Golf Analytics

How Golfers Win

Golf Analytics Primer

This is a continually updated overview of the major findings in golf analytics. Much of the work here is linked to my own research, supplemented by more general work which is cited and linked to outside sources.

Basic Facts of Golf:
– Strokes are the currency of golf, just like possessions in basketball, downs in football, and outs in baseball. All golf analytics must be done with the idea that the point of the game is the finish with fewer strokes than the field. Shots and strategies that achieve that are positive and those that detract from that are negative. Mark Broadie introduced the strokes gained model and I have been using the Z-Score model (which measures strokes better than the field).

– Performance versus the field is all that matters. Shooting a 66 at the Humana Challenge is much worse than a 69 at a typical US Open set-up. The same can be said about other stats. Hitting 70% of your greens at Kapalua is terrible; hitting 60% of your greens at Harbour Town is above-average.

– Every shot should be evaluated relative to the expected outcome before it is played. An approach shot from 200 yards in the fairway is the same difficulty as one from 150 yards in the rough.

Repeatability of Performance:
– Performances over samples ranging from multiple seasons to single seasons to the first few months of a season regress to the mean. A lot of different tools, such as Bayesian weighting to simple regression to Tour average, can approximate true talent, but golf performance even over a full-season of rounds is roughly 30% randomness. When predicting performance over a large sample, large samples of prior performance are 3-4 times more important than the most recent few months.

– Putting and short game performance are much more affected by randomness in samples ranging from several tournaments to half-seasons to full-seasons, largely because golfers play fewer meaningful putts/short shots than drivers/long shots.

– Younger golfers are more likely to retain their short-term performances than older golfers (ie, there’s more signal in the small sample performances of younger golfers).

– Small sample improvements driven by putting regress to the mean much more than small sample improvements driven by tee to green play (ie, there’s more signal contained in small sample deviations from normal driven by tee to green play than putting).

– Golfers retain around 20% of short-term hot streaks over the next months.

Components of Performance:

– Driving Distance is generally more important than Driving Accuracy in evaluating performance off the tee.

Tournament winners typically play ~12 strokes better than their normal performance, split roughly 40/60 between putting and tee to green performance.

– Performance in the first round has almost no predictive value to projecting second round performance; any value is provided solely by tee to green (tee shots and approach shots) performance. First round putting has zero predictive value when projecting second round performance.

– Long putting performance is more random than short putting performance. Players who over-perform because of making long putts regress more to the mean than those who over-perform because of making short putts.

Aging Curves for Pro Golfers:
– Golfers generally improve slightly between 20 and 30, plateau for much of their 30s, and decline sharply from their late 30s. A top ten golfer on Tour at their peak will normally decline to no better than average by their late 40s.

– Golfers mainly decline as a result of their long game (approach shots and drives). The long game accounts for at least 80% of the typical decline from the late 30s onward.

– Putting is not a major source of improvement or decline, while golfers typically improve their short game with age.

Miscellaneous:
– The very best college golfers generally play above-average on the PGA Tour to start, however very few golfers beyond the top 10-15 ever become good enough to regularly play on the PGA Tour.

– Evaluating golf course difficulty on scorecard length ignores that courses vary in length mainly due to the distribution of holes by par. Many par 72 courses are longer than par 70 courses but actually play shorter. Courses that are long due to having long par 3s and 4s are harder than those that are long due to having more par 5s.

Advertisements

10 responses to “Golf Analytics Primer

  1. robertmorrism November 16, 2014 at 12:02 PM

    That explains Gary Player, and Tom Watson, less so Ray Floyd. Interesting – and how do you correlate the Senior tour stats, with say, the PGA Tour, or European tours?

  2. jalnichols November 16, 2014 at 2:12 PM

    Senior Tour stats are still stuck in the pre-ShotLink dark ages AND there’s no significant pool of players playing both Tours regularly (like exists between Web.com/PGA or European/PGA). So it’s pretty difficult to put the actual stats into any sort of context. Because so few Senior Tour players actively participate on the other major Tours, it’s not a big focus of mine.

  3. robertmorrism November 18, 2014 at 11:19 AM

    No, I understand. Can I ask – do you have stats on “better wind players”, or “tight players” than others. I noticed you had Kucher playing more agressively off the tee block on tight courses, though hitting it shorter than most, and I was wondering how he hits and shapes the ball, low or high, draw or fade. I remember Nicklaus saying, fade into the wind on tight courses and hold the ball up in the wind, using the entire fairway to control the flight instead of aiming it down the middle and using half the fairway.

    Just thinking aloud.

  4. jalnichols November 18, 2014 at 2:41 PM

    That’d be great to have, but historical wind data is difficult to find.

    And it really depends on how you define “tight”. Some courses have tight fairways, but you can land a drive anywhere in an ~80 yard range and be in play. Others have normal sized fairways, but the trees are tight or there’s water in play a lot. My guess would be that Kuchar is better on the latter because he’s one of the best on Tour at staying out of trouble off the tee. The guys who hit it more straight can afford to hit driver on the tight courses and still stay out of trouble.

    Kuchar generally draws it.

    • Jake Saladin August 23, 2015 at 3:03 PM

      Kuchar now is trying to miss more of his tee shots to the right as his approach shots from the left rough are not near as good as his from the right (using the book “Every Shot Counts” he determined this). He usually plays a fade on approach shots.

    • Myles August 8, 2016 at 10:11 AM

      You could use another stat to grade the “tightness” of a course. Something like average distance to the hole after the regulation shot (2cd for par-4, 3rd for par-5). Sure, other factors could affect this stat, but it’s probably a pretty good indicator. Perhaps blend in a bit of fairways-hit to filter out some other factors (like thickness of rough).

  5. Pingback: What’s Wrong With Justin Rose? | Golf Analytics

  6. AL (@alenin) April 27, 2015 at 3:33 PM

    Is anyone out there going to develop a program to allow amateurs like me to upload golf course shot data so that I can have my own strokes gained/lost analysis?

    • jalnichols April 27, 2015 at 4:21 PM

      I think there are a number in development or already released. I don’t have any experience using any, but a Google search might yield some examples.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: