Professional Documents
Culture Documents
Accentuated Loss Aversion in PGA Tournaments: Does Company Matter?
Accentuated Loss Aversion in PGA Tournaments: Does Company Matter?
Abstract
Professional golfers are loss averse, expending more effort on par putts than on birdie putts, thereby having greater
success with par putts. However it is unclear if this behavioral bias alters under competition with increase in number of
competitors. Our study used PGA putting data along with shot, hole, round and event level information from 29 PGA
tournaments played in 2013 (106,370 attempted putts birdie, par and bogie) to test this contention. Our model
explained over 50% variance in putting success by adding psychological variables, i.e., effect of competitive company,
that is, effect of partner putting for birdie and number of players remaining. N-effect appears to have a negative effect
on the difficult (psychologically) task of putting for birdie. Overall, the results support our hypothesis that the gain frame
of birdie shows pronounced effect of competition field size (n-effect). The results imply that as tournaments increase in
size/competition, player-groupings, and the variance in players performance due to the grouping characteristics will
become crucial parameters for skill and performance evaluation in sports. It is recommended that inclusion of
psychologically meaningful variables in sport analytics will improve model predictability.
Keywords: Putting success, loss aversion, proximity to standard effect, n-effect, social comparison, ShotLink
Golf is a mental game the players operate under psychological factors such as the recently highlighted loss aversion in
reference to the par [1] and dynamics of risk taking by golfers [2]. This paper explores another crucial factor namely the
competition - effort link suggesting that the effort which a player exerts is a function of the number and nature of
competitors presents [3]. This social comparison is driven by player pairing and field size. As golf gains more popularity,
and the tournaments increase in size, the effect of social comparisons on the efforts exerted and the resultant
performance becomes a critical condition in players performance evaluation.
In line with the expectations of prospect theory [4], professional golfers are loss averse: they do not exert equal effort
across different greens. Instead they expend more effort on their par putts (loss frame) than they do on their birdie putts
(gain frame). Due to greater efforts exerted on par putts than the birdie putts, ceteris paribus, putt accuracy is greater for
par than for birdie putts ([1]). This birdie par differential in effort is significant even after controlling for alternative
explanations like learning, player abilities, green-specific effects and rank position in the field. These empirical findings
demonstrate that behavioral biases are not limited to lab-based studies but they extend the sports (a real market) as well
[5]. For example, professional golfers are highly skilled and experienced agents facing large stakes [6] [7] and high
competition[8]. These factors such as skill, experience, chance and competition are generalizable to all most all the
sports. However, in spite of the utility of including psychological/behavioral phenomenon such as loss aversion in
performance evaluation, these constructs remain rarely utilized especially in the game of golf.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Assistant Professor, Strategic Management Area, Indian Institute of Management (IIM) Kozhikode, KL, India
Assistant Professor, Department of Humanities and Social Sciences, Indian Institute of Technology (IIT), Delhi, DL,
2 Assistant Professor, Department of Humanities and Social Sciences, Indian Institute of Technology (IIT), Delhi, DL,
India
1
2
1!
!
!
In reference to the construct of social comparison, the situational factors such as proximity to a standard and the
number of competitors impact social comparison and the resultant competitive behavior [3]. For instance, applying the
social comparison perspective to golf, it can be expected that the playing partner provides a physically proximate
performance standard. Performance in the previous rounds is an important determinant of playing pairings; the playing
partner is quite likely a close competitor. For instance, if the playing partner is shooting for birdie, it is expected that the
player will exert additional effort in order to better or match his playing partner and perform up to the proximate
standard (proximity to a standard effect)[9], and we expect putting performance to improve.
H1: Putting performance improves when playing partner is putting for birdie
The second psychological phenomenon we examine is the n-effect [10]. Social facilitation literature focuses on the
difference between tasks done alone or in a small group, and finds that when individual tasks are performed in the
presence of others, motivation to exert greater efforts increase, thereby increasing the efforts [11]. However, in a
competitive environment, a further increase in N reduces social comparisons and thereby agents motivation to
compete and perform [10]. At times, in experimental studies it has been found that with competitive interactions, the neffect is not robust [12]. One of the reasons could be that social facilitation improves performance on well learned tasks
while having the opposite effect in difficult tasks [3]. Therefore, assuming that the gain frame putting of birdie poses a
difficulty [1], we hypothesize N-effect to further impede birdie putting performance.
H2: Larger field decreases birdie putting performance
Golf tournament mechanisms such as player pairing and its effects on individual players with reference to proximity to a
standard and the n-effect might be intricately linked to field size (i.e., number of players in the tournament). PGA
Tournaments pair players in rounds 3 and 4 of a tournament, primarily on the basis of their performance in previous
rounds. While tournament characteristics, star pairing, career performance and marketing elements determine player
pairings, the previous round performance is a key element. The mere presence of a playing partner introduces relative
performance dynamics and through the above discussed psychological phenomenon could affect player putting
effectiveness.
Method
Data: We used Shot, Hole, Round and Event level information of 29 of the PGA Tour tournaments played in 2013 (see
Table 1) from the link: https://stats.pgatouhq.com/. In these tournaments, players are cut on the basis of the
cumulative scores of players in the first two rounds. Typically, half to two thirds of the starting field survive the cut and
continue playing in rounds 3 and 4. In rounds 3 and 4, the remaining players are grouped together, again on the basis of
their performance on the first two days of the tournament. When about 70 to 78 golfers remain, rounds 3 and 4 would
feature playing pairs the case in the 29 tournaments which were used in this analysis. Playing partners in a particular
round of golf have the same tee-off time (Round level information) and hole-sequence (Round level information was
available in the hole level information of ShotLink). The Hole-sequence determines the hole on the golf course where
that particular playing pair would start its round. We used this information to create playing pairs and create the dummy
variable for whether the playing partner is putting for birdie or not putting for birdie.
[INSERT TABLE 1 ABOUT HERE]
As cited in the literature, we employed controls to account for alternative explanations for the variance in putting
performance: learning, player abilities, green-specific effects and position in the field ([1]). We measured the learning on
the attempted putt using the count of prior putts attempted on the reference green by the player or playing partner ([1]).
The competitive field position of the player is determined by comparing the players score if the putt is successful, with
that of the score of all the players who had completed the same number of holes. This position is used to control for the
motivation hypothesis, that is, if a player is performing well, s/he is motivated to sustain that performance [8]. The other
variables in our analysis were: distance of the attempted putt, whether the putt is for Birdie, Par or Bogey, number of
!
players in the field (to test N-effect hypothesis), player ability and hole specific effects obtaining them directly from
stroke level data of ShotLink3 ([1]).
[INSERT TABLES 2, 3, 4 ABOUT HERE]
Model:
Pr (make putt) =
F (Birdie/ Par, Distance, Previous putts on Green, Position in field, Tournament-Round-Hole characteristics, Player
Ability, Partner putting for Birdie [H1], n-effect [H2]) + e
We used logistic regression to estimate putting success of playing pairs, in rounds 3 and 4 of 29 PGA Tour tournaments
in 2013, a total of 106,370 attempted putts birdie, par and bogie attempts. As mentioned in the earlier section, the
variables used in our analysis were: distance, putt type (birdie, par, bogey), whether playing partner is putting for birdie,
and n-effect (median-based cut off of number of players in round 3 and 4). Control variables were: player ability,
learning, player position in the tournament, and difference across holes. The dependent measure was putting success.
Results: Results of the logistic regression supports hypothesis 1. The proximate standard of playing partner putting for
birdie improved birdie and par putting performance. Playing partner putting for birdie acts as a proximate standard and
improves the overall putting performance of the player. The number of competitors in the tournaments also impacts the
difficult task of putting for birdie negatively (hypothesis 2). While there is a negative impact on the task that is perceived
to be relatively easier (psychologically), that is, putting for par, the effect is marginal in terms of statistical significance for
a large number of observations that were analyzed. The results of logistic regressions are presented in Table - 5 where we
begin with the test of the hypotheses and then add the control variables. It is observed that the results are fairly robust to
the addition of control variables. This indicates that the alternative explanations of player ability, learning, competitive
field position and differences across holes do not entirely account for the difference in putting accuracy which is in line
with the findings of [1] and more importantly the psychological phenomenon of proximity to a standard and the n-effect
have an effect along with loss aversion.
