee that Broadie's numbers assign to a much lower HCP.

Spoiler

 

The Accuracy Table below prescribes the dimensions of the area into which a scratch or bogey golfer is
expected to hit shots of various lengths 67 percent of the time. It is used to assist in evaluating the effect of
obstacles around the target

 

 

Edited July 22, 2016 by natureboy

[CarlSpackler]

Last summer, the course I played often had 3 mowers broken down, so the rough got to be VERY long. Any ball hit into the rough was almost impossible to find even if it rolled just through the fairway into the rough. The course played significantly tougher, yet the slope/rating stayed the same.

[natureboy]

Just now, CarlSpackler said:

Last summer, the course I played often had 3 mowers broken down, so the rough got to be VERY long. Any ball hit into the rough was almost impossible to find even if it rolled just through the fairway into the rough. The course played significantly tougher, yet the slope/rating stayed the same.

I think this is something the daily average 'correction' that's being planned would account for.

[SG11118]

1 hour ago, natureboy said:

I disagree with limiting to one score per player from a set of tees. The accuracy of the central limit theorem is dependent on large 'N'. If you took an average for an individual player it's essentially like taking all their scores anyway. This is also why I would think multi-year averages (like for Broadie's SG baseline) for the rating would be more accurate (and less volatile) than single-year numbers that might reflect unusual weather or course maintenance.

 

Eventually I would probably be in favor of allowing all of the scores count.  However, during the transition between two systems, I am not.  In general, I think the current CR and Slopes are decent, but there are probably outlier courses where the CR may be off by a stroke or two and the Slope might not be perfect.  I want these courses to get adjusted to something more proper as quickly as possible, and if we flood the plot for that set of tees with a bunch of scores from players with handicaps that are also out of whack because they are often playing this course, it is going to take a long time to fix the problem.  I figured if a player is playing a set of tees often, take their average score and handicap on that set of tees for the year and plot that point only - at least during the transition. 

[natureboy]

3 hours ago, SG11118 said:

Eventually I would probably be in favor of allowing all of the scores count.  However, during the transition between two systems, I am not.  In general, I think the current CR and Slopes are decent, but there are probably outlier courses where the CR may be off by a stroke or two and the Slope might not be perfect.  I want these courses to get adjusted to something more proper as quickly as possible, and if we flood the plot for that set of tees with a bunch of scores from players with handicaps that are also out of whack because they are often playing this course, it is going to take a long time to fix the problem.  I figured if a player is playing a set of tees often, take their average score and handicap on that set of tees for the year and plot that point only - at least during the transition. 

The other way to look at this is to focus first on the travelling population of golfers and ID'ing statistical discrepancies between their expected rounds and that predicted by the CR and slope. That would identify the 'problem' courses that have less 'portable' ratings and would inform tweaks to the existing course rating system and / or regional discrepancies.

I see what you are saying about having a few rounds from visiting players getting drowned out by many home course data points as far as determining a slope line and overweighting a few rounds from the travelling players. The potential issue I see with this is that a few rounds by an individual golfer are statistically quite variable. If you don't have a good data set of visitors for a certain tee it's less valuable information, and the 'correction' may be more off target than the original. Maybe a minimum of 10 scores per player for comparisons would be better as a threshold, or something like at least 50 or 100 'visiting' golfers scores per tee?

Personally, I'd rather work on tweaking the existing system using actual scores data as it would be less disruptive and more 'evolutionary' while keeping in place a mostly very good system for rating a 'brand new' course without anyone playing it.

Edited July 22, 2016 by natureboy

[GolfLug]

Would it make sense to skew the HCP to give more weight to the most recent round and less weight progressively to older rounds?

 

[newtogolf]

44 minutes ago, GolfLug said:

Would it make sense to skew the HCP to give more weight to the most recent round and less weight progressively to older rounds?

 

I don't think so, it would make sandbagging easier.  Handicap is supposed to represent your potential to score not just reflect your current scoring trends.  

[iacas]

8 hours ago, natureboy said:

I don't think so. The accuracy table for male bogey golfers indicates the width of the area into which a bogey golfer is expected to hit their 200 yard tee shot 2/3 of the time as 33 yards wide. It doesn't say half-width of the area.

No. Now I can say definitively that you are way off. We consider a significantly larger area. Pretty sure it was 120 yards and is now 100.

Read down farther in the 2012-2015 manual you've found online and you'll see that it talks about that table only in what makes something a "significant factor." Essentially, to simplify, you're taking the tree rating and the measurement and multiplying them. The trees within 50 yards (to both sides, so 100 yards wide, in 2016-17) are considered and rated. We simply rate how thick they are, how difficult they are to escape, their number, etc. So if you get a 2 for tree value (very low) and they're far away from the center of the line of play (but inside that 100 yard width), then you might get "2 x 1" basically and call those trees a "2" for that hole. But the same "2" trees very close to the fairway might be a "5" for proximity - they're a significant factor, hence the 5 - and they might rate out at a "10" total (5 x 2).

Super thick trees 50 yards away from the center of the fairway might be an 8, but since they're far away, it's 8 x 1 = 8.

Again that's not exactly how it works (multiplying two integers), but you've misread that chart or misunderstood what it is saying several times, and I'm pointing it out to you again. That chart only shows the border of what counts as a "significant factor." It doesn't say "trees outside of 33 yards are not considered." And I've said that in the past, too.

[Rainmaker]

On April 21, 2016 at 0:14 PM, Rainmaker said:

I haven't really "embraced" the handicap system but, from my outsider's perspective . .it's too complicated.  It's too difficult to establish and maintain (have to belong to a club or some kind of hc service).  People can easily cheat it with selective posting or by not knowing/following the rules, etc.   I think there is an unspoken understanding that there is cheating going on at nearly all handicap tournaments . .even if the prize is just a sleeve of balls, lol. 

Plus I can't rationalize the reason for it, in the first place.  Let's create a way so bad golfers can beat good golfers.  Hmm.  That's odd to me.  The rare times I play for money it's never very much money and I always offer to play "straight up" even if I'm totally going to lose.  If the other guy happens to know my game and wants to give me some strokes . . ok.

I'd rather just play off scratch and lose all the time. 

 

 I finally decided to establish a handicap. I haven't yet but one of the first things I noticed is the course ratings favor distance too much. That is - short courses have excessively low ratings - they aren't that much easier!  Now I have to play higher rated courses unless I want to sandbag.  I'm not planning on competing so I want the lowest index I can legitimately get - which means playing longer, higher rated courses. Not sure if that's truly a handicap issue or a course rating issue....

[Aflighter]

Unless you play from the tips then sometimes the hcp for that hole wont make sense.Id imagine its based off playing from the tips. It might be hardest hole from tips but the tees on up its not.

[natureboy]

On 7/22/2016 at 8:34 PM, iacas said:

No. Now I can say definitively that you are way off. We consider a significantly larger area. Pretty sure it was 120 yards and is now 100.

Read down farther in the 2012-2015 manual you've found online and you'll see that it talks about that table only in what makes something a "significant factor." Essentially, to simplify, you're taking the tree rating and the measurement and multiplying them. The trees within 50 yards (to both sides, so 100 yards wide, in 2016-17) are considered and rated. We simply rate how thick they are, how difficult they are to escape, their number, etc. So if you get a 2 for tree value (very low) and they're far away from the center of the line of play (but inside that 100 yard width), then you might get "2 x 1" basically and call those trees a "2" for that hole. But the same "2" trees very close to the fairway might be a "5" for proximity - they're a significant factor, hence the 5 - and they might rate out at a "10" total (5 x 2).

Super thick trees 50 yards away from the center of the fairway might be an 8, but since they're far away, it's 8 x 1 = 8.

Again that's not exactly how it works (multiplying two integers), but you've misread that chart or misunderstood what it is saying several times, and I'm pointing it out to you again. That chart only shows the border of what counts as a "significant factor." It doesn't say "trees outside of 33 yards are not considered." And I've said that in the past, too.

I can definitively say you are inventing a straw man argument, or are not reading what I have written. I have never said what's bolded in your quote above.

In referencing the accuracy chart, I am not talking about or comparing it to the zone within which the rating system considers trees. I am saying that the 'model' of the bogey golfer embodied in the accuracy table significantly differs from Broadie's model of a ~ bogey golfer. That's where the 16 yard difference lies. The trees being under-weighted in general for higher handicap golfers is another separate point.

I expect the accuracy chart is mostly taken into account for hazard factors within the landing zone, but is to me a symbolic implication the current rating system may have an overall unrealistic expectation of high handicap accuracy, whereas its model for scratch golfers is pretty good per Broadie's data.

I expect that this model mismatch bleeds into an under-estimation of how often higher handicappers are likely to stray far from fairway center line and therefore how much the scoring impact of trees are given in the weighting formula. As you say 50 yards away from the center line is considered 'far away' relative to the center line or edge of the expected landing area, and acts as the cutoff at which trees are no longer taken into account or are weighted very low (a '1') if just within the cutoff distance.

Per Broadie's data, the 50 yard from center line cutoff is at the edge of the ~ bogey golfer's 2-sigma / 95% dispersion width. While for a scratch golfer, their 95% dispersion width is ~ 66 yards and the 50 yard from center line cutoff represents their 3-sigma dispersion width.

So essentially, 99% of the scratch golfer's expected shots are within the 50 yard from center line / 100 yard total width cutoff. Most of the scratch golfer's errant shots will land within the range where the tree hazard is weighted more than the ones at the edge of the expected bogey golfer dispersion. Per Broadie a ~ bogey golfer's 3-sigma distance is ~74 yards from center line (147 yards total). About 5% of the expected average bogey golfer's tee shots are entirely outside the cutoff.

But from a statistical standpoint (per Broadie's data), a scratch golfer's drive to 33 yards from center line is equally bad to a bogey golfer's 49 yard miss of the center line. Why not have a tighter cutoff for consideration of the hazards and the gradation of hazard rating within that distance for the scratch rating? Aren't scratch golfers expected to be more accurate?

Under the rating system, even if the trees were quite close to fairway center line and resulted in an 8x10=80 score, the overall weight given to them is not high relative to the course length. You'd be multiplying the 80 score by something like .14, which itself is then is a sub-contribution of several other factors summing to maybe 15% of the overall rating, while distance alone is weighted in the vicinity of 85% of the overall rating.

IMO these elements of the current rating system put a slight burden of higher / unrealistic performance expectations on the 'model' bogey golfer and by extension through the slope system, higher handicappers in general.

Edited July 24, 2016 by natureboy

[iacas]

Just now, natureboy said:

I can definitively say you are inventing a straw man argument, or are not reading what I have written.

I've read it all, and you continue to be wrong. It's really that simple. You don't understand the purpose or role of the accuracy chart you like to cite, and you do not understand how trees are rated in the course rating system. ALL trees are considered within a VERY wide range, and the handicapping system is about potential, not average ability. You continue to get it backward, too: rate the trees as more important and the 15.7 golfer becomes a 13.2 golfer… and gets fewer strokes from a scratch golfer.

But it's interesting, as this is the first time "you are inventing a straw man" was used as a straw man…

I see no point in hashing this over for the umpteenth time with you. You're set in your incomplete understanding of this, and we're both repeating ourselves.

Have a great evening. Maybe go play some golf, eh?

[natureboy]

On July 22, 2016 at 8:50 PM, Rainmaker said:

Not sure if that's truly a handicap issue or a course rating issue....

They are inter-related.

11 hours ago, Aflighter said:

Unless you play from the tips then sometimes the hcp for that hole wont make sense.Id imagine its based off playing from the tips. It might be hardest hole from tips but the tees on up its not.

Doesn't each tee have it's own rating?

28 minutes ago, iacas said:

I've read it all, and you continue to be wrong. It's really that simple.

According to Broadie, what is the expected dispersion width for the average 90-shooter? How does that compare to the accuracy chart for a bogey golfer in the course rating guideline? The USGA bogey golfer model does not reflect reality as accurately as its model for the scratch golfer. I notice that you keep avoiding addressing that specifically, but AFAIC, it's really that simple.

28 minutes ago, iacas said:

You continue to get it backward, too: rate the trees as more important and the 15.7 golfer becomes a 13.2 golfer… and gets fewer strokes from a scratch golfer.

I'm aware how it would affect HCPs. I'm for accuracy & portability.

IMO a statistical excess of players playing better than their course HCP on non-home courses tends to enhance a mis-perception of 'all those sandbaggers' out there. Perhaps with increased accuracy in course comparisons there would be less 'need' for the arbitrary  'bonus for excellence'?

[iacas]

33 minutes ago, natureboy said:

According to Broadie, what is the expected dispersion width for the average 90-shooter? How does that compare to the accuracy chart for a bogey golfer in the course rating guideline? The USGA bogey golfer model does not reflect reality as accurately as its model for the scratch golfer. I notice that you keep avoiding addressing that specifically, but AFAIC, it's really that simple.

Once again… it's not a "model" in the way you keep saying.

33 minutes ago, natureboy said:

IMO a statistical excess of players playing better than their course HCP on non-home courses tends to enhance a mis-perception of 'all those sandbaggers' out there. Perhaps with increased accuracy in course comparisons there would be less 'need' for the arbitrary  'bonus for excellence'?

Where are "all those sandbaggers"? Where are all the players beating their course handicaps away? They don't exist. And the bonus for excellence makes sense.

And the fact that you have this the opposite - you want trees considered more, which would increase slope, which would possibly increase course ratings, either by just rating trees more or expanding the zone of what trees comprise a "significant factor" - which would again make a 15.3 into a 13.1. If anything, an argument could be made that trees are over-rated now, because if players are hitting into them as often as you suggest they are, they're going to shoot higher scores, and they aren't a 15.3 but maybe they're a 17.8.

You also seem to have no real understanding of what a "6" value for trees is like, or how that affects a course's rating if all the trees were 1s, or there are trees on just one side of play, or the idea of the corridor of the shot for the scratch versus bogey golfer, despite me hinting at or outright saying some of these things in related posts in the many threads in which you've brought this tired topic up.

But here's the thing, at the root of it… aside from some data on how accurate a class of players are from one guy, and a chart you don't seem to understand from last year's manual… you're just using your hunches and what you think in your mind is common sense. You don't seem to understand or know much about the course rating system, how and where and why trees are rated, etc.

And finally, let's pretend that for one minute the accuracy table you're fond of citing was changed to perfectly reflect Broadie's stats, even though his own collection of stats for average golfers pales in comparison to some others out there… the USGA might just change the math in the "significant factor" math and how quickly the values degrade to "not a factor" to arrive at the same average course ratings and slopes. And where would that leave you? Because at the end of the day, the course rating math is a black box to you, or grey at least. You would find yourself with the same knowledge you have now, but would instead agree with the "accuracy table" even though the grey box was spitting out the same numbers in the end.

So I'm done. Your premise is backward, based on what appears to be very clearly a deep misunderstanding of some things, and I'm tired of discussing this topic across multiple threads.

So that's enough, please.

[Chilli Dipper]

Pardon my ignorance on the topic, but how was a slope rating of 113 determined to be "average?" I'm looking through the course ratings of all the regulation-length courses near me, and I'm hard-pressed to find one with a slope under 120 from the white tees.