2

Dominance and Chance

There is a widely held and mistaken view in armchair sports analysis that the great teams separate themselves from the merely average ones in their ability to win the close games. If a baseball team is in first place in early July and is also winning a disproportionate number of one-run games, media analysts are guaranteed to point this canard out repeatedly. It’s a belief based on the gut-level assumption that most games are decided by a run or two. Ergo, teams that triumph over the long haul must necessarily do well in those close games.

In fact, as those who study baseball closely have realized for some time, the actual relationship between winning close games and winning a lot of games is surprisingly casual. The reason is as evident as it is overlooked: outcomes of close games are more likely to be determined by chance occurrences. When one says a team is winning the close games, what it means to some extent is that they’re lucky. Luck is not much of a strategy.

Far stronger than the relationship between winning close games and winning a lot of games is its opposite, the relationship between winning routs and winning a lot of games. The best teams don’t squeak by their opponents; they pound them. The worst teams don’t get edged out at the margins; they get hammered. In 2017 eight Major League Baseball teams averaged at least a half run per game more than their opponents. Those eight had a .596 winning percentage. At the other end of the spectrum, ten teams averaged at least a half run less than their opponents. The winning percentage of those teams was .428. Of the eight teams that excelled in run production, seven fared worse in one-run games than they did generally with a combined .555 winning percentage in those close games, 41 percentage points below their collective season-long average. The ten teams at the extreme negative end of the run-production scale had a combined .454 winning percentage in one-run games, 26 percentage points better than their season-long winning percentage. Eight of the ten did better in one-run games than they did generally.

There is no reason to believe that in other sports—and notably for this book’s purposes in golf—the statistical axioms are different. You don’t excel by winning the close ones; you excel by dominating. But as one examines the role of dominance in top-level golf, one would be well served to take a brief detour into recent history.

In 2007, encompassing fourteen medal-play tournaments, Tiger Woods averaged 67.79 strokes per round. The tour as a whole averaged 70.94 strokes. Put another way, the field needed 100 strokes to play the same number of holes Woods could play in 95.5. That means Woods was 4.5 percent better than the average player.

As with any other carbon-based life form, Tiger had his good weeks and his bad weeks, although in Tiger’s case the bad weeks weren’t very bad. What one might consider his “normal” performance range over that period—mathematicians refer to this as the first standard deviation of his performance—was about 19 percentage points, meaning that about two-thirds of the time Tiger’s score fell between 93.6 and 97.4 percent of whatever the field was doing in a given week.

Most of the remaining one-third would be encompassed by a second standard deviation. That would encompass about 97 percent of all events, in Tiger’s case extending his predictable performance range to a maximum of 99.3 percent and a minimum of 91.7 percent of the field average. Any Woods performance above 99.3 percent of the field average or below 91.7 percent of that same number, while possible, would be freakishly rare. In fact, there were none in 2007. His worst effort came at the Players Championship, an even-par 288 that was good for a tie for thirty-ninth (his score being 98.9 percent of the field average). His best was 93.0 percent at the PGA Championship, which he won by three shots. A simple graph illustrates the percentage likelihood of a Woods performance at various levels relative to the field average for the 2007.

There was a 99 percent likelihood that Woods’s score would exceed the field average (point A), approximately a 50 percent likelihood that it would be 95 percent or less than the field average (point B), and at least a measurable chance that it could go as low as 92 percent of the field average (point C). On tour, 95 percent is often a winning range, so you can readily understand why Woods won seven tournaments in 2007.

Since Woods does not play in isolation, these numbers really need to be interpreted in the context of his competitors. On the 2007 PGA Tour, approximately two dozen golfers averaged 99 percent of or below the field average for the events in which they played. That is to say, their average was consistently within two standard deviations of the top of Tiger’s predicted range of 99.3 percent, giving them at least a plausible chance of matching him on one of their good weeks. Although literally anybody on tour is good enough to conceivably win a tour event, those two dozen constitute the most likely challengers to Woods. Consequently, the most meaningful way to assess Tiger is to compare him to them.

Of those nearly two dozen, the most consistently dominant in 2007 was Ernie Els, whose scores averaged 98.1 percent of the field average for the fifteen medal-play events in which he competed. Els’s normal variation—his first standard deviation performance—was 2.5 percent, which is wider than Woods’s 1.9 percent. (He’s not alone in that; only Chris Riley matched Woods’s performance variation in 2007, and Riley’s average score, 99.7 percent of the field average, was far higher than Woods’s. The average variation among the top men’s pros was about 2.5 percent.) The bottom line is that Els could expect to use between 95.6 percent and 1.006 percent of the field-average number of strokes, with a prospect that his performance could fall as low as 93.1 percent or rise as high as 1.031 percent.

Chart the 2007 Els against the 2007 Woods, and Ernie’s problem beating Tiger comes into focus.

