PGA Tour: How Good Is The Strokes Gained Metric?

The PGA Tour issues a plethora of metrics to its members and fans. With so many metrics to choose from the question inevitably arises with regard to which of these metrics best measure a golfer’s performance over the course of a shot, a round, a tournament, a year and/or a career? The generally accepted answer is the Strokes Gained metric.

The purpose of this article is to identify the weaknesses and strengths of these metrics and to put forward improvements. To note, all data throughout this article is from the PGA Tour website unless stated otherwise.

The Strokes Gained Metric: What is it?

Every shot a player makes on the PGA Tour is measured against the average outcome of that particular (or very similar) shot by all players on tour. If a player’s shot is better than the average outcome, they will have a positive Strokes Gained number. If a player’s shot is below average, they will have a negative number. Every shot in a round is assigned a Strokes Gained statistic, and at the end of the round they are all summed together to give the Strokes Gained: Total metric.

First adopted by the PGA Tour in 2011, Mark Broadie, who created the original Strokes Gained: Putting metric, explains: “Strokes gained recognizes that sinking a 20-foot putt represents a better performance than sinking a three-foot putt, even though they both count as a single stroke on the scorecard. Strokes gained assigns a number to this intuition.”

Example: Strokes Gained: Putting

In the above example, both players have an 8-foot putt. On average it takes a Tour player 1.5 putts to sink an 8-foot putt.  Player A took 1 putt and therefore beat the average by 0.5 strokes; this is shown by a positive Strokes Gained statistic (+0.5). Player B took 2 putts from 8 feet and therefore exceeded the average by 0.5 strokes; this is shown by a negative Strokes Gained statistic (-0.5). The difference in Player A and Player B’s Strokes Gained statistic is one, which makes sense as Player B took one more putt. To note, an improvement of 0.5 strokes per round increases a player’s earning potential by 73 percent.

Overview of the Strokes Gained metrics

Following creation of the original Strokes Gained: Putting metric, five others have been devised to measure other player attributes. These are Strokes Gained: Off-the-Tee, Strokes Gained: Approach-the-Green, Strokes Gained: Around-the-Green, Strokes Gained: Tee-to-Green and Strokes Gained: Total. Of these six Strokes Gained metrics, two (Strokes Gained: Tee-to-Green and Strokes Gained: Total ) are summations of a combination of the other four as illustrated in the diagram above.

How accurate are the Strokes Gained metrics?

The level Tiger Woods reached is undeniably higher than any other golfer in the modern game and the Strokes Gained: Total captures this. Judgementally therefore, it seems to be measuring what it is intended to measure pretty well. Whilst there may be a few surprises in the other Strokes Gained metrics, these tend to show the players one would expect to see both near the top of the rankings and near the bottom. For example, Rory McIlroy is continuously high up in the Strokes Gained: Off-the-Tee metric which is of no surprise whatsoever due to his most-outstanding, proficient swing.

How can the PGA Tour improve its Strokes Gained metrics?

Improvement One: Do not rely on distance alone

The problem

First, the inherent difficulty in a shot cannot be solely captured by distance. Now, whilst distance is a key factor in most shots’ proximity to the hole, it is not the only one, and, at times, you could even question whether that is the most important factor in measuring performance.

Is there a difference in difficulty between a downhill big-swinging left-to-right putt than a straight putt of the same distance? Absolutely. Putting difficulty is heavily dependent on how much the ball will turn and how much elevation change the ball experiences. Yet, these are not factored into the calculation.

In the examples above, Putt A is twice as long as Putt B and according to Strokes Gained: Putting, Putt A takes, on average, 0.25 strokes more than Putt B. In other words, Putt A is 15.34% harder to make than Putt B. Yet, this is unlikely to be the case and more likely to be misleading as slope and break are not accounted for. Acknowledging other factors such as wind (particularly when strong), green speed and grass type affect putting difficulty, factors such as slope and break are likely to capture putting difficulty as much, if not more, than distance does.

Another example of other factors significantly contributing to a shot’s difficulty is when a player has missed the green and within 30 yards of the edge of the green. This statistic is measured by Strokes Gained: Around-the-Green. In the examples below, both Shot A and Shot B are 20 yards from the hole with Shot B from the fairway whereas Shot A is from the rough. According to Strokes Gained: Around-the-Green, Shot A takes, on average, 0.19 strokes more than Shot B. In other words, Shot A is 7.92% harder than Shot B.

This metric indicates that Shot B is hit, on average, approximately 9 foot and Shot A is hit, on average, 6.5 foot, a difference of 2.5 foot. Now, despite taking into account the lie, Shot A appears immensely more difficult due to it being short-sided (missing to the side of the green where the flag is positioned). It is near certain that Shot A is more difficult, yet the metric indicates otherwise, thus, again, misleading. These two fairly obvious examples highlight that when distance is short, other factors are key to measuring golf skill/performance. However, this is not the only issue with solely factoring distance.

The second problem in the use of an average is that it hides the variability of hole difficulty. Each hole on the PGA Tour is given an average (expected) score based on its distance and nothing else. So, for a 450-yard par 4, the average score is about 4.1 and is collected from all the scores played on all par 4s 450 yards in length (or very similar). It goes without saying that there would be variability in scoring from one 450-yard hole to another. In other words, not all 450-yard par 4s are of equal difficulty.

One caveat here might be that the reason the hole might be harder/easier is due to a harder/easier tee shot and/or approach shot and/or green to putt and chip on. For example, if the hole average was higher solely because of a challenging approach shot, then the same tee-shot should not be rewarded differently any differently, but the approach shot should be. Though it could be argued that the pressure to hit the fairway is higher due to a more demanding approach shot, more credit should be given to the tee shot despite the actual difficulty being the same as the average par 4 of that length.

Third, by only measuring distance, the variability of each shot is lost. The image below shows two holes that are both 420 yards long and have an average score of about 4 strokes. Now, Hole 1 has water on both sides of the fairway, giving a more challenging tee shot than Hole 2 which has no hazards as well as a wider fairway. On the other hand, Hole 2 has a more challenging approach shot than Hole 1 due to hazards (water) and a more challenging green shape.

With both holes the player hits a 250-yard tee shot on the fairway and then 170-yard approach shot to 35 feet and two putts. These are the averages required to score zero on all Strokes Gained metrics for the hole. Assuming putting difficulty is the same for both holes from 35 feet, the Strokes Gained: Tee-to-Green metric is zero, the same for each hole. However, what is lost (i.e., not captured) is the performance and contribution of a golfer’s different attributes within the hole. As shown in the table above, the difficulty of the tee-shot on Hole 1 and the approach shot on Hole 2 is lost alongside the relatively easier approach shot on Hole 1 and tee shot on Hole 2. In effect, the granular differences have been cancelled out.

Solution 1: Use specific/individual holes as the averages

The PGA Tour tends to play the same courses/holes each year and performance on a hole/shot should be measured against the averages of that hole/shot. Whilst sample sizes will of course be smaller, they are likely to be large enough. And when sample sizes are not large enough, for example a new tournament/course, then use the current average system, with each new year of data replacing one year of the current (generic) average. This will improve the accuracy of what the golfer is actually measured against. It will also allow for players’ attributes to be measured to a higher level of granularity.

One issue is where a hole is modified (i.e., new bunker, new tee position), which is likely, and an adjustment is needed. For example, if a new tee box makes a par 4 ten yards longer, increasing it to 460 yards, a comparison of the expected difference is called for (i.e. expected strokes for 460 yards – expected strokes for 450 yards). This would result in adding 0.1 to the previous holes difficulty. So, if the average score for the hole is 4.25, the extra length increases the average score to 4.35.

Improvement Two: Improve Sample Size

The problem

Not all shots are measured. This is a significant issue to the Strokes Gained metrics. Below are four tables showing the top 20 ranked players in each of the Strokes Gained metrics for the whole 2019 season.

The difference between actual and measured rounds not only highlights that many shots are not measured but also the variability between players, ranging from 60-90%. The average measured rounds for the whole PGA Tour is around 75%, meaning that 1 in 4 shots are not captured in the Strokes Gained. This is like handing out the Best Player award at the end of a football season based on only 75% of matches played, with individuals ranging between 60% to 90%.

Emiliano Grillo was ranked third in Strokes Gained: Approach-the-Green in the whole of last season with his measured rounds only accounting for 64.71% of his actual rounds played, meaning that slightly more than 1 in 3 shots he hit into the green were not measured. Roughly, one quarter of the top 20 players in each metric above is measured on about 1 in 3 shots actually hit in the entire season.