[INSERT TABLE 5 ABOUT HERE]
Out of the 310 PGA professionals who constituted our sample, 142 have at least 30 attempted birdie putts with partner
also putting for birdie (that is, proximity to a standard), as well as at least 30 attempted Birdies without the partner
putting for birdie (that is, without proximity to a standard). It is to be noted that 86% of these 142 PGA professionals
have better conversion rates with their birdie putts when their partner is also putting for birdie. We carried out nonparametric tests to compare the proportions of successful putts in two groups first the group where golfers playing
partner was putting for birdie on that hole and the other group where the playing partner was not putting for birdie. We
carried out a simple z-test of proportions between putts attempted with and without proximity to a standard (with or
without competitive company) controlling for the distance of the attempted putt as well as Birdie versus Par. We find
that for Birdies attempted from different distances, success rates are indeed better when the partner is also going for
birdie (Table 6a). Similar to results of our previous analysis, we find a significantly better success rates with par attempts
as well (see Table 6b).
[INSERT Tables 6, 7 about here]
In the logistic model (see Table 5), we had already controlled for learning effects. Another approach we took was to
isolate only those situations when the golfer merely has the knowledge that his partner would also putt for birdie, and no
opportunity to observe his partner putt. We analyzed putts attempted by players who were in the top deciles of
tournament standings (control for competitive motivation) at the time of the attempted putt. Amongst these, we focus
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
https://stats.pgatouhq.com/ We thank ShotLink for access to the ShotLink database as well as the kind support
provided to us on various queries.
3
3!
!
on putts that were the first ones attempted in the playing group on that particular green (control for learning). We find
that proximity to standard (partner also putting for Birdie) has a significant positive impact on success rate of putts in
the difficult task of putting for birdie (Table 6c) whereas performance on the relatively easier task of putting for Par is
not affected significantly. In our logistic regressions, we found support for the decline in performance due to n-effect for
the relatively (psychologically) difficult task of putting for birdie. However, distance seems to play an important role in
how n-effect impacts putting success. For instance, in the non-parametric tests, decline in putting performance is
statistically significant only in the short putt (from distances less than or equal to two hand a half feet), for both birdie
and par (Tables 7a, 7b). Further, this decline in putting performance due to the n-effect is found to be not statistically
significant in the non-parametric tests of first putts on the green (Table 7c).
Taking the results of our parametric (i.e., stringent assumptions) and non-parametric (realistic assumptions) hypothesis
tests, it appears that loss aversion (birdie-par) could indeed be accentuated by competitive company (via the proximity to
standard effect). We find some support for the hypothesis that the n-effect would have a negative impact on the
relatively difficult task of putting for birdie [3]. Though this result is not robust in our non-parametric tests, similar
results have been observed in other studies (e.g., [12]).
Discussion
Using a fraction of the observations of our predecessors, our model explains little over 50% variance in putting success
by adding critical psychological variables, i.e., effect of competitive company (through the proximity to a standard effect)
that is, effect of partner putting for birdie and number of players remaining. N-effect does appear to have a stronger
negative effect on the more difficult (psychologically) task of putting for birdie, however the statistical support is not
robust in the non-parametric tests carried out. Overall, the results support our hypothesis that the gain frame of Birdie
shows pronounced effect of competitive company (proximity to standard effect), and to a lesser extent to competition
field size (n-effect). With an increase in tournament size, player groupings and effects of grouping on players success
will gain importance.
So, although golf as a sport is celebrated for the strong internal locus of competition (player against the course OR
player against himself in the context of the golf course), subtle yet strong psychologically forces with roots in the social
psychology of competition are at play (proximity to standard, n-effect). Stalwarts like Justin Rose (2014 Quicken Loans
National champion), Jason Dufner (2013 PGA Championship winner and 2011 Championship runner-up), Kevin
Stadler (2014 Waste Management Phoenix Open champion) and Tiger Woods belong to that select group of 45 golf
professionals who putt more accurately for birdie when they have competitive company even when these birdie
attempts are identical or more difficult than attempts made without the benefit of the proximity to a standard.
Additionally, these effects are directly linked to the decisions made by tournament directors in terms of field sizes and
an erstwhile un-researched area, playing pair formations. Our work joins a growing list of researchers focused on
behavioral and psychological aspects of sporting competition. It builds the case for a more nuanced approach to
analyzing and evaluating performance, as well as for formation of playing groups, which is an integral part of Golf
tournaments.
