Wednesday, February 19, 2014

John Paul Newport: Not the Sharpest Wedge in the Bag II

John Paul Newport is the golf columnist for the Wall Street Journal.  His lack of analytics can often lead him astray as it did in his column of July 2-3, 2011 (“Fighting Back Against Sandbaggers”).  First, he took some tables off the Internet, and assumed they were valid—this is never a good idea or good journalism.  Second, some knave misled him on the interpretation of the tables and the vagaries of the Slope System.  This resulted in multiple errors. Third, Newport pronounced the Tournament Point System (TPS) the Salk vaccine for sandbagging on the basis of two phone calls.  I doubt the efficacy of the TPS, and strongly believe sandbaggers will be eating the few remaining cockroaches when our planet comes to its end.  More detailed comments are made below.   Excerpts from Newport’s column are in italics.   

Among the most fascinating pages of Knuth’s website…is one titled “Odds of Shooting an Exceptional Tournament Score.”    A player in the 13-21 handicap range, for example, will shoot better than his course handicap only one round in every six.

The table Newport cites on Knuth’s website is also Appendix E of the USGA Handicap System.  Part of that table is shown as Table 1 below.  When the probabilities shown in the Table were first developed there was no Slope System and the probabilities were related to handicaps.  When the Slope System was introduced handicap ranges were switched to index ranges.  Such a conversion is not theoretically correct, but intellectual consistency was never the USGA’s strong suit.   Newport should have used an index range of 13 to 21.9 to be in accordance with the USGA

Table 1
Appendix E - Exceptional Tournament Score Probability Table[1]

Net Differential
Handicap Index Ranges
The values in the table are the odds of shooting a net differential EQUAL OR BETTER than the number in the left column.

Looking at Table 1, an inconsistency should be obvious.   If a player has an index between 13 and 21, he is 20 percent less likely to score a net differential of zero or better than a player with either a lower or higher index.  This is the only case where the probability of any level of performance (e.g., -2,-3…-10 net differential) does not increase or stay the same as index is increased.  Why is this index range cursed with such poor performance?  I suspect it has to do with poor research technique.  Any difference in probability among groups for scoring a net differential of 0 or better may not be statistically significant.  This hypothesis cannot be confirmed, however, since the USGA does not release its studies for peer review.    When I mentioned this anomaly to the USGA, the probability for the 13.-21.9 index player was changed from 6:1 to 5:1 in the 2012-2015 edition of The USGA Handicap System.  The USGA never admitted a mistake nor cited new research for the change.

The odds of a midrange player shooting eight strokes better than his course handicap are 1 in 1,138.  The odds of him doing that twice in 20 rounds (much less than in a two- or four-round tournament are 1 in 14,912).

There are several errors in this quote.  Table 1 show the probability of shooting a net differential of  -8 or better, not just -8.  Newport also errs in using handicap instead of index.  An example will make this clear.  Assume a player is a 13.0 index.  He plays a course with a course rating of 72.0 and a Slope Rating of 150.  His course handicap is 17.  Assume he shoots eight strokes better than his handicap (i.e., his net score is 64 and his gross score is 81).  The player’s handicap and net differentials are:
Handicap differential = (81-72) x 113/150 = 6.8
Net differential = 6.8 -13.0 = - 6.2 rounded to -6.0
So in this case,  the probability of a player shooting 8 strokes or better than his handicap is 323:1 and not 1,381:1.  The probability of the player doing that twice in 20 rounds is 1 in 3,385, and not 1 in 14,912. (Note: These latter probabilities are taken from a Table shown in Knuth’s website that also has technical problems that are not discussed here.)   In his defense, Newport was just using “handicaps” as instructed by Knuth.  The fact that Newport did not notice the inconsistency between “handicap” and “index” demonstrates Newport’s shortcomings as journalist.

But clubs or associations that run numerous handicapped or net tournaments are using another more effective Knuth invention called the Tournament Point System.

Newport judged the Tournament Point System (TPS) effective based on little evidence.  It appears the TPS has been around for at least 7 years and possibly much longer.  If it was that effective, it should have swept the nation.  It appears Newport only found two clubs using the Point System, one of which may have been Knuth’s home club in California.  The other club that extolled the TPS was Lost Creek Country Club in Austin Texas.  A call to the Lost Creek pro shop in 2014, however, revealed the TPS it is no longer in use.   
A club in Oregon implemented a version of the TPS nine years ago.  It was not a fair test, but it did reveal critical flaws in the system:
Not Effective at Many Clubs – Many clubs do not have a large number of major tournaments necessary for effective implementation of the TPS.  Often, the Member-Guest is the only tournament with significant prize money.  Under the TPS, a player could win the Member-Guest every year and not be penalized.  If minor events are put under the TPS, the strategic player could chose to sit out these tournaments rather than risk a penalty for an up-coming major tournament.  The artful sandbagger could also go unpunished under the TPS.  If the TPS cuts two strokes, he merely submits scores that raise his handicap by two strokes.
Double Whammy – Assume a player with a 15 index had two net tournament differentials of -8.  Under section 10-3 of the USGA handicap system, a player’s index could be reduced to 8.2.  On a course with a 133 slope rating his handicap would come down from 18 to 10.  Under the TPS his handicap would be further reduced by another three strokes making him a 7 handicap.  This is draconian.  Now the tournament committee may think this unfair and not apply the TPS reduction.  But if the objective of the TPS was to remove discretion, it would have failed.
Too Complicated and Time Consuming – The TPS is not as simple as suggested in the Newport’s column.  When the Oregon club used the TPS for weekly and flighted events, just about everyone was getting a penalty.  The latest version of the TPS has now been changed to alter the point distribution and penalties as a function of flight size and number of tournaments played.  And who is going to keep track of all of this?  Pros certainly don’t want to do it.  Volunteers were relied upon at the Oregon club.  This led to haphazard record keeping.  Some tournaments were recorded, others were not.  Members showed their dissatisfaction by declining to participate in tournaments.  The TPS was quietly killed at the end of the season.

In summary, Newport’s main fault appears to be he believes anything that contains a modicum of mathematics or statistics.  This should be a disqualifying fault for being a Wall Street Journal columnist.  Unfortunately, it is not.

[1] Note: Appendix E has been abbreviated here.  It actually shows odds up to a -10 net differential

Tuesday, February 4, 2014

John Paul Newport: Not the Sharpest Wedge in the Bag - I

Introduction - John Paul Newport is the golf columnist for the Wall Street Journal.  He does well when his column is purely descriptive (e.g., my round with Donald Trump), but struggles when the subject is more technical.  His difficulty is exemplified by two columns where Newport extolled the work of Mark Broadie and his “strokes gained statistic.”[1]  Both columns make the same points, and both promote Broadie’s forthcoming book, Every Shot Counts.   More problematic is why Newport is promoting a book that is still “forthcoming” some 20 months after he first wrote about it. 
Both columns argue that Broadie’s work has debunked myths about the determinants of scoring.  If Newport had a background in player performance research, he would have known that 1) the myths listed in the column are straw men or myths that have been debunked a long time ago, and 2) the putting myth may not have been debunked at all.   If he had a technical background, he could have challenged Broadie on some of the author’s findings.  And finally, if Newport weren’t so enthralled, or overwhelmed, by Broadie’s mathematics he could have examined the value of Broadie’s research to the average golfer.  Each of these shortcomings is examined in turn.
Demystifying Broadie -Many of Broadie’s findings can be found in Cochran and Stubbs’ book, The Search for the Perfect Swing,[2] hereinafter referred to as Cochran.  In fact, even the “strokes gained” nomenclature was first used by Cochran.  Cochran studied 40 golfers at one tournament at minimal cost.   Broadie’s research tab was probably in the millions of dollars when the cost of collecting the data is added.  
Myth #1 - Driving Accuracy is Important
Newport wrote that Broadie’s research has shown “driving contributes 28% of top player’s advantage, and the main factor there is raw distance, not accuracy.”[3]  Forty six years ago Cochran wrote “In fact, putting it rather loosely, distance from the tee seems to count for more than accuracy.”[4]  So there was no myth about the primacy of accuracy or if there was it was debunked in the last century. 
Myth #2 -Long Approach shots are not that important.
Newport wrote “The biggest factor in lower scores, however, is approach shot accuracy. It contributes 40% of the advantage that the Tour’s top players enjoy over their peers…For Woods, by far the top strokes-gained player between 2004 and 2012, it was even more: 46%.”[5]  Of course, Cochran found the importance of the long approach shot and even entitled a chapter “Long Approach Shots—Where Tournaments are Won “ [6]  Sorry, no myth debunked here.
Myth #3 – Lay up to a good distance is better than getting it as close as possible.
Newport  wrote that Broadie concluded ”that laying up short to a green to a comfortable distance, such as 100 yards doesn’t generally lead to better scores than knocking the ball as close as possible.” [7] This is consistent with Cochran’s finding that pros hit their half of their approach shots between 7.5% and 8 percent of the range from which each stoke was played .  Moreover Cochran found that the number of shots to get down from approaches of more than thirty yards from the hole could be estimated by:[8]
                Number of shots to get down = .0044·d +2.35
Where ,
                d = the distance in yards.
Broadie reported a similar estimating equation:[9]
                Number of shots to get down = .0041·d + 2.38
In essence, both recognized the importance of getting close except in certain circumstances.
Myth #4 –Putting is not golf’s most crucial skill.
Newport wrote “The No. 1 shibboleth Broadie’s research debunks is that putting is golf’s most crucial skill.”  Broadie and Cochran have a small disagreement here, but that could be due to methodology or semantics (i.e., how “crucial” is defined).   Cochran looks at the proportion of putts holed from various distances for the top nine and bottom nine finishers in the tournament studied.  The top nine were more proficient than the bottom nine at every distance.  Cochran estimates the top nine players gained (emphasis added) 1.8 strokes per round over the bottom nine by better putting alone.   Since the difference in scoring between these two groups was 4 strokes, 45% of the difference was the result of putting.  That is, putting was crucial in determining the tournament’s outcome.
Newport wrote that “Broadie finds that on average putting contributes 35% of Tour winners’ advantage.”  This is not far from Cochran’s estimate of 45%.  Since Cochran’s comparison is with the bottom finishers and Broadie’s comparison is with the average finisher, it is likely that both would conclude putting contributes to approximately 35% of a Tour winners’ advantage.  Newport takes this figure of 35% and implicitly concludes that putting cannot be that important if 65% of a winner’s advantage is due to something else.[10]    There are three problems with Newport’s reasoning. 
First, the largest contributor to a golfer’s success depends on how you breakdown the components of scoring.  Table 1 below shows the contributions different shots make to Tiger Woods “strokes gained” statistic.  From Table 1 you would conclude the two biggest contributors to the success of Tiger Woods are driving and putting.  You would not use Newport’s construct that driving is not that important since it only contributes 22% of Wood’s total strokes gained (i.e., the other parts of his game contribute 78%).

Table 1
Strokes Gained by Tiger Woods (2003-2010)[11]

Long Tee

Second,, if a player does not gain strokes by putting, it does not necessarily mean he putted poorly.  Let’s assume a three hole tournament.  Player A sinks 2 out of three putts from 10 feet.  Player B sinks 1 out of three putts from 30 feet.  Player A’s stroked gains putting is .83, while Player B’s strokes gained putting is .94 strokes.[12]  Player A wins the tournament, but comes in second in the “strokes gained putting” statistic.   Unlike Cochran, Broadie does not compare the putting ability of players for the same type of putt.   A player who hits par fives in two and sticks his irons close, may not have an impressive “strokes gained putting” statistic.  This is not the same as saying such players are not good putters.
 Third, the importance of putting in determining who wins may be due to its variability.  Players typically drive the same distance each week, but can have extraordinarily good and bad putting tournaments.  Using Broadie’s 2010 data, there was essentially no correlation between average “strokes gained putting” and money winnings per tournament (see Appendix).  The PGA Tour has a non-linear payoff, however, and a good putting week can vault a player to the top and to a big pay check.  Putting is clearly not the sole determinant of who wins, but for many players it has been crucial.
Lack of Critical Analysis – Newport fails to question anything that has a hint of mathematics.  Here are some examples:
What errors are involved in Strokes Gained Analysis – It appears Broadie uses the same benchmark for all courses on the Tour.  This could lead to errors.  Assume putts are tougher to make at a course because of the undulation and speed of a green.  A player hits an iron from 160 yards to 10-feet of the flagstick.  Now from 160 yards the player is expected to complete the hole in 2.98 strokes. [13]  From 10-feet Broadie estimates it would take an additional 1.61 strokes.[14]  Broadie would estimate the strokes gained from the approach to be 0.37 strokes.  But what if the probability of making a 10-foot put on this course were 1.81 strokes?  In that case, the strokes gained by the approach shot would be 0.17 strokes, not 0.37 strokes.  If the putt was made, Broadie would estimate the strokes gained by putting at .61 when it should be .81 strokes.   Broadie does try to normalize the “total” strokes gained.  It is not apparent if errors in the estimates of strokes gained by shot category are corrected.
Some Benchmarks Are a Little Strange - Broadie estimates is takes 2.99 stokes to complete a 120-yard hole if the ball is teed up.[15]  He also estimates it takes only 2.85 strokes to complete a hole from 120-yards from the fairway. The advantage of playing from the fairway vanishes at distances of 140-yards or more.  Broadie does not explain this anomaly.  Broadie also estimates it would take 6.1 strokes to complete a hole if the player is in a sand bunker some 600 yards from the hole.  It is not clear on what data set such an estimate could be made.
The Listing of the Most Difficult Courses on the Tour Appears Biased Toward short Courses -   TPCSawgrass and Harbour Town Golf Links are listed as the first and third most difficult courses on the PGA Tour.[16]  These are two of the shorter courses on the Tour.  Broadie’s methodology assumes both courses play to their yardage (i.e., scores should be lower there reflecting the short yardage).  The “effective” playing lengths of these courses, however, are much longer.   The increased effective length stems from the driver not being used on many holes at these two courses.  If the difference between actual and effective length is 300 yards, Broadie would underestimate the benchmark total score by 1.32 strokes (i.e., 300 yards·.0044).   If the effective yardage was used in the strokes gained analysis, the difficulty ranking of each of these course drops considerably.
Usefulness to the Average Golfer - If Broadie’s research was valuable to the average player, Newport’s over promotion of the forthcoming book would be understandable as a public service.   Newport’s columns, however, do not address how the average player would go about using the findings.  The promo for the Broadie’s book claims “Broadie uses analytics to uncover the secrets of the game of golf.”   This is pure hyperbole.  There are no secrets, and to imply that there are is misleading at best. The publisher’s promo goes on to ask “What does it take to drop ten strokes from your golf game?” and implies the answer lies within the book covers.   I suspect the answer is the typical bromide – hit it longer, straighter, and improve around the greens.  
The real policy question that Newport avoids is this: A player has limited money and time to invest in improving his game.  What is the most cost-effective use of those resources?    A player should be able to spot his deficiencies without Broadie’s book.  When he plays with better players, he needs to note what part of his game leads to higher scores than his playing companions.   He may realize he does not hit it as far, but also recognize increasing his distance is not in the cards because of his physical limitations.  Conventional wisdom says he will see that his short game suffers in comparison.  The better player does not skull his chips, flail away in sand bunkers, or three-putt with regular frequency.  Since short game skills are a matter of technique rather than physical strength, concentrating on this part of the game seems most cost-effective use of an average golfer’s resources.    Maybe Newport will have a third column that debunks this conventional wisdom.  Until he does, a prospective reader should pass on Every shot Counts and save $24.92 (plus shipping and taxes).  Remember, every dollar counts.

Money Winning per Tournament versus Strokes Gained Putting
Table A shows the strokes gained putting and average winning per tournament for a sample of players.  A linear regression was performed with Money/Tournament as the dependent variable, and Strokes Gained Putting as the independent variable.  The estimated equation was:
                Money/Tournament = 91,000 +25,600 ·Strokes Gained Putting    R2 = .02
The coefficient of the Strokes Gained Putting variable was not significant at the 95% level of confidence.
Table A
Strokes Gained Putting and Money Winnings per Tournament for Selected Players

Strokes Gained Putting
Donald, Luke
Casey, Paul
Petterson, Carl
Wilson, Dean
Chalmers, Greg
Gay, Brian
Goosen, Retief
Kuchar, Matt
Collins, Chad
Johnson, Zack
Wi, Charlie
Imada, Ryuji
Sim. Michael
Jones, Matt
McCarron, Scotrt
Merritt, Troy
Baddeley, Aaron
Na, Kevin
Molder. Bryce
Snedeker, Brandt
Westwood, Lee
Stricker, Steve
Furyk, Jim
Els, Ernie
Toms, David
Harrington, Padraig
Poulter, Ian
Fowler, Ricky
Duval, David
Johnson, Dustin
Choi, K.J.
Weir, Mike
Woods, Tiger
Appleby, Stuart
O'Hair, Sean
Mickelson, Phil
Cabrera, Angel
Sabbatini, Rory
Garcia, Sergio
Mediate, Rocco
Singh, Vijay
Scott, Adam
Mayfair, Billy

[1] Newport John Paul, “The Short Game’s the Thing? Nope,” Wall Street Journal, June 8, 2012 and Newport, John Paul, “Research Debunks Golf Myths,” Wall Street Journal, January 16, 2014.
[2] Cochran, Alastair and Stobbs John, The Search for the Perfect Swing, The Bootlegger, Grass Valley, CA, Copyright 1968.
[3] Newport, 2014.
[4] Cochran, op. cit., p. 184
[5] Newport, 2014.  Broadie reports the average strokes gained from the long tee by the top ten players  is .61 strokes.  The average strokes gained by these groups is 1.84 strokes.   The advantage due to driving would then be 33% of the total not 28%.  The difference could be due to a different time frame.  In his article “Assessing Golfer Performance on the PGA Tour” Broadie uses data from 2003-2010, and not 2004-2012.  Interestingly, Broadie reports different numbers in two different papers.  In “Putts Gained: Measuring Putting on the PGA Tour” Tiger Woods in ranked 116th in putts gained for the year 2010.  Tiger Woods was ranked 91st in 2010  in his paper “Assessing Golf Performance on the PGA Tour.”  
[6] Cochran, op. cit., pp. 196-202.
[7] Newport, 2014.
[8] Cochran, op. cit., p. 198.
[9] Broadie, Mark, “Assessing Golfer Performance on the PGA Tour,” Columbia University, 2011, p. 8.
[10] Newport, 2014.  Newport writes “65% of a winner’s advantage comes from off the green.”
[11] Broadie, Assessing Performance, , p. 18.
[12] Broadie, Mark, “Strokes Gained, Measuring Putting on the PGA”, Columbia University, 2011, p.4.
[13] Broadie, Assessing Performance,  p.31.
[14] Broadie, Strokes Gained,  p.4.
[15] Broadie, Assessing Performance, p. 31.
[16] Broadie, Assessing Performance, p.26..