Monday, September 15, 2014

The Four-Ball Stroke Play Adjustment Under Section 3-5

The USGA recommends different procedures for adjusting handicaps under Section 3-5 depending upon the format of the competition.  In foursomes and Chapman competitions, the team playing the course with the higher Course Rating adds d-strokes (d being the difference in Course Ratings) to its team handicap.  In four-ball competitions, however, each player has d-strokes added to his handicap.
In foursome and Chapman formats, the effect of the adjustment is certain and known in advance.  The teams playing from the tees with the higher Course Rating will have their net scores reduced by  d‑strokes.  The effect in four-ball stroke play, however, is not certain.  If the two players have the same handicap, the effect will be to reduce the team score by d-strokes, the same as in foursome and Chapman competitions.  If the difference in handicaps between partners is equal to or greater than d, the expected reduction in team score will be d-strokes but can range from zero to 2d-strokes.
The probability that a team will gain more than d-strokes increases with the value of d.  As shown in the Appendix below, if d is equal to 3, a team has a 34 percent chance of reducing its team score by more than 3-strokes.  In a large field competition, the overall winner will most likely come from the teams playing the tees with the higher course rating.[1]   To eliminate this inequity and to make the handicap adjustment under Sec. 3-5 consistent over all forms of competition, it is suggested the adjustment for four-ball should also be a reduction in team score by d-strokes.


Appendix
Expected Reduction in Net Score

To simplify the model, it is assumed that only five scores on a hole are possible—eagle. birdie, par, bogey, and double bogey.  The probability of making each score for the two players is shown in Table A1 below.
Table A1
Probability of Making Various Scores

Score
Player 1
Player 2
Eagle
a1
b1
Birdie
a2
b2
Par
a3
b3
Bogey
a4
b4
Double Bogey
a5
b5

Now on any hole there are 25 possible outcomes as shown in Table A2.  Assuming the scores by each player are independent, the probability of each outcome is shown column 3.  Assume that d is equal to 1 so that each player gets an additional stroke, and the handicap of Player 1 is at least one stroke lower than the handicap of Player 2.  Column 4 shows the results for when Player 1 gets an additional stroke.  As an example, if both players have eagled the hole, the additional stroke does not result in a reduction in the total score (i.e., player 2 by definition already has a stroke on that hole).  That is why zero is shown in Column 4 for the eagle-eagle outcome.
Similarly, Column 5 shows the reductions when Player 2 gets an additional stroke, but Player 1 does not.  The eagle-eagle outcome results in a reduction of -1 since Player 2 now strokes on the hole.
Table 2A
Reduction in Net Score for Possible Outcomes

(1)

Player 1
(2)

Player 2
(3)

Probability
(4)
Player 1 Gets +1 Stroke
(5)
Player 2  Gets +1 Stroke
Eagle
Eagle
a1·b1
0
-1
Eagle
Birdie
a1·b2
-1
0
Eagle
Par
a1·b3
-1
0
Eagle
Bogey
a1·b4
-1
0
Eagle
Double Bogey
a1·b5
-1
0
Birdie
Eagle
a2·b1
0
-1
Birdie
Birdie
a2·b2
0
-1
Birdie
Par
a2·b3
-1
0
Birdie
Bogey
a2·b4
-1
0
Birdie
Double Bogey
a2·b5
-1
0
Par
Eagle
a3·b1
0
-1
Par
Birdie
a3·b2
0
-1
Par
Par
a3·b3
0
-1
Par
Bogey
a3·b4
-1
0
Par
Double Bogey
a3·b5
-1
0
Bogey
Eagle
a4·b1
0
-1
Bogey
Birdie
a4·b2
0
-1
Bogey
Par
a4·b3
0
-1
Bogey
Bogey
a4·b4
0
-1
Bogey
Double Bogey
a4·b5
-1
0
Double Bogey
Eagle
a5·b1
0
-1
Double Bogey
Birdie
a5·b2
0
-1
Double Bogey
Par
a5·b3
0
-1
Double Bogey
Bogey
a5·b4
0
-1
Double Bogey
Double Bogey
a5·b5
0
-1


The probability that Player 1 successfully uses an additional stroke on a hole to lower the team score is:
1)            p = a1·b2+a1·b3+a1·b4+a1·b5+a2·b3+a2·b4+a2·b5+a3·b4+a3·b5+a4·b5
The probability that Player 2 successfully uses an additional stroke on a hole to lower the team score is:
2)            q = a1·b1+a2·b1+a2·b2+a3·b1+a3·b2+a3·b3+a4·b1+a4·b2+a4·b3+a4·b4 +a5
Assume the difference in Course Ratings is “d” strokes.  The probability that Player 1 lowers the team score on “n” holes is:
3)            P1(n) = (d!/(n!(d-n)!) )· pn·(1-p)d-n
Similarly, the probability that Player 2 lowers the team score on “n” holes is:
4)            P2(n)  = (d!/(n!(d-n)!))·qn·(1-q)d-n
To evaluate these probabilities, it is necessary to know the likelihood of making each score for both players.  In previous posts, reasonable estimates of these likelihoods have been used and are presented in Table 3A below.
Table 3A
Probabilities of Scoring

Score
5-Handicap
10-Handicap
Eagle
.005
.003
Birdie
.140
.090
Par
.450
.350
Bogey
.310
.380
Double Bogey
.095
.177

Based on the assumptions in Table 3A, the estimates of p and q are shown in Table 4A.
Table 4A
Probability of a Reduction in Team Score for One Additional Handicap Stroke

 Player
Probability
Estimated Probability
Player 1
P
.44
Player 2
q
.56

Two cases are examined to demonstrate how the competition could be affected by using Sec. 3-5.  In the first case “d” is one stroke.  The probability of each player lowering the team score is shown in Table 5A below:


Table 5A
Probability of Team Score Outcomes When d =1

Outcome
Formula
Estimated Probability
0
(1-p)·(1-q)
.25
-1
p·(1-q) + q·(1-p)
.50
-2
p·q
.25

The expected reduction in team score is -1 stroke.  There is a 25 percent chance, however, a team will lower its score by -2 strokes.
Table 6A shows the probability of various outcomes when the difference in course ratings is 3 strokes and the difference in handicaps is 3 or greater.
Table 6A
Probability of Team Score Outcomes When d=3

Outcome
Formula
Estimated Probability
0
P1(0)·P2(0)
.015
-1
P1(1)·P2(0) + P1(0)·P2(1)
.092
-2
P1(2)·(P2(0) +P1(1)·P2(1) + P1(0)·P2(2)
.235
-3
P1(3)·P2(0) + P1(2)·P2(1) + P1(1)·P2(2) +P1(0)·P2(3)
.315
-4
P1(3)·P2(1) +P1(2)·P2(2) + P1(1)·P2(3)
.235
-5
P1(3)·P2(2) + P1(2)·P2(3)
.092
-6
P1(3)·P2(3)
.015


Table 6A indicates the probability of reducing a team score by the full 2d strokes declines as d increases.  The probability of reducing a score by more than d-strokes, however, increases with d.  In this case, a team has a 34 percent chance of reducing its net score by more than 3-strokes.




[1] The winner will most likely come from the teams playing the tees with a lower course rating if those teams have their handicaps adjusted downward by d-strokes. 

Tuesday, August 26, 2014

The Birth of "113"

The USGA has long claimed that the slope of the line relating average score to handicap is “1.13.”[1] The USGA, however, has never presented any empirical evidence supporting its claim.  I suspect that is because no such evidence exists.  At the time of the claim, the USGA had neither the computing power nor the data to make an estimate of the slope.  It is possible that a small and unpublished USGA study found the slope to be around “1.13.”  Such a limited study, however, could not claim that “1.13” was a finding that applied to every course.  It is more probable that the universal “1.13” came from a theoretical calculation.  To make that calculation, two assumptions had to be made: 1) The standard deviation of a player’s score is a linearly increasing function of handicap, and 2) The standard deviation is, on average,  the same for all players of the same handicap regardless of course difficulty.

The necessity of these assumptions can be found through an examination of the equations of the USGA Handicap System.[2]  Assume the standard deviation of a scratch player is σ.  The standard deviation of a handicap player is postulated to be (1 + a·Handicap)·Ïƒ.  In a normal distribution, the average of the better half of differentials will be .8·Ïƒ below the average of all differentials.  Therefore, the equation for a player’s average score can be written as:

      1)  Average Score = Course Rating + Average of Ten Best Differentials +.8·(1 + a·Handicap)·Ïƒ

Since a player’s handicap is equal to .96 times the average of his ten best differentials, equation 1 can be rewritten as:

      2)  Average Score = Course Rating + Handicap/.96 +.8·(1 + a·Handicap)·Ïƒ

The slope is the average score of a handicap player minus the average score of a scratch player divided by the handicap of the player:

       3)  Slope = (Average Score of Handicap Player – Average Score of Scratch Player)/Handicap
                      = ((Course Rating +Handicap/.96 +.8·(1 + a·Handicap)) –                                                                                                  (Course Rating +.8· σ))/Handicap
                      = (Handicap/.96 +.8·Ïƒ +.8·a·Handicap·Ïƒ  - .8·Ïƒ)/Handicap    
                      = 1.04 +.8· a· σ

The 1.04 represents the part of the slope due to the USGA’s Bonus for Excellence.  Scheid has written that the value of “a” is .05.[3]  To have the slope of the line relating average score versus handicap be 1.13, the standard deviation of the scores of a scratch player must be 2.25. And so “1.13” was born in equation 4: 

          4)  Slope = 1.04 + .8·2.25 ·.05 = 1.13

The slope of “1.13 depends on two assumptions that have never been proven to be valid.  The Slope System, however, does not depend upon “1.13” being the actual slope.  If the real slope was 1.05, then all Slope Ratings would be off by 7.6 percent (1.13/1.05).  Such a proportional error would not affect the efficacy of the Slope System (i.e., handicaps and indices would remain the same).





[1]  Scheid contends that studies of large numbers of golfers show that player means increase with USGA handicap at a slope near 1.13.  Scheid does not cite any of these studies, however.  See Scheid, F. J., “On the normality and independence of golf scores, with various applications,” Science and Golf, E & FN Spon, London, 1990, p. 151.
[2] The USGA Handicap System 2012-2016, United States Golf Association, 2012.
[3] Scheid, loc. cit.

Thursday, July 17, 2014

Why Is “113” in the Slope Handicap System?

Introduction and Summary - The number “113” has a ubiquitous presence in the USGA Handicap System.   It is used as a multiplier in calculating a player’s handicap differential.  It is used as a divisor in determining a player’s handicap.  It is not clear, however, how this seemingly strange number came to be such an integral part of the Slope System.  Three possible reasons are examined here.  First, a case can be made that “113” stems from the theoretical underpinnings of the Slope System.  A slope of 1.13 is assumed by the USGA to be the slope of the line relating a player’s average score with his handicap.  The USGA’s Course Rating Model, however, actually measures the slope of the line relating the average of the better half of a player’s scores and his handicap.  The slope of this line is 1.04.  To determine a course’s Slope Rating, the USGA takes the result of the Course Rating Model and multiplies it by 1.13/1.04.    Therefore, theory may explain where “113” came from, but not why it was chosen over “104”.
Second, the USGA has stated “113” is used since it is the empirically derived average Slope Rating on a standard American golf course.[1]  Since the Slope Rating is essentially a statistical index, there can be no empirically derived average.  Any average Slope Rating would depend upon the USGA’s choice of a reference value.  The USGA’s choice of “113” determined the reference value, not the other way around.  
Third, the USGA selected “113” for bureaucratic reasons.  This is the most likely explanation.  The USGA could have chosen any number for the reference Slope Rating—“100” would have been the ideal choice—without altering the handicaps of any player.[2]  The number “113” was probably chosen because it gives the illusion of precision and adds mystique to the Slope System.  In the end, however, the selection of “113” unnecessarily complicated the Slope System but did not affect estimates of Indices or Handicaps.  Each of the possible reasons behind “113” is examined in turn.
 1. The Number “113” Stems from Theoretical Underpinnings of the Slope System - The theory behind the USGA’s Slope system was first put forward in an article by Stroud and Riccio (hereinafter referred to as Stroud).[3]   Stroud argued the “slope” of a course can be defined either of two ways.   First, the “slope” can be defined as the slope of average score versus a player’s Perfect Valley handicap where Perfect Valley is the reference course.  Second, the slope can be defined as the average of a player’s ten best differentials versus his handicap at the reference course.  The reference course would have a slope of 1.13 if a player’s average score is plotted.  The slope would be 1.04 if the average of a player’s ten best differentials is plotted.
This would seem to explain where “113” came from.  Unfortunately, the USGA’s Course Rating Model does not measure the slope of average scores versus handicap.  The USGA does not define either the Course Rating or the Bogey Rating with mathematical precision.  Course Rating, for example, is defined as the playing difficulty of a course for a scratch golfer based on distance and obstacle factors (i.e., the USGA Course Rating Model).[4]  Knuth, however, writes that Course Rating Model attempts to predict the average of the better half of scores for the scratch and bogey player.[5]  Using Knuth’s definitions of the Ratings, the slope actually measured is:

                1)            Slope = (BDBP – BDSP)/Bogey Handicap
Where,
                                                BDBP = Average of the ten best differentials of a bogey player
                                                BDSP = Average of the ten best differentials of a scratch player
                                                Bogey Handicap = Handicap of the Bogey Player at the Reference Course

By definition the slope of this line is 1.04 (1/.96).  The USGA could have chosen to go with "104" as the reference Slope Rating.  Instead, they chose to multiply both sides of the equation by 1.13/1.04.  This resulted in the USGA’s equation for the Slope Rating (Note: The Bogey Handicap has to be 20.16 for this to work.  This is different than the USGAS’s definition of the Bogey Handicap which is 20.0.  This small discrepancy, however, does not affect the workings of the Slope System):

                2)            USGA Slope Rating = 5.381·(Bogey Rating – Course Rating)

The USGA has never published any evidence that the slope of the line plotting average score versus handicap is 1.13 at all courses.  For this to be true a crucial assumption has to be met: A player’s standard deviation of scores must increase at the same rate at all courses.  That is, on average the standard deviation of scores for any handicap level is not dependent on the Slope Rating.  The USGA adopted “113” even though it is based on this tenuous theory.  By adopting “113” instead of “104”, the USGA also increased the standard error of the estimate for the Slope Rating.  In summary, there is nothing in the theory of the Slope System that would make “113” a clear choice.            

2. The number “113” was Chosen Since It Was the Average Value of a Standard American Course – Knuth has written that “113” was an empirically derived average Slope Rating.  Knuth does not document this finding.  An examination of how the Slope Rating is estimated shows that “113” was not an average.
The USGA used a multivariate regression model to estimate the Course and Bogey Ratings as a function of yardage and obstacle values.  The USGA’s Course Rating Model does not measure the Slope Rating directly, but rather the difference between the Bogey Rating and the Course Rating.  As an example, Table 1 shows the difference in Ratings for five courses.  The only way “113” can be the average Slope Rating is if the originator (i.e., the USGA) assigns the Slope Rating of “113” to a course that has the average difference between its Bogey and Course Rating (Course 3 in this example).  In essence, the Slope Rating of “113” was not empirically derived, but a conscious selection by the USGA.[6]

Table 1
Difference Between Bogey and Course Rating


Course
Difference Between Bogey and Course Rating (Strokes)
1
19.0
2
20.0
3
21.0
4
22.0
5
23.0
 
3) “113” Selected for Bureaucratic Reasons – In the USGA’s Handicap System, any number (except zero) chosen for the Slope Rating of the reference course will yield identical results.    As an example, let’s compare the Stroud Slope Rating based on a reference course Slope Rating of "104" with the USGA Slope Rating of “113.”   Table 1 presents the Slope Ratings under the two methods for courses where the difference between the Bogey Rating and Course Rating is 26, 21, and 16 strokes.

Table 2
USGA and Stroud Slope Ratings

Course
Bogey Rating – Course Rating
USGA Slope Rating
Stroud Slope Rating
1
26
140
129
2
21
113
104
3
16
86
79

 The two methods yield different estimates of the course Slope Rating, but identical estimates of a player’s index and handicap.   To see this, Table 3 presents the index of a player who has ten best differentials equal to DIFF on the Course 1.  His Index is the same under both Slope Ratings.

Table 3
Index of a Player with Ten best Differentials Equal to DIFF


USGA Index
Stroud Index
Index
=.96·Diff·(113/140) = .77· DIFF
=.96·DIFF·(104/129) = .77·DIFF
               
Table 4 presents this player’s handicap at the three courses.

Table 4
Handicap of a Player with Ten Best Differentials Equal to DIFF

Course
USGA Handicap
Stroud Handicap
 1
=.96·DIFF(113/140)·(140/113) = .96·DIFF
=.96·DIFF(104/129)·(129/104) = .96·DIFF
2
=.96·DIFF(113/140)·(113/113) =.77·DIFF
=.96·DIFF(104/129)·(104/104) =.77·DIFF
3
=.96·DIFF(113/140)·(86/113) =.61·DIFF
=.96·DIFF(104/129)·(79/104) =.61·DIFF

Handicaps are determined by the ratio of the difference between the Bogey Rating and the Course Rating at any course.  The Slope Rating chosen for the course of average difficulty has no effect on the efficacy of the Slope System.

So why did the USGA choose “113” as the reference Slope Rating?   When the USGA named the new handicap system the Slope System, it obviously needed a slope.  It is possible “113” was selected since it was easier and more understandable to talk of a player’s average score than the average of his ten best differentials. 

A more sinister reason is USGA did not want a simple explanation of the Slope Rating and chose to make a player’s handicap a function of two rather meaningless numbers—the Slope Rating and “113.”  This added complexity and may have served to dampen criticism of the Slope System.  After all, it is difficult to criticize what you don’t understand.

Another possible reason for the USGA’s decision is that it wanted to give the illusion of precision.   The number “113” gives the impression that it was derived through exact science.  For example, if you hear that the attendance at a golf tournament is 25,000 it sounds like a rough estimate.  If the attendance is announced 25,301 it appears as an accurate estimate even though the method of counting is exactly the same.  The number “113” does not sound like it was drawn out of a hat, even though that is not far from the truth.

If the new handicap system had been named something like Handicap Adjustment for Course Difficulty there would have been no reason to introduce a slope.  Instead of the Slope Rating, there would have been something termed the Adjustment Rating.    A reference Adjustment Rating of “100” would have given intuitive meaning to the rating.  The Adjustment Rating would be the percentage of a player’s index he would receive as his course Handicap.  For example, if the Adjustment Rating was 120, a player’s Course Handicap would be 120 percent of his index.

Conclusion – There appears to be no strong theoretical or empirical basis for choosing “113” as the reference value for the Slope System.  There are possible bureaucratic reasons to explain the USGA’s decision to select “113.”  In the end, however, the decision unnecessarily complicated the Slope System but did not affect estimates of USGA Indices or Handicaps.

  






[1] Knuth, Dean, “A two parameter golf course rating system,” Science and Golf, The Proceedings of the World Scientific Congress of Golf, E & F.N. Spon, London, 1990, p. 143.
[2] The Slope Rating is essentially an index and any number can serve as the reference value.  The USGA Index, however, is not an index.  This is confusing and may explain why some countries have not adopted the term “Index.”  Australia for example, uses the term “Golf Australia Handicap” for what the USGA calls a player’s Index. Canada uses the term “handicap factor” instead of Index.
[3] Stroud, R.C., and L.L Riccio, “Mathematical underpinnings of the slope handicap,” Science and Golf, The Proceedings of the World Scientific Congress of Golf, E.FN. Spon, London, 1990, pp. 135-140.
[4] The USGA Handicap System 2012-2015, United States Golf Association, Far Hills, NJ, p. 11.
[5] Knuth, Dean, “A two parameter golf course rating system,” Science and Golf, The Proceedings of the World Scientific Congress of Golf, E.F. Son, London, 1990, p. 143.
[6] There is little evidence to suggest that “113” is the average Slope Rating in the United States.   A cursory observation of Slope Ratings listed by state golf associations indicates that the average Slope Rating is much higher than 113. Moreover, the reference Slope Rating for women is also 113.  What is the likelihood that the average Slope Rating is the same for both men and women?