In other words, 90 percent of the time, Tiger 2007 would shoot what Ernie 2007 would shoot about half the time. Conversely, Ernie shot Tiger’s 50th percentile score only about 15 percent of the time. From this, roughly estimating Ernie’s prospects for actually beating Tiger is simply the product of Ernie shooting at a certain level multiplied by the prospect of Tiger not doing so. Let’s assume, for illustration, that Tiger shot his 50th percentile score. The chart estimates that there was about a 15 percent chance in 2007 of Els scoring at that same level. Fifty percent times 15 percent is 7.5 percent. Ergo, when Tiger played his normal game in 2007, Els had approximately a 7.5 percent chance of beating him. In reality, Ernie did a bit better than that, besting Tiger twice in the fourteen 2007 PGA Tour events in which they both competed and matching his score two other times. In the remaining ten tournaments, Tiger beat Ernie.

One can make the same estimates up and down the performance range. It is known that Ernie’s normal score was about 98 percent of the field average. It is known that there was about a 90 percent prospect that Tiger would shoot within 98 percent of the field average—which means there was a 10 percent chance he wouldn’t. So if Ernie merely shot his normal round, his prospects of beating Tiger were 50 percent times 10 percent, or about 5 percent.

Thus far, the task of beating Tiger in 2007 sounds pretty much hopeless, which it would have been if the contest were simply between Woods and any single competitor. But a typical PGA Tour field consists of about 150 entrants, whose range of talents is remarkably compressed. Among the nearly 200 players who competed in at least ten tour-certified tournaments in 2007, the scoring average range between the very best (Woods, 67.79 strokes) and very worst (Todd Hamilton, 72.83 strokes) was less than 7 percent. Given the potential week-to-week variance for so many bodies, that is as inconsequential as it is possible to get.

Even so, during his fourteen 2007 medal-play starts, Woods was bested by only 68 different players, a fascinatingly small number considering that fourteen tournaments, each consisting of about 150 competitors, amounts to roughly 2,100 opponents. The news was actually worse than that for Woods’s pursuers, because 33 of the 68 did it in the two tournaments where Tiger played poorly (by his standards), the Palmer and the Players. In the other dozen tournaments in which Woods competed, just 35 players managed to finish in front of him even once. Only 3 men—Stewart Cink, Sergio Garcia, and K. J. Choi—beat Woods as many as three times.

It has already been established that the upper end of Tiger’s expected performance range in 2007 was 99.3 percent of the field average. In 2007 only 13 men’s tour players produced a dominance rating of .993 or lower, and those 13 can probably be characterized as the guys Tiger needed to pay attention to. Listed from most dominant to least dominant, those 13 included:

As demonstrated earlier, none of these guys individually posed a consistent threat to Woods, since in 2007 he was likely to shoot their midpoint score about 90 percent of the time. But considered as a pack, they were a more legitimate force. Their average dominance rating was about 98.8, and the average variance of their performance was about 2.5 percent, setting the low end of the first standard deviation of their predicted performance at about 96.3 percent. That number, not far from Tiger’s own 95.5 percent midpoint, meant that in any given week, the question of Tiger versus this particular field of most likely pursuers was fundamentally a toss-up, an assertion borne out by Woods’s seven victories in his fourteen starts. His seven defeats came to Choi (the Memorial and the AT&T National), Singh (the Palmer), Harrington (the British Open), Johnson (the Masters), Phil Mickelson (the Deutsche Bank and the Players), and Angel Cabrera (the U.S. Open).

In addition to their wins head-to-head against Woods, those 13 players won six of the thirty-two events in which Woods did not take part. Again, that proportion of wins makes sense because when Tiger was removed from the picture in 2007, the talent spread among all other golfers was so small that what developed was pretty much a fair fight. While the average dominance rating—about 98.9 percent—of the 13 most likely challengers was definitionally better than the 100 percent dominance rating for the “field,” the “field” had one huge advantage, size.

To this point, largely for purposes of simplicity, the analysis has hinged on the prospect of Tiger 2007 shooting his midpoint score. There is no law stipulating that he would do so, and when he posted a score in the lower third of his personal range, the fact that his range was so much lower than anybody else’s made him essentially invulnerable. His performance chart suggests that in 35 percent of his 2007 tournament competitions, Woods should have put up a four-round total that was 5.4 percent lower than the field average for the tournament in question. (In fact, he did even better, with very low scores six times in fourteen medal-play starts, or 42.9 percent of the time.) Even among the 13 tour players identified as Woods’s most legitimate threats in 2007, only 2—Els and Furyk—had even a 10 percent statistical chance of separating themselves from the field by 5.6 percent. That puts their individual likelihood of matching Woods when he goes low at 35 percent times 10 percent, or 3.5 percent. And in real life, Tiger’s statistically improbably good season made the likelihood just 4.7 percent. In real life, Tiger played eight events in which his score was better than one standard deviation below the average dominance rating of his plausible pursuers—that is, below 96.3 percent—and he won seven. Combined, Els and Furyk had twelve opportunities to beat Woods when he was playing at that level, and they went a collective zero for twelve. Here’s the full table, with the lowest dominance rating in boldface.

