Tuesday, March 16, 2010

MIT vs. Hound Dog

Looks like Hound Dog, tatkins, and I have a lot more competition when it comes to golf stats: there's a team at MIT using advanced mathematics and supercomputers to develop a new way of isolating putting skill from other factors. Their measure is called "putts gained per round," and it sounds pretty impressive to me. By comparing the length and kind of each putt taken to tour averages for similar putts, and totalling up the gains and losses over the course of a round, this stat improves on total putts, putts per green in regulation, and total length of putts holed, the three standard stats on the PGA Tour. Apparently, ShotLink is trying to incorporate this measure into their record-keeping, the tour is planning to roll out many more stats from ShotLink, and the MIT team is already trying to develop other performance stats.

So unless Hound Dog has a supercomputer hidden in his basement, it's looking like the only thing that's going to keep us LPGA stats guys in business is the poverty of the tour. Given that they can't afford to keep any performance stats in Asian events, they're a long way from collecting the kind of data that would allow them to generate tour averages for different kinds of putts. I'm guessing that they'd need measures for the speed, slope, break, and length of every putt taken by every player on tour for at least a full season before they could roll out a similar stat....


Hound Dog said...

MIT vs Hound Dog? You should definitely take MIT. But like you say, if the LPGA doesn't (or can't afford to) make the necessary data available, it doesn't matter whether you have a supercomputer in your basement or an abacus.

Their findings with the PGA data, however, might give us clues about how/where to look in our restricted dataset. We should stay tuned.

Mike said...

You know, guys, there's always the possibility that this will be a totally useless stat. According to what I read about the stat over at Devil Ball, it sounds like it could be misleading. For example, after saying that it tells us how many strokes a player "gains" on the field by putting, Jay Busbee writes:

"The winner? Luke Donald, who could count on gaining an extra 0.905 strokes on the field by his putting skills in 2009. Steve Stricker, who ranked No. 1 in putting average, ranks No. 69 in putts gained, mainly because he's exceptionally good with his approach shots."

So it sounds like this is a measure of "recovery" putting -- that is, how you putt after you've missed the green. (Otherwise Stricker wouldn't be penalized for hitting good approach shots.) But doesn't this mean that "putts gained" is actually a measure of how well you chip and pitch? And in Donald's case, if he had to hit a chip first, then didn't he lose more strokes than he "gained" on the field? He certainly didn't gain any strokes on Stricker!

This just seems to be a way of isolating a single aspect of the "scrambling" stat, and I'm not sure it really tells us anymore than we knew before.

The Constructivist said...

Jay's summary is decent, but a little off. They compare the length and other factors of every putt you take, so however you got to the green doesn't matter. Correct me if I'm wrong, HD!

Mike said...

TC, I went back and looked at the Golf Digest article you referenced and the WSJ article that it referenced... and I'm still not convinced about this new stat. I have two questions about it:

1) The WSJ article says MIT tried to figure out why Stricker was penalized so much, and they determined that "he happened to play on the 'easiest' greens of all 166 players sampled." So they're saying Donald played tougher courses than Stricker last year? Stricker played more events than Donald, and both played all the majors, all the WGCs, and all the FedEx playoff events, so that doesn't sound right to me. This most likely means that Stricker is being penalized for playing more tournaments, especially if those tournaments aren't the big ones -- not a message the Tour wants to send to the big names!

2) The WSJ article also says the baseline figures are adjusted for 2 factors -- relative difficulty of the greens and relative strength of the field. Soooo... it makes no difference whether the putt is for eagle, birdie, par, bogie, or "dreaded other"? Doesn't the meaning of the putt have some bearing on how well a player putts? And I know you can't really measure something like that... but it does seem that a birdie putt "gains more strokes against the field" than a par putt. This sounds suspiciously close to agreement with Busbee's comment; Steve obviously had a lot more birