'Shots Gained' Explained

A crash course on the statistics we use when modeling golf, also called 'Strokes Gained'

I am hoping to have some occasional ‘evergreen’ posts, usually on Fridays. In theory, these will be less topical and more educational.

What Are ‘Shots Gained’ Anyway?

The majority of PGA events have this fancy tool called Shot Link installed, it’s what allows the TV to show you the trailing line behind the ball after a guy drives. It’s also collecting a ton of data that can be consumed. We’ll get to the types of stats we can work with in a second, but let's start with how the numbers should be interpreted.

When a shot is ‘gained’, it means it was gained relative to the field for that round. If we used raw data, there are too many variables that are ignored; courses all have varying challenge levels, and the factor wind plays will change round to round.

An example…let’s say someone hits 8 out of 14 fairways. Without any context of his peers, we have no idea if 8 is good or bad. If the average was 10, then 8 was a below average day. If the average was 6 fairways, then 8 likely led to an above average day. In the world of ‘shots gained’, an average of 6 fairways by the field with our 8 hit would be represented ‘+2 shots gained fairways’.

Here is a much longer explanation of how they are calculated.

Types of Shots Gained

There are a lot of different ways to breakdown the data, but it all starts with the main four categories:

• Shots Gained Off The Tee (SG: OTT) - Exactly like it sounds, this shows how someone is performing from the tee box on non-par 3’s.

• Shots Gained Approach (SG: APP) - This is going to show how someone if performing hitting into greens; historically this is the most predictive statistic.

• Shots Gained Around The Green (SG: ARG) - Measures performance from shots within 30 yards of the edge of the green; stat most likely to be ignored on easy courses.

• Shots Gained Putting (SG: P) - Obvious one, but quite relevant; this is most useful with some filters on the data (more on that in a minute).

Parallel to these stats above, there are some more granular stats you will see me reference:

• Fairways Gained - Measures purely fairways hit, useful on courses with very penal rough

• Distance - Exactly as it sounds, this ends up being factored in frequently

• Good Drives Gained - The closest to SG: OTT, this counts fairways AND any missed fairway that still led to a GIR

• Proximity Ranges - We can see stats for approach shots from any range, set at 25 yard intervals; some courses set up consistent approach ranges, so this stat becomes useful from time to time

• Sand Saves Gained - Measures a players performance relative to field from the sand, useful on courses where bunker count is higher than usual

• Scrambling Gained - Similar to Sand Saves, but also considers any missed green in regulation (GIR); more useful than Sand Saves in my opinion

• 3 Putt Avoidance - Straightforward metric, most useful on courses with large greens or anything known for difficult putting

• Putting Ranges - We have stats on any range of putting at five foot intervals; I have found that 5-10 foot putting performance (filtered to course settings) has been a valuable metric to include in safety models

• Par 3/4/5 Ranges - Some courses will have holes of similar distance, and we have stats on player performance by hole length at 25 yard intervals; often useful when the longer courses come around

• Birdie or Better Gained (BoB) - Measures birdie count relative to the field; we use this at true birdie fests (example player: Matt Fitzpatrick is not great at this, so birdie fests are a time to look elsewhere)

• Opportunities Gained - Similar to BoB, but this measures how often a player had a realistic chance for a birdie; not factoring if they converted like BoB

• Bogey Avoid Gained - The inverse of BoB, this is seeing who cards the fewest bogeys (example player: Matt Fitzpatrick is great at this, so a tougher course is going to suit him well against the field)

How Can We Use Shots Gained?

Every course on tour is going to require slightly different skills. When we do our homework on researching the course and what to expect, we can start to identify which of the statistics above will be most predictive. For example, on a longer course we will want to factor in ‘Distance’, whereas a short course may not consider it at all. Ultimately, we are trying to find a ‘horse for the course’.

All of the statistics listed in this email can be pulled in a variety of ways; filters exist to hone in on the most useful ones. When modeling, I like to strive for a blend of recent form and historical fit - I am able to do that by using different filters on the statistics above, then combining them into a ‘weighted model’..

Using Putting as an example for a filterable stat (SG: P), we may value overall putting performance across the past two years, but it likely becomes more useful if we filter SG: P by the type of grass that week’s course has on the greens. What you will find is that some guys perform much better on certain grass types, opening the door for an edge on the odds we find on them. The same filtering can be applied to any metric, choosing the right ones being the part we refine over time.

Another filter option I have found useful is to filter by certain courses. This provides value in finding who has played well at exact courses in the past (course history), but can also find value by showing who played well on comparative courses. Each week, I try to establish which other courses on tour are similar to the one at hand, which makes ‘Comp Courses’ a weekly staple in my modeling.

To find this data, we have a variety of options. The PGA themselves publish a lot of data we can use, but with fewer built-in filter options. Data Golf is likely the most robust source of data, but can be hard to work with unless you and more advanced technically. I have tried most of them, and found Fantasy National to be the best option for cost/value. When I show statistics in my model, or end results, the majority of data is coming from there.

Closing Thoughts

As you learn more about the data, working with it each week becomes pretty fun. We’ll find angles to consider with undervalued as well as overrated guys and can learn more about when to have certain players in mind. That being said, this is still not an exact science. You, loyal reader, have likely played golf before and grown frustrated by how inconsistent you are at times - while the pros are way better than you, the same thing happens all the time.

If you use my models, someone else’s, or make your own, I encourage you to treat them as just one piece of the puzzle. You need to decide the guys you like each week and shop for the best odds - I can say from experience, my betting results are much better when I treat the model as a guide and not an answer key.

Best of luck now and in the future in our quest to make money gambling.

Matt / Ziggy

If you are reading this, you probably know me personally. But either way, thank you for taking the time to read this far, I appreciate it. If you know anyone else who likes pro golf and/or betting, please forward this to them and encourage them to sign up.