Economist Golf Oracle Dan Rosenheck’s Interview

If you’re a golfer, one of the main reasons you chose this sport and enjoy watching it resides in the fact it is unpredictable! Well, you might be wrong…

Following Harvard studies, and a long sports journalist career at The Economist mainly, Adam Rosenheck, today data’s editor created the Economist Advantage Golf Likelihood Estimator, a major tournament outcome tool that succeeded in predicting Sergio Garcia’s 2017 Masters Win.


1-Considering your resume and The Economist forecasting editorial strategy, when you became data editor, we could probably have predicted you would create such tool, but what we could not is why on golf and not on Baseball for example?

Well, I already do have a baseball projection system of my own, which has done pretty well in online forecasting competitions.  And there are lots of other outstanding baseball prediction systems out there.  I was very surprised to find out that there wasn’t already a well-known and highly developed publicly available live projection system for golf (Ken Pomeroy‘s is the closest, but he’s the first to admit it’s pretty bare-bones, and it only lives on Twitter).  Also, golf is just a dream sport to model: there are no interactions between players, essentially only one thing to count (strokes), and enormous datasets available online.  That means you can be pretty confident that a good statistical system should be able to capture a lot of what actually goes on.

2-We know that other tools exists, but they are mainly used for profit purposes  (I think about bookmakers). What is your leitmotiv for creating such tool? 

No different than anything else The Economist does–we produce compelling content, which we monetize via subscriptions and advertising.  There are millions of golf fans out there who would love to know exactly where a tournament stands, and they’ll need access to The Economist to find out.

3-We hear the commentators use more and more data to tell X player’s chances of win for a hole, or a tournament, not always successfully. How yours is better?

As I said before, the only other win-probability system I know of is Ken Pomeroy’s, and it doesn’t incorporate nearly as many factors as ours does.  I believe the 15th Club has one that might well be as good or better, but I think it’s proprietary.  It would be interesting to see rival models sprout up to see which is the most accurate–but hopefully ours will prove to be so good that no one will feel the need! 🙂

4-Your model could not predict Jordan Spieth 2016 Masters collapse. But my question is, how many times such event happens?

Depends on how you’re defining the collapse.  At his biggest lead, heading into the back nine, he had about a 95% chance to win, so it was 1-in-20 that he would manage to lose.  But the *way* he collapsed–the quadruple bogey on the 12th hole–was much less probable.  EAGLE gave him a one-in-1,350 chance of hitting a quad or worse on that short par 3.

5-Do you think the players could use your tool one day to determine the tournaments they’ll play, or for a better cause, to improve their game? 

Sure, I think EAGLE or something like it could be a handy tool to determine which courses and fields of competitors would offer players the best odds of victory or a strong showing, and the corresponding accolades, rankings gains and earnings.

6-I remember when for the first time I had a class with a pro on the course, he would just tell me the strategy I needed to put in place for each hole because he had those insights. Would you see your tool formula applied to amateurs to better enjoy/play the game and improve the pace of play?

I’m not really sure EAGLE would serve much of a purpose for amateurs.  It doesn’t tell you anything about how to play–it just predicts how professionals will perform.

7-What EAGLE 3.0 will look like?

The next step would be to incorporate more granular data.  Right now, EAGLE operates exclusively with stroke totals from each hole.  But there’s a wealth of information available by going one level deeper, and examining shot-level rather than hole-level data–driving distances, driving accuracy, greens-in-regulation percentage, putting statistics, strokes gained and lost, number/location/type of hazards, etc.  And working in weather data would be terrific too.  It’d be fascinating to see whether bringing all of that into the model yields a quantum leap forward in terms of prediction accuracy, or merely a marginal improvement.

8-Do you think your tool can take of the magic of the game? If not, what are the remaining factors that make it so unpredictable?

I’m afraid « the magic of the game » is outside of EAGLE’s purview.  Like every other part of life, golf will always offer surprises and we’ll never be able to predict it perfectly.  But the better EAGLE gets, the more we will understand about how the sport works, and why some players tend to win more than others.

Wanna know more about this tool, check this video of Dan presenting his tool at MIT.

Dan Rosenheck

Dan Rosenheck pictured above presenting Eagle at MIT in 2016.

Votre commentaire

Entrez vos coordonnées ci-dessous ou cliquez sur une icône pour vous connecter:


Vous commentez à l’aide de votre compte Déconnexion /  Changer )

Photo Facebook

Vous commentez à l’aide de votre compte Facebook. Déconnexion /  Changer )

Connexion à %s