This issue is further exacerbated during the season. Currently, as of writing, about half of the tournaments have been completed for 2019-2020 PGA Tour season (excluding the playoffs). The table below shows the current number-one-ranked golfer in each Strokes Gained metric (Tyrrell Hatton also in Strokes Gained: Tee-to-Green and Strokes Gained: Total). This is because small (sample) sizes yield large results – the difference between measured and actual rounds is significant.

Tyrell Hatton’s Strokes Gained average score is measured in only half of his approach shots whilst Sergio Garcia’s Strokes Gained average score is measured in only 2 out of 5 shots he hits off-the-tee and Jason Day and Louis Oosthuizen’s average scores are from roughly 2 out of 5 shots measured. What does this mean? Are the players’ averages inflated or deficient to what they should be? Possibly both. However.

If Sergio Garcia’s expected average was 2.0 then you would expect with more measured rounds throughout the season, his current (season’s) average of 1.222 to rise towards his expected average. However, if Sergio Garcia’s expected average was 1.0, then you would expect with more rounds measured, his current average to reduce towards his expected average. This result is known as regression to the mean. A person’s performance, over time, reverts back to their average.

Using the current season’s (mid-point) Strokes Gained: Total, the graph above shows that two players are currently on for the 4^th and 5^th best Strokes Gained: Total scores ever recorded, with a certain Tiger Woods holding the top three positions. This is again a result of the Strokes Gained sampling issue. Though this is not uncommon whilst a season is still only halfway through. Regression to the mean suggests that both Hatton and McIlroy’s totals will reduce over the course of the year. Adding up McIlroy’s last two seasons (2.551 and 1.412) along with this season’s current total (3.078) and dividing by 3 gives him an expected three-year average of 2.17. Similarly, with Hatton (0.938 + 0.489 + 3.078) / 3 gives him an expected three-year average of 1.5. The conclusion to draw here, is that despite both performing higher than their expected average, Hatton is outperforming his expected average by 105.2% whereas McIlroy is only by 16.9%, thus the strong likelihood of Hatton’s season average will drop. (However, due to COVID-19, this season will see a significant reduction in tournaments played and regression to the mean will have less of an impact.)

Solution

The metric must have data for all rounds (i.e., actual rounds must equal measured rounds). The fact that some players enter more tournaments will always be a factor and is not an issue provided a lower limit on the number of rounds played is introduced for inclusion in the rankings. Obviously, the ideal solution is for all tournaments that are currently not using ShotLink to start using it, namely CJ Nine Bridges, the WGC-HSBC Champions, the WGC-Match Play, The Masters, Zurich team event and all rounds at Pepple Beach.

Until this is achieved though, missing data points must be filled – an obvious requirement. One way to do this is to use a player’s individual two-year rolling average instead (see table below). For explanation purposes, the season has eight tournaments (A,B,C,D,E,F,G and H). Tournaments B and E do not measure data at the granular level, i.e., missing data. Working backwards, for the 2019 tournaments that do not have data, use the two-year rolling average Y. For the 2018 tournaments that do not have data, use the two-year rolling average X.

A two-year rolling statistic would be used to fill in missing data for the current season. However, what happens if a player has joined the tour or when the data stops at 2004? To fill in the missing data of the first two seasons, whenever that may be, another statistic is needed. For the 2017 season, fill in the missing data with n(1), which is calculated by multiplying the season average n by 0.75 (ratio of measure rounds to actual rounds).

Improvement 3: Further Adjustments Needed (Field Strength)

Problem

Adjustments are made for round difficulty, but no adjustment is made for the quality of players measured against. Strokes Gained: Total and Scoring Average factor in the course’s difficulty for that round. Thus, acknowledging the impact the course has on a player’s statistic, but fails to register the ever-changing player make-up of the field.

As mentioned previously, the nature of the PGA Tour allows players to choose which tournaments they would like to play in and those in which they do not participate. A typical season now has up to 50 regular tournaments with an additional three for the Playoffs. Rory McIlroy and Webb Simpson – the top two ranked in the Strokes Gained: Total mid-way through the 2019/20 season – have only played against each other once this season. They have, in effect, been sitting different exams papers that have different mark schemes and different pass/fail (i.e., the average) marks. This is exactly how a PGA tour season works.

