• Analytics Blog
August 7th, 2018
In today's game, where players surround themselves with an entourage of consultants and advisors,
swing coaches are daily fixtures in the lives of PGA Tour pros. And while the nature of the
player-coach relationship can vary both in its philosophy and intimacy, it's exceedingly rare to find a
player who does not use a swing instructor in some capacity. Intuitively, we assume that coaches
are beneficial to a golfer's performance (if not, why would they have them?). However, to what extent can a coach
alter the natural ebbs and flows of a golfer's performance? Did the Butch Harmon 'overhaul' of Tiger's golf
swing lead to the further elevation of his play, or would Tiger have followed a similar career trajectory
As with many of the variables thought to influence golfer performance, the impact of a swing coach is
difficult to quantify (and likely impossible for players who have only worked with one coach). Fortunately, the
precarious nature of professional golf lends itself to tinkering, and there are numerous examples of
players changing swing instructors at some point during their career. In this article, we analyze
the impact of coaches in golf by documenting golfers' performance before and after their coaching
change. This analysis is mostly descriptive: it is very difficult to isolate a 'causal effect' because golfers
do not change coaches randomly (i.e. they choose to do it), and also because golf scores can be very noisy.
For our analysis we gathered the dates of 39 coaching changes for PGA Tour players between 2004 and
2017 (accurate to the year and month). Ideally our analysis would have included a larger
sample of players, but finding details on coaching changes proved to be a difficult endeavour. With these dates in hand, we then
calculated the 30-round moving average of each golfer's adjusted strokes gained for the two years before and the two years after the coaching switch.
The moving average (an average of their 30 most recent rounds on a given date) is a less noisy measure of a golfer's recent performance and is much more
visually informative than the raw round-level strokes-gained data. All of the data is presented in the dynamic graph below. Note that
the x-axis is scaled by number of rounds before or after the switch (and not date). Also, not all players in the sample had data available 2
years prior to their coaching change, and a few players did not stay with their new coach for the full 2
years after. As a result, the lengths of the plots differ for each player. Plots are
grouped by swing coach, so be sure to explore all of the data using the dropdown menu!
When toggling through the different player-specific plots, it's clear that the strokes-gained trajectories following a coaching change vary widely across players. Most
show sizeable fluctuations in performance both 'pre' and 'post' instructor change, and it's not immediately obvious
what effect, if any, changing swing coaches has on a player's performance. That being said, it appears the
majority of players experienced a decline in performance prior to a coaching switch. This fits with intuition,
as a performance 'slump' is one of the main reasons cited by players for changing coaches. After
the coaching change is where the plots become less predictable: some players experience an
immediate improvement in performance (Phil Mickelson, Justin Rose), while others continue their decline
(Rickie Fowler, Gary Woodland). As we move further away from the date of the change, most plots fluctuate
considerably and it's our impression that many players oscillate around a level of performance that is
similar to their 'pre-slump' level of play. This fits with the idea that swing coaches are perhaps most
valuable in offering a new perspective that helps a player recapture their previous form, as opposed to
drastically changing or improving their game. We do note however, that our views on the benefits (or lack
thereof) of coaching in golf are likely biased by our own personal experiences.
As it relates to the individual plots, we wanted to highlight two players that we found interesting. The first
is Rickie Fowler, who best exemplifies one of the more common coaching change narratives. Prior to
beginning his work with Butch Harmon, Rickie experienced a decline in performance;
immediately following the switch, this decline continued as Rickie and Butch worked on swing changes. Finally, as Rickie became comfortable with the
changes (or so the narrative goes), his game reached a higher level of performance than at any earlier point in his career. While there are other players that follow this
general pattern (Rory Sabbatini for example), there are many players who do not.
One such example is Jimmy Walker, who also began working with Butch Harmon in December of 2012. His
improved performance from 2013-onwards (culminating with his win at the 2016 PGA championship) is
sometimes attributed to this coaching change. Examining Walker's plot, we see that he was already
experiencing a steady rise in his game prior to beginning his work with Butch. And while his game
continued to improve under the tutelage of his new coach, it's possible that Walker's improved play would
have come to fruition regardless of the coaching switch (it's also entirely possible that it would not have).
In this next figure we summarize the data visualized above by aggregating over coaches. We look at the changes in each golfer's
30-round moving average over 3 time periods: the start of our sample to the time of the switch; from the switch up to 1-
year post-switch; and finally from 1-year post-switch to 2-years post-switch. These player-specific changes are then averaged for each coach.
Here are the results:
The first column of the table is meant to capture the trend in performance leading up to a coaching change. As was fairly evident in
the first figure, the majority of coaching changes occurred after (or while) players were experiencing a decline in
performance. The second and third columns show that, in general, players experienced a small
improvement in their performance during their first year with a new coach, and then more or less maintained their
level of play during the second year post-coaching change. It is interesting to note that, overall, the
decline in performance (-0.37 strokes-gained per round) prior to the coaching change is similar to the increase in performance
(+0.31) in the first year following the change. Given the small number of players included in our analysis
we are hesitant to draw strong conclusions, but this does support the notion that while the swing
coaches in our sample did not dramatically improve the quality of their players long-term, they did help
them return to their 'pre-slump' level of play.
As far as specific coaches are concerned, the data looks promising for John Tillery and Scott Hamilton. Their players tended to experience
improvements in performance in the first year following their coaching changes (+0.98 and +0.69 SG per round, respectively).
This improvement continued during their second year with Tillery's players showing a further improvement in
performance (+0.53), while Hamilton's maintained their level of play (+.09). While we don't think our sample is
comprehensive, we did find 5 and 6 players for Hamilton and Tillery respectively, which we think is a
reasonable representation of their coaching history on the PGA Tour.
So what should we make of this data? As mentioned from the outset, these patterns should not be taken as causal. Golfers switch coaches
for all sorts of reasons apart from just what we perceive be to trends in performance (e.g. strokes-gained averages). Because we can never
know the true counterfactual (i.e. what their performance would have looked like without a coaching change), conclusions can only be
speculative. However, with a decent sample size (39 players), we still think these data are useful for thinking about the interaction between
coaching changes and golfer performance. We saw that most players experienced declines in performance leading up to instructor changes, followed
by (on average) an increase in performance in the first year post-switch which roughly offset the previous decline. This aggregate pattern of course does hide a lot of
player-specific variation. This analysis also highlights the difficulty in attributing the impact of specific events (e.g. coaching changes, personal events, etc.)
to a golfer's performance. Even aggregating data to a 30-round moving average, we still see significant fluctuations in all of the player-specific plots over short
periods of time.
Because we can't resist, we'll end this by stirring the proverbial 'pot' as it relates to coaching and performance. If you are looking for a lesson, John Tillery
or Scott Hamilton (and not Butch Harmon or Sean Foley) could be the two best coaches to consider!