Event Woods Els Furyk Other

Buick Invitational

96.0

DNS

DNS

Doral

95.2

97.6

100.4

Wachovia

93.8

98.2

102.3

Bridgestone

93.4

98.6

DNS

PGA

93.0

94.0

99.9

Deutsche Bank

96.2

DNS

101.2

95.5a

BMW

93.9

99.3

98.9

Tour Championship

94.0

102.4

99.9

Note: DNS = did not start. a. Phil Mickelson.

Fast-forward now to 2015. The most dominant player on the PGA Tour was Jordan Spieth, with an average of 97.9 percent and a standard deviation of .024. That means the first standard deviation of Spieth’s expected performance range—a performance he could be expected to deliver about two-thirds of the time—was between 95.5 percent and 100.03 percent of the field average. In fact, Spieth made twenty-three starts on the PGA Tour in 2015, of which eighteen landed in the first standard deviation and two within the second standard deviation. The remaining three represented the three times Spieth missed the cut.

Spieth won five times on tour in 2015, two fewer than Woods in 2007, but perhaps even more remarkable given the difficulties Spieth faced that Woods did not. As noted above, Woods faced 13 competitors who could reasonably project to score within two standard deviations of Woods’s upper performance range, 99.3 percent. The upper limit of the second standard deviation of Spieth’s performance in 2015 was 1.027 percent. Aside from Spieth, there were 147 players on tour in 2015 who compiled dominance ratings of 1.027 or lower. Repeat: Woods had 13 significant challengers in 2007; Spieth had 147 in 2015.

Not only was the competition deeper in 2015 than 2007, it was far closer. Recall that in 2007, the runner-up to Woods’s 95.5 dominance rating was Els at 98.1. That’s a spread of 2.6 between first and second. In 2015 the spread between Spieth and the runner-up, Jason Day (98.0), was just 0.1. The third most dominant player, Henrik Stenson, was at 98.2, followed by Rory McIlroy and Bubba Watson, both at 98.6. Within the 2.6 spread that separated Woods in first and Els in second during 2007, Spieth faced 46 competitors in 2015.

Beyond that, whereas the first standard deviation of Woods’s performances was 1.9, it was 2.4 for Spieth. That means Spieth was less predictable than Woods—witness his three missed cuts.

Let’s repeat the Woods 2007 chart, this time substituting Spieth and his 147 significant challengers for 2015. This time the low end of the first standard deviation of the performance of Spieth’s challengers is 0.8 with a group average of 1.000, meaning that all events in which Spieth’s dominance rating was 99.2 or below will be studied. There were eighteen such events, Spieth winning five. Here’s the comparison of his dominance rating in each event to that of his 2 closest competitors, plus all others in the group of 147. (Winners’ dominance rating are in bold face.)

Event Spieth Day Stenson Other

Waste Management

97.8

DNS

DNS

96.3a

AT&T Pro-Am

98.4

98.0

DNS

96.1b

Northern Trust

97.4

DNS

DNS

97.0c

WGC Cadillac

98.6

99.3

96.9

95.2d

Valspar

96.6

DNS

96.9

Valero Texas

95.8

DNS

DNS

94.4e

Shell Houston

96.9

DNS

DNS

96.9f

Masters

94.5

100.5

99.4

RBC Heritage

98.3

DNS

DNS

95.4g

Colonial

97.5

DNS

DNS

97.2h

Memorial

96.7

CUT

DNS

96.0i

U.S. Open

96.0

97.8

99.5

John Deere

96.0

DNS

DNS

British Open

97.0

97.0

100.2

96.7j

Bridgestone

97.2

97.6

96.9

94.8k

PGA

95.2

94.1

100.8

BMW

98.2

94.3l

97.9

Tour Championship

96.3

98.8

97.7

Note: DNS = did not start. a. Brooks Koepka. Spieth was also beaten by Hideki Matsuyama, Bubba Watson, Ryan Palmer, Martin Laird, Jon Rahm, and Graham Delaet. b. Brandt Snedeker. Spieth was also beaten by Nick Watney, Charlie Beljan, Dustin Johnson, and Pat Perez. c. James Hahn. Spieth was also beaten by Paul Casey and Dustin Johnson. d. Dustin Johnson. Spieth was also beaten by J. B. Holmes, Bubba Watson, Adam Scott, Louis Oosthuizen, Bill Haas, Webb Simpson, Kevin Na, Rory McIlroy, Ryan Moore, Ryan Palmer, Danny Willett, Jim Furyk, Rickie Fowler, and Lee Westwood. e. Jimmy Walker. f. J. B. Holmes beat Spieth in a playoff. g. Jim Furyk. Spieth was also beaten by Kevin Kisner, Troy Merritt, Brendan Todd, Matt Kuchar, Sean O’Hair, Branden Grace, Louis Oosthuizen, Bo VanPelt, and Morgan Hoffman. h. Chris Kirk. i. David Lingmerth. Spieth was also beaten by Justin Rose. j. Zach Johnson. Spieth was also beaten by Louis Oosthuizen and Marc Leishman. k. Shane Lowry. Spieth was also beaten by Bubba Watson, Justin Rose, Jim Furyk, Robert Streb, Brooks Koepka, David Lingmerth, and Danny Lee. l. Spieth was also beaten by Daniel Berger, Scott Piercy, J. B. Holmes, Rory McIlroy, Rickie Fowler, Dustin Johnson, Hideki Matsuyama, Cameron Tringale, Kevin Na, and Bubba Watson

Spieth in 2015 simply wasn’t as good as Woods was in 2007. Spieth’s 68.94 stroke average of 2015 was a full stroke higher than Woods’s 67.79 . . . and then some. In theory, that might be accounted for by tougher courses or conditions . . . except it’s not. In 2007 the tour-wide stroke average was 70.94; in 2015 it was 70.95. It also undermines the oft-proposed theory that it’s tough for Spieth—or anybody else—to win today because the competition is simply better and deeper. Nope, Woods was just that much better than everybody else. That’s why he faced only 13 players in 2007 whose scores suggested a plausible chance of competing with him, while Spieth faced 147 such qualified challengers in 2015.

It also validates the idea that with the talent level so close and widespread, performance correlations are highly likely to come and go from week to week.

Finally, it establishes that Spieth and his contemporaries have a long way to go if they aspire to replicate Woods’s brilliance of a decade ago.

Evidence buttressing this conclusion can be readily found in the early records of the two men. From the moment he joined the tour at the start of the 2013 season through 2015, Spieth played in eighty events. Woods’s first eighty events as a pro began on September 1, 1996, in Milwaukee and continued through late March 2000. The two stars-to-be were roughly parallel in age, both in their early twenties. Side by side, here’s how they stacked up:

Woods, 9/1/96 to 3/27/00 Spieth, 2013–2015

Starts

80

80

Rounds

303

288

Wins

20

6

Top 5

40

23

Top 10

50

32

Cut

1

12

Stroke average

68.14

69.84

Field average

71.23

70.90

Dominance rating

95.7

98.5

With respect to dominance and chance, the current situation on the women’s tour roughly parallels that on the men’s tour. In 2015 the most dominant player on the women’s tour was Lydia Ko, with a rating of 97.1. But, like Spieth, she was closely followed: by Inbee Park (97.3), Stacy Lewis (97.8), Shanshan Feng (98.0), Lexi Thompson (98.2), and a host of others. The upper limit of the second standard deviation of Ko’s expected performance was 1.023, a level averaged or exceeded by eighty-nine pros on the LPGA Tour in 2015. Of thirty-one tournaments on the LPGA’s Tour schedule in 2015, twenty-six were won by competitors with a season-long dominance rating below 99.8 and seventeen with a dominance rating below 98.6.

Again as with the PGA Tour, this remarkable competitive balance, while normal, is not necessarily required. As with Woods on the PGA Tour, one needs retreat only to 2007 to find dominance winning out. That year Lorena Ochoa had a dominance rating of 94.9. That means Ochoa was marginally even more dominant on the women’s tour than Woods was on the men’s tour. Her performance that season, which included eight victories—two of them majors—in twenty-five starts, translated to a first standard deviation scoring range between 93.7 and 97.3 percent of the field average, with a second standard deviation range between 91.1 and 98.7. Those numbers combined with her omnipresence—Ochoa played in twenty-four fields in 2007—suggest that more of her seventeen top-five finishes should have been wins.

As with Woods, few competitors actually had a legitimate chance to beat Ochoa unless they played near their absolute peak. She, after all, wasn’t going to help them. (In her twenty-four 2007 starts, Ochoa never posted a four-round total that was higher than 99.2 percent of the field average, roughly the high end of her second standard deviation.) But for the record, here are the sixteen who tended to perform within two standard deviations of Ochoa’s norm, listed in order of their own dominance scores:

These sixteen averaged 98.2 percent of the field average score for the year, making that a good first benchmark. Aside from that lone 99.2 referenced above, Ochoa beat 98 percent of the field average every time out. As for Lorena’s own average score of about 94.9, only Creamer had as much as a one-third prospect of equaling that. The average—and this is for Ochoa’s sixteen closest pursuers—was about 10 percent.

And that was if she just shot her average game. If Lorena dialed her own performance up to the level she reached about one-third of the time, a score that was about 94.2 percent of the field average, plausible contenders began to fall away quickly. While there was better than a one-in-four chance that Ochoa would record a score that was 94 percent of the field average—and in fact she did so eight times in 2007—only Creamer, Pettersen, and Sorenstam could calculate as little as even a 10 percent chance of doing so.

In fact, Ochoa never lost in 2007 when she shot below 94 percent of the field average and lost only twice when she went below 94.8 percent of that target. Both defeats came in playoffs and both to likely suspects. At the Ginn Tribute, Castrale beat Ochoa in a playoff after both players shot 279 against a field average of 296.7. At the Long’s Drug event, Ochoa and Pettersen tied at 277 against a field average of 294.5, Pettersen winning the playoff.

And as with the men’s tour, none of this has anything to do with a deeper talent pool. In 2015 Ko’s 69.44 stroke average was about 3 strokes better than the 72.37 average for all qualifying players. In 2007 Ochoa’s 69.69 stroke average was 3.5 strokes better than the 73.34 average for all qualifiers. Relatively speaking, Ochoa in 2007 was a half-stroke better than Ko in 2015.

One can learn something about the microreasons winners of PGA Tour events succeed from week to week by dissecting the results of individual events. In doing so, it would be helpful if the actual event winners were, one might say, unexpected. After all, if Jordan Spieth wins, the most likely explanation is “Duh . . . he’s Jordan Spieth!” This exploration focuses on four events from the 2015 PGA Tour season, each of them won by a player who ranked outside the top fifteen in stroke average during that season.

McGladrey Classic, October 23–26, 2014, Sea Island Golf Club, Sea Island, Georgia

Winner: Robert Streb, 266

Margin: Playoff over Brendon de Jonge and Will Mackenzie

The McGladrey is one of those off-season events that attracts a field consisting of second-tier pros largely there in order to make a buck. None of the top 15 in the final 2014–15 stroke-average list showed up; Streb, who would finish 18th, was highest ranked. De Jonge, statistically a cowinner, finished 48th; Mackenzie, who did not compete in enough events to develop a full statistical profile, finished 126th.

So the first explanation for Streb’s victory was simply that he was the most talented player in a weak field. Nor do the particulars deliver much in the way of profound elaboration . . . either for Streb or for de Jonge. For the 71 players who completed four rounds, data publicly exist on 62 in four categories of analysis: driving accuracy, driving distance, greens hit in regulation, and putts per round. Here’s what the data show:

Driving accuracy: Relatively speaking, Streb was a laser that week. He hit 73.21 percent of his fairways, exceeding his season-long 59.47 percent performance by 13.74 percent. Still, Streb—and de Jonge, for that matter—ranked in a 13-way tie for only 23rd in driving accuracy with, among others, Justin Leonard, who finished thirteen strokes behind them, and Martin Flores, fourteen back. First in driving accuracy, at 85.71 percent, was Chad Campbell, who finished nine strokes behind the leaders. Combine that with what has already been established about the general insignificance of driving accuracy, and it’s impossible to make the case that Streb’s heightened ability to find fairways led to his victory.

Driving distance: Streb’s drives averaged 297 yards that week, almost precisely matching his 297.1 yard season average. (For the record, de Jonge was a fraction shorter than usual, averaging 282.8 yards off the tee, 2.5 yards below his average.) To the extent distance aided Streb’s win, it had less to do with his performance than with the nature of the field. Although ranking only 36th on tour in driving distance for the season, Streb had the 5th-longest average among those in the McGladrey field, behind only Tony Finau (who finished in a tie for 14th), Jason Kokrak (tied for 66th), Justin Thomas (71st), Daniel Berger (tied for 22nd), and Hudson Swafford (tied for 12th). Since one knows that distance, like accuracy off the tee, is not a significant predicator of success on tour, the most one can say was that Streb’s relative length benefited him against a weak field.

Greens in regulation: Chapter 1 established that the ability to approach the green well was a significant factor on tour. Streb hit 73.61 percent of his greens in regulation that week, a performance that was slightly but not profoundly better than his 69.14 percent season average. Yet it ranked only a very modest 40th among the four-round field. The leader, Stewart Cink at 86.11 percent, finished in a tie for 32nd. (De Jonge, at 83.33 percent, tied for 3rd in GIR that week.) It’s safe to say that Streb’s victory did not turn on his mastery of his fairway irons.

Putts per round: At 28.3 putts per round, Streb tied Fabian Gomez and Nicholas Thompson for number 1 in this category. One could, then, argue that this was the decisive element. To an extent, it’s fair to say that Streb got warm, if not hot, on the greens at Sea Island; his seasonal average of 29.1 putting strokes—102nd tour-wide and 38th among those completing four rounds that week—was eight-tenths of a stroke higher than his field-leading performance that week. (De Jonge, for the record, averaged 30 putts per round that week, about one stroke worse than his 29.03 season-long average.) Obviously, that eight-tenths of a stroke advantage Streb had over his usual self could go a long way toward explaining a playoff win.

2015 Arnold Palmer Invitational, March 19–22, Bay Hill Course, Orlando

Winner: Matt Every, 269

Margin: 1 stroke over Henrik Stenson

Every finished 182nd on the season’s stroke-average table, so his victory at Bay Hill can reasonably be described as one of the least likely on that season’s tour. The Palmer field, while not a major-caliber one, contained names you’re familiar with: Stenson, Hideki Matsuyama, Louis Oosthuizen, Jason Day, Zach Johnson, Rickie Fowler, and Adam Scott. Forty-one of the 64 players in that field who completed four rounds, and who competed enough to develop a full-season data profile, finished ahead of Every in the FedEx Cup rankings, and literally all of them finished with better season-long stroke averages than Every’s 72.42. Yet Every won. How?

Driving accuracy: Every hit 73.21 percent of his fairways, good for a five-way tie for 30th among that week’s full field, yet far better than his season-long 52.51 percent performance. In raw numbers, that translates to a dozen more fairways for the tournament, or three per round. However lightly one considers the ability to hit a fairway, hitting three more per round than usual can’t be a bad thing.

Driving distance: Every coupled his accuracy with a 290.3-yard-average distance, about 3 yards better than his 286.6-yard season average and 31st best that week. But Scott, who tied for 35th, led the field at 308.4 yards, 18.1 yards longer than Every. In fact, 48 players finished 2015 with longer average driving distances than Every, while just 16 were shorter. So if the 3 yards Every picked up that week did anything, they at most kept him afloat. They certainly were not an explanation for victory.

Greens in regulation: If there was a key to Every’s victory at Bay Hill, this plainly was it. He hit 80.56 percent of his greens in regulation, second best in the field (behind Francesco Molinari, 81.94 percent) and far better than his season-long standard. For the year Every would average just 58.1 percent of greens in regulation, ranking dead last among those finishing four rounds at Bay Hill. In raw numbers, his improvement amounted to sixteen greens for the tournament, or four per round. When you go from last to second in the field in the most significant skill, you have given yourself a chance.

Putts per round: Every required 28.3 putts per round at Bay Hill, tying him for 15th with three other players. The leader, Shawn Stefani at 26.5 putts per round, finished in a tie for 21st place, 11 strokes behind Every. Plainly this wasn’t a putter’s tournament. Every’s performance on the greens was basically in keeping with his profile: he averaged 28.67 putts per round for the season.

2015 Crowne Plaza Invitational, May 21–24, Colonial Country Club, Fort Worth, Texas

Winner: Chris Kirk, 268

Margin: 1 stroke over Jordan Spieth, Brandt Snedeker, and Jason Bohn

Sandwiched between the Players and the U.S. Open, the Colonial is one of those Texas tour events that the best players occasionally drop in on but don’t consider an appointment stop. In 2015 the event drew four members of the eventual FedEx Cup top 10—Spieth, Zach Johnson, Danny Lee, and Charley Hoffman—and 3 of the 10 lowest season stroke averages, Spieth, Zach Johnson, and Paul Casey. Kirk, who would rank 93rd in stroke average, would hardly have stood out amid such company. Yet he built a 2-stroke lead on Spieth, number 1 in both FedEx Cup and stroke average, after three rounds and held him off on Sunday to win by 1. How does that happen?

Driving accuracy: Don’t look here. Kirk hit just 50 percent of fairways, nearly 12 percentage points worse than his season-long average, seven fewer than Spieth, and only 59th best among the 69 players qualifying for this study.

Driving distance: Or here. Kirk averaged just 277.1 yards off the tee, about 9 yards less than his 2015 average and just 39th best in the field. For the record, Spieth was also off his average . . . but at 285.5 yards he could afford to be.

Greens in regulation: And certainly do not look here. Kirk hit just 63.89 percent of the greens in regulation that week, the 7th-worst performance in the field and fractionally worse than his standard 64.54 percent for the year. Which leaves only . . .

Putts per round: And here one arrives at Colonial’s bottom line. Kirk required just 25.8 putts per round, the best performance in the field and nearly 3 strokes superior to his 28.6 putt per round season-long average. This was especially fortuitous against Spieth, the tour’s number-1-ranked putter in 2015 at an average of 27.82 putts. At Colonial he needed 28.5, just far enough off his norm to cost Spieth at least a playoff berth. By the way, Kirk ranked 37th on tour in 2015 in putts per round. In essence he got and stayed blisteringly hot on the greens, and that more than offset all of his other flaws. So when announcers say putting is the decisive skill in golf, they’re right . . . occasionally.