Top 15 Ranked Golfers in 2019/20 season (so far) on Scoring Average

Despite streaky players, at times, outperforming consistent players (e.g., bigger variability in scores can lead to more money won), it is not a jump in faith to state that better players, on average, score better (i.e., lower). So, when a tournament has a stronger field (i.e., the average player in the tournament is better than the average player on tour), the strokes taken per round will be lower (i.e., better).

Solution

The diagram below shows that a player’s Strokes Gained statistic for both the first round and second round is +0.2. Now on the first day, the field’s average is +0.5, suggesting that the course was half a stroke easier due to, perhaps, very accessible pin positions. After this adjustment, the player’s Strokes Gained: Total is -0.3. However, due to a greater number of better players than normal playing in the tournament, who shoot, on average, lower scores (an average of +0.3 Strokes Gained), then the player’s final Strokes Gained: Total is zero. In other words, the course difficulty of +0.5 is accounted for by +0.3 of the field strength.

The second day, Player A, again has an original Strokes Gained: Total statistic of +0.2. This time, the course difficulty adjustment was -0.2, suggesting that the course was playing harder than normal, perhaps some gusty winds. Now, again, it was the same field, so a second adjustment is made giving Player A’s Strokes Gained: Total of +0.7. This makes sense because with Player A’s original Strokes Gained: Total is the same as well as the field strength being the same, it is just the difference in course difficulty between rounds. In other words, Player A’s performance was much greater in round 2 due to the course’s conditions that day being much harder.

If Player A made the cut, then rounds 3 and 4 would need to have a different field strength adjustment as the field is now different. It would be expected that a slightly higher field strength occurs at the weekend as, on average, the better players make the cut. Now, remember that Strokes Gained: Total is just a summation of four Strokes Gained metrics: Off-the-Tee, Approach-the-Green, Around-the-Green and Putting. This second adjustment should be happening within each of these before the summation. The graphic below shows a player having a Strokes Gained: Total of +0.4 for one round. For explanation purposes, it has been divided amongst the four metrics evenly, stating that the Player gained +0.1 strokes on the field in each attribute.

The course difficulty of -1.0 suggests that the particular round was one whole stroke harder. Now, why was this? For this example, it was the final round of the tournament and the greens had become incredibly hard and fast. This resulted in the field’s Putting being 0.5 strokes worse than their normal, 0.3 strokes worse Around-the-Green and 0.1 strokes worse for both Approach-to-Green and Off-the-Tee. The second adjustment for field strength has also been broken into the four individual metrics, suggesting that better than average short game players were playing but who were slightly less than average at putting.

After making adjustments at the granular level, the Player can now make fair comparisons of each of his departments. For example, his Around-the-Green metric actually shows his good initial performance of (+0.1) was even more impressive due to trickier greens that day (+0.3) and that his peers possessed a better than average short game (+0.2), resulting in a total of +0.6. These invaluable small differences would then be calculated every single round and produce more of an insight as well as making fairer comparisons between players, especially as when they do not play the same course.

Strokes Gained Summary

Are Tiger Woods’ SG: Totals inflated or deficient?

ShotLink data was not available prior 2004 when Woods was arguably at his best. Missing ShotLink data only stops the granular data being calculated. It is possible to calculate the Strokes Gained: Total metric since Tiger Woods turned professional as it can be computed using just 18-hole scores. The above data is from the founder of the stokes gained metric who factored in field strength.

We now know that not all tournaments are included in the calculation, about 75% in fact. One of the tournaments not included in Woods’ totals is the Masters in which his results are exemplary. Furthermore, many of the WGC are not measured, something again Woods had an outstanding record in, further suggesting that the above totals are deficient. What would have Tiger’s true Strokes Gained: Totals? What would have been his best… over 5.0 perhaps?

The Stat Squabbler concludes:

1. The original concept of Strokes Gained is great but is very basic. All the suggestions in this article can be, and should have been, very easily integrated by the PGA Tour as most of the data is already stored in their databases.
2. Further refinements need to be continuously made which would lead to a very impressive sporting metric (there are few at present).
3. Discrepancy between actual rounds and measured rounds is a serious issue, especially during the season when this is magnified. However, this is easily managed for Strokes Gained: Total as only 18-hole scores are needed to compute, leaving no discrepancy between measured and actual rounds. The only difference is that the metric is calculated on final round scores as opposed to summing granular data.

Do you agree with the Stat Squabbler: How good is the Strokes Gained metric? What do you think is Tiger’s true best Strokes Gained: Total?

Comment below.