Acknowledgement
We thank ShotLink for access to the ShotLink database as well as the kind support provided to us on various queries.
We thank the conference organizers and the sponsors of MIT Sloan Sports Analytics Conference 2015 for providing
generous conference support.
References
[1]Pope DG and Schweitzer M (2011). Is Tiger Woods loss averse? Persistent bias in the face of experience,
competition, and high stakes, American Economic Review, 101, 129 157.
[2] Serkan O, Smith JK. 2014. Risk taking in competition: evidence from match play golf tournaments. Journal of
Corporate Finance (forthcoming).
[3] Garcia SM, Tor A and Schiff TM. 2013. The psychology of competition: a social comparison perspective.
Perspectives on Psychological Science, 8 (6): 634 650.
[4]Kahneman, Daniel and Tversky, Amos. 1979. Prospect theory: an analysis of decision under risk. Econometrica, 47
(2): 263 91.
[5]Levitt SD and List JA. 2008. Homo Economicus Evolves. Science, 319: 909-10.
[6]List JA. 2003. Does market experience eliminate market anomalies? Quarterly Journal of Economics, 118 (1): 41
71.
[7]List JA. 2004. Neoclassical theory versus Prospect theory: evidence from the marketplace. Econometrica, 72 (2):
615 25.
[8]Koszegi B and Rabin M. 2006. A model of reference-dependent preferences. Quarterly Journal o Economics, 121
(4): 1133 65.
[9]Garcia SM, Tor A and Gonzalez RD. 2006. Ranks and rivals: a theory of competition. Personality and Social
Psychology Bulletin, 32, 970-982.
[10]Garcia SM and Tor A. 2009. The N-effect. Psychological Science, 20(7): 871-877.
[11]Zajonc RB, Heingartner A and Herman EM. 1969. Social enhancement and impairment of performance in the
cockroach. Journal of Personality and Social Psychology, 13 (1): 83-92
[12]Vandegrift D and Holaday B. 2012. Tests of the n-effect and proximity to a standard. Journal of Neuroscience,
Psychology and Economics, 5 (3): 182 192
Appendix I
Table 1: Tournaments in Analysis Dataset 2013
Tournament
Number,
Code
Event Name
1 (20)
2 (30)
3 (50)
4 (60)
5 (70)
6 (90)
7 (100)
8 (120)
9 (130)
10 (140)
11 (150)
12 (170)
RBC Heritage
13 (180)
14 (190)
15 (200)
16 (210)
17 (220)
18 (230)
19 (240)
20 (260)
Travelers Championship
21 (270)
AT&T National
22 (280)
23 (290)
24 (300)
25 (320)
26 (340)
27 (360)
Wyndham Championship
28 (370)
The Barclays
29 (380)
Table 2: Data Summary No: of Successful putts by Putt type and Partner putting for Birdie (or not)
Putt for
Partner
putting for
Birdie
BIRDIE
NO
BIRDIE
YES
BIRDIE TOTAL
PAR
NO
PAR
YES
PAR TOTAL
BOGIE
NO
BOGIE
YES
BOGIE TOTAL
GRAND TOTAL
Successful putt
rate (%)
No: of attempted
Putts
23%
29%
28%
71%
91%
83%
89%
93%
90%
60%
9,124
36,580
45,704
20,694
29,356
50,050
7,500
3,116
10,616
106,370
Table 3: Data Summary No: of Successful putts by Putt type and Field Size greater than Median (or not)
Putt for
Field Size GE
Median (76)
BIRDIE
NO
BIRDIE
YES
BIRDIE TOTAL
PAR
NO
PAR
YES
PAR TOTAL
BOGIE
NO
BOGIE
YES
BOGIE TOTAL
GRAND TOTAL
Successful putt
rate (%)
No: of attempted
Putts
28%
28%
28%
82%
83%
83%
90%
91%
90%
60%
21,299
24,405
45,704
23,738
26,312
50,050
5,199
5,417
10,616
106,370
Partner
putting
for
Birdie
BIRDIE
NO
BIRDIE
YES
BIRDIE TOTAL
PAR
NO
PAR
YES
PAR TOTAL
BOGIE
NO
BOGIE
YES
BOGIETOTAL
GRAND TOTAL
Succe
ssful
putt
rate
(%)
23%
29%
28%
71%
91%
83%
89%
93%
90%
60%
No: of
attempted
Putts
9,124
36,580
45,704
20,694
29,356
50,050
7,500
3,116
10,616
106,370
Avg.