2015 Memorial Tournament, June 4–7, Muirfield Village, Dublin, Ohio

Winner: David Lingmerth, 273

Margin: Playoff win over Justin Rose

Featuring the Jack Nicklaus imprimatur, the Memorial is one of those midseason events that occupies the vast space just below the majors. As such it draws a representative field, including star-level talent. Spieth, Justin Rose, Dustin Johnson, and Phil Mickelson all played in 2015. Against such a field, Lingmerth, who would finish 75th in stroke average, would have hardly drawn any betting interest. Yet he held off Spieth by 2 strokes on Sunday and then defeated Rose in a playoff. From time to time, a relative outsider succeeds on tour simply because PGA fields are by their nature highly competitive. The outcome of the 2015 Memorial was decided less by what Lingmerth did remarkably well and more by the stars’ failures to separate themselves from the pack.

Driving accuracy: With a 76.79 percent accuracy rate, Lingmerth easily outperformed his season-long 65.51 percent average. But fairway accuracy isn’t a big deal on tour, and anyway Spieth, Johnson, and Rose all also outperformed their season-long measurements. In fact, the field averaged 70 percent of fairways, nicely above its 62.48 percent average for the season. So Muirfield’s fairways were docile for everybody.

Driving distance: At a 289.5 yard average, Lingmerth was not one of the tour’s longest hitters, and at Muirfield he failed to reach even that standard, averaging just 279.6 yards. Johnson led the field at 326.1, Rose also topped 300, and Spieth averaged 293.5 yards. In short, driving distance was not a determinative factor in the outcome. If it had been, Johnson, who eventually tied for 13th, would have breezed home.

Greens in regulation: Lingmerth hit 68 percent of his greens, modestly better than the 65 percent he usually hit that season. Rose, Spieth, and Johnson were all a couple of percentage points off their usual averages, but in none of those cases did the differences amount to more than a quarter of a green per round. To the extent this is a big deal, it’s because for the season as a whole, Spieth, Rose, and Johnson all averaged between 3 percent and 6 percent more greens in regulation per tournament than Lingmerth. At the Memorial, Lingmerth led all three of those bigger names—just by a few points, but he did lead them.

Putts per round: Lingmerth averaged 26.5 putts per round, fractionally better than Rose and fractionally worse than Spieth or Johnson. Again, the key here is that Lingmerth stayed close, essentially giving fate a chance to tip the scales in his favor. For the season at large, he averaged 0.84 of a stroke more per round than Spieth. In the season-long averages, that 0.84 translates to the difference between being 1st in putting (Spieth) and being 40th (Lingmerth). At the Memorial, Spieth was 2nd in putting behind Patrick Reed, and Johnson was 5th . . . but Lingmerth, with only the 20th-best season-long average in the field, was 7th.

In the four tournaments studied, each was won by a player not generally ranked among the game’s elites. One of those victors, Streb, won by dint of standing out as the best in a mediocre field. A second, Every, won by monotonously hitting fairways and greens. A third, Kirk, won by overcoming a series of shortcomings with his putter. A fourth, Lingmerth, won because none of his competitors did anything to distinguish themselves when he was enjoying a better than average week. Collectively, they suggest that there is no single key to winning in a given week . . . which is why picking winners in a golf tournament is a fool’s errand. This will come as a terrible letdown to the wishes and hopes of Draft Kings types. But unless Tiger is in his prime, the competitive space between players is simply too narrow, and—at least on the PGA Tour—their performance is governed too much by artistry rather than mechanics.

Is it possible, however, to predict future stars? Is the ability to drive a ball long distances, or on an unerring path, or to hit a green in regulation, or to scramble from off the green for par, or to putt well, or any combination of those skills, useful in identifying who is likely to become a star?

On the 2016 PGA Tour were 21 players fitting a pair of criteria that would be useful in addressing the answer to that question. The first condition is that they were active on the Web.com Tour—the PGA’s top minor-league tour—a reasonable number of seasons earlier . . . let’s say five. The second condition is that since graduating to the PGA Tour, they were successful. For purposes of this discussion, “success” is defined as having at least $1 million in post-Web.com earnings. In descending order of post-2011 career earnings, the 21 players who fit both criteria are Russell Knox, Jonas Blixt, David Lingmerth, Danny Lee, James Hahn, Jason Kokrak, Roberto Castro, Brendon Todd, Brian Stuard, Ken Duke, Scott Brown, Billy Hurley III, Erik Compton, Greg Owen, Kyle Reifers, Will Wilcox, Dicky Pride, Tyrone Van Aswegen, Luke List, Steve Wheatcroft, and Tim Wilkinson. Your initial reaction will be that this is not a collection of the tour’s front-rank players. There’s no Jordan Spieth, no Jason Day. It is, however, a solid representation of the tour’s middle class. In combination, the 21 can claim 13 tour wins, 190 top 10s, and about $100 million in career winnings, the poorest among them, Wilkinson, still registering $1.8 million as of the end of 2016.

