Quote:
Originally Posted by Doug Funnie II
I had some free time this morning at work so I decided to take a quick look at some FedEx cup stats.
I looked at FedEx cup points per event played, average driving distance, % of drives in the fairway, and putts gained to perform my analysis. I assumed all metrics were normally distributed.
Some quick observations:
- Bubba Watson really is an incredible driver of the golf ball. His Total Driving Z Score (defined as (driving distance - tour average driving distance)/ St. Dev of Tour Driving Distance + (driving accuracy - tour average driving accuracy)/St. Dev of Tour Driving Accuracy) is 3. The next highest guy, JB Holmes is ~1.87. The difference between Bubba Watson and JB Holmes is about the same as the difference between JB Holmes and Matt Kuchar. However I think this is probably due at least in part to the fact that driving distances are likely not normally distributed.
-Mike Weir is twice as bad at driving the golf ball as Bubba Watson is good (relative to their peers, of course).
-John Merrick and David Toms are really good at both driving and putting.
-Of the top 5 guys in terms of points per event (Luke Donald, Charl Schwartzel, Nick Watney, Bubba Watson, and David Toms) all of them are in the top 30 for total driving + putting. FIGJAM and KJ are # 6 and #7 respectively, and frankly their driving + putting scores kind of suck. This could indicate that they do a lot of the stuff between the drive and getting to the green extremely well (true), or that they have 1-2 solid performances and a bunch of mediocre ones, but given the distribution of how points are awarded, their point totals are artificially inflated above expectation.
Now how do these Z scores translate to predicting who has the highest fed ex cup points per event? Honestly, I'm not sure because I didn't look at any other variables yet, so it's hard to get a grasp of the significance of the effect. Overall it appears that an improvement in total driving by 1 Z score (so adding 8.3 yards off the tee at no reduction in accuracy, or adding 5.8% to your driving accuracy at no reduction in distance, or some combination of the two) leads to around 11.5 more fedex cup points per event played. By comparison, a 1 Z score improvement in putts gained (increase of .43) would predict a 9.1 point improvement in points per fedex cup start. Both regressions are statistically significant at the 1% level, however both still have relatively low predictive powers as the R squared coefficients are .178 for driving and .110 for putting.
Ok, I updated and expanded my analysis from last week with the most current data and what I hope to be slightly better methodology. I changed some things about how total driving Z scores were calculated (I believe the old methodology was biasing the data in a preventable way) and also looked at average distance to the hole after approach and scrambling %. The results are similar:
-Bubba Watson is still a disgustingly good driver of the golf ball. I believe that he hits below average approaches given his tee shot locations (although that's very hard to isolate). He is one of the worst scramblers and also a negative putter. If he were to bring his scrambling game up to average (without an improvement in putting or approach shots), my analysis predicts that he'd be ~7th in average scoring. Currently he's 86th.
-Sergio is a good driver and scrambler. He is below average with approach shots and very poor at putting. My analysis predicts that he should have around the 105th best scoring average. Currently he's 7th. He is amongst the most conspicuous "overperforming" outliers along with Kyle Stanley, Ricky Barnes, and Chris Kirk. Peter Tomasulo and Roland Thatcher appear to be large underperformers. Mike Weir, Billy Horschel, Rocco Mediate, and Ernie Els have all been playing right in line with their results, which is to say, terribly.
-In terms of overall quality of predictions, my bastardized multivariate regression is relatively good. The four metrics that I looked at explain roughly 58.4% of the variance in scoring. When I redo this analysis for a 3rd time I'm sure I'll find more holes and tighten it up a bit more. A bigger sample size won't hurt either.
-Scrambling appears to be the most important of the variables, supporting the hypothesis that the N-1 shot is the most important. A 4.3% improvement in scrambling yields a half shot saved per round.
-Total driving is second most important. A 7.3 yard increase off the T or a 5.5% bump in accuracy corresponds to a .35 shot decline in scoring average.
-Putting is a close third. I still haven't really looked at how the putts gained metric is computed, so I haven't mulled over whether it is appropriate to use in this sort of analysis, or if the risk of multicolinearity is too high to look at on its own.
-Average distance after approach is a distant 4th. That surprised me. The best explanation here is probably that when the average shot lands 25 feet from the stick, and 95% of golfers have an average between 18 feet and 31 feet, there really isn't much to be gained from a small improvement (say going from 25 to 22 feet).