Dista
nce to
Hole
(inche
s)
Average
Putt
distance
Distance to
Hole
(inches)
280
245
252
77
35
53
43
35
41
134
8
8
8
5
2
3
2
1
2
5
!
Average
Position
given
successful
putt
35
34
34
35
34
34
35
34
35
34
Avg.
field
size
76
76
76
75
76
76
75
76
76
76
Avg.
putting
sequence
(in the
playing
pair)
1.3
2.0
1.9
2.4
3.5
3.1
3.1
4.6
3.5
2.6
Proporti
on of
observati
ons
under
Field
size GE
76
54%
53%
53%
52%
53%
53%
51%
52%
51%
53%
2. Base Model
(Robust SE)
3. Player ability
(Roust SE)
-0.01372***
-0.8150***
0.5445
0.4328***
-0.0141***
-0.7456***
0.5270***
0.3985***
4. With all
controls
(Robust SE)
-0.0153***
-0.414 ***
0.410 ***
0.175 ***
-0.4174***
-0.3932***
-0.560 ***
-0.04703*
-0.02050
-0.148 *
0.0632
0.1094
A. Distance
B. Putt (1) Birdie (with respect to Par)
(2) Bogey (with respect to Par)
C. Playing partner putting for Birdie
D. Putt type n-effect
1. Base
Model
(Robust SE)
-0.136***
-1.032***
0.6076***
0.4468***
Constant
2.485***
2.538***
3.400***
3.375 ***
Observations
106,370
106,370
106,367
106,367
24,486
24,266
23,908
22,328
***
***
***
***
46.8%
46.9%
47.7%
50.7 %
Note: *** reflect 0.000 p level. ** reflects 0.05 p level, * reflects 0.10 p level
Distance from
Hole (feet ')
No Competitive
Company
Success
With Competitive
Company
Success
Overall
Success
z-statistic
pvalue
< 2.5'
51%
745
64%
2,039
62%
3,922
6.72
0.00
51%
2,139
55%
5,959
55%
12,906
3.68
0.00
>=12.5'
10%
6,240
11%
2,552
11%
28,876
2.99
0.00
TOTAL
23%
9,124
29%
28%
45,704
10,550
(b) Comparison of Success Rates with and without Company Attempted Par Putts
Distance from
Hole (feet ')
No Competitive
Company
Success
With Competitive
Company
Success
Overall
Success
z-statistic
pvalue
< 2.5'
97%
7,248
99%
17,933
99%
25,280
16.90
69%
10,528
81%
8,550
75%
21,074
19.52
>=12.5'
16%
2,918
19%
151
16%
3,696
2.58
0.00
TOTAL
71%
20,694
91%
26,634
83%
50,050
(c) Comparison of Success Rates of First Putts on Green Proximity to Standard Effect
Partner not for Birdie
Success
Overall
Success
Statistic
p-value
Birdie
15%
606
21%
1,385
19%
1,991
2.97
0.00
Par
53%
272
53%
30
53%
302
0.08
0.47
1,415
23%
2,293
878
N < Median
Success
N >= Median
Success
Overall
Success
z-statistic
pvalue
100%
890
50%
1,530
62%
3,922
26.60
55%
6,257
54%
6,649
55%
12,906
0.43
0.33
>=12.5', <22.5'
11%
14,152
11%
14,724
11%
28,876
0.05
0.48
TOTAL
28%
21,299
28%
24,405
28%
45,704
(b) Comparison of Success Rates with and without n-effect Attempted Par Putts
Distance from
Hole (feet ')
< 2.5'
N < Median
Success
N >= Median
Success
Overall
Success
z-statistic
pvalue
100%
11,463
98%
13,507
99%
25,280
14.18
75%
10,393
76%
10,681
75%
21,074
(1.11)
0.87
>=12.5', <22.5'
17%
1,882
16%
1,814
16%
3,696
0.26
0.40
TOTAL
82%
23,738
83%
83%
50,050
26,312
Overall
Success
zStatistic
p-value
Birdie
18%
912
19%
1,079
19%
1,991
0.61
0.27
Par
55%
145
50%
157
53%
302
(0.84)
0.80
23%
2,293
1,057
1,236