Let’s take that group of 21 and ask a question: Looking at their 2011 performance on the Web.com Tour, did the group as a whole—or its individual members—stand out in any of the correlative skills measurements?

There were nearly 130 players sufficiently active on the Web.com Tour in 2011 to compile a statistical résumé, so the collection of 21 represents the roughly 15 percent who went on to greatest success. To ascertain whether that success might have been predicted based on 2011 Web.com data, one can begin by examining the ordinal placement of each of the 21 among the full field of nearly 130. If they bunch toward the top in any particular area, then that area will look suspiciously like a predictor.

Here’s how the individuals, and the group as a unit, did in each of the five correlative skills.

Driving accuracy: Russell Knox was the 4th most accurate player on the Web.com Tour in 2011. Knox hit 73.57 percent of his fairways. This is interesting if only because he is Russell Knox, and Knox has already been established as the most successful 2011 graduate of the Web.com Tour. For the record, he has since won two tournaments and more than $10.1 million. Billy Hurley, however, is the only other of the 21 players who ranked higher than 15th in driving accuracy as a Web.com pro; Hurley’s 72.41 percent of fairways was good for 8th place. The 65.36 percent accuracy average of the group was marginally better than the average for all Web.com pros that season, but it translated to only an average ordinal placement of 59.5, barely breaching the upper half. Nothing remarkable there.

Driving distance: At 318.6 yards, the longest driver on the Web.com Tour in 2011 was Jason Kokrak. The 5th longest, at 314 yards, was Luke List, and the 10th longest, at 308.9 yards, was Greg Owen. All 3, by the way, have lost a little muscle in the intervening years. Kokrak was down to 307.4 yards in 2016, List down to 306.9, and Owen down to 297.9. Those distances, it should be noted, were still good enough to rank 6th, 8th, and 39th on the big boys’ tour.

Kokrak, List, and Owen were, however, the exceptions. The average 2011 driving distance of the 21 Web.com successes was 295.29 yards, the average ordinal placement a totally middle-of-the-road 65th. That’s roughly where one would have likely wound up with any randomly selected group of 21 from the full field. At the other end of the scale from Kokrak, List, and Owen, Billy Hurley (273.8 yards) and Tim Wilkinson (282.6) both ranked among the bottom 10. Conclusion: driving distance is showy, but not indicative.

Greens in regulation: The data for GIR is not markedly different from the data for either driving distance or accuracy. Of the 21 eventually successful players, 2—Roberto Castro (4th) and Knox (8th)—ranked among the top 10. One, Brendon Todd (117th), ranked among the bottom 10. The group average of 69.36 percent of greens hit in regulation was slightly but not significantly better than the full-field average. The average ordinal rating was 57.19, about 2.25 places more favorable than driving accuracy but hardly a predictor of success.

Scrambling ability: If there is a predictor of future success, the data suggest this may be it. By converting 67 percent of his missed greens into pars or better, Tim Wilkinson ranked 1st on the Web.com Tour in 2011. Meanwhile, Dicky Pride, Billy Hurley III, Danny Lee, and David Lingmerth ranked 3rd, 4th, 7th, and 9th, respectively. That’s 5 places among that season’s top 10 occupied by current PGA Tour regulars. Among the second 20 were Erik Compton (11th), Knox (14th), and Kokrak (16th). Against those 7 in the top 16, only one (Wilkinson, 116th) ranked among the bottom 30. Tour-wide, the average Web.com player converted 58 percent of his scrambling opportunities in 2011; the select 21 converted 61.14 percent. That may not sound like much, but had the group been a person, they would have ranked among the top 35 on tour that season. Their average ordinal placement, 36.05, is 7 places higher than for any of the other measurements.

Putts per round: On the 2011 Web.com Tour, Blixt (2nd), Wheatcroft (3rd), Todd (5th), and Lingmerth (7th) all ranked among the top 5 for fewest putts per round. As with scrambling, there are few negative counterbalances; only Castro (102nd) ranked among the bottom 30. The average ordinal placement of the 21 was 43.6, noticeably less profound than the average ordinal placement for scrambling ability but also notably more substantive than for any of the other three categories.

All of that digression back to 2011 doesn’t really prove anything beyond a shadow of a doubt about the long-term predictive abilities of the correlative data. But it does at least underscore the view that the surest way of escaping golf’s minor-league system is by polishing that short game . . . something your pro has probably already told you. (Of course he charged you $100 or more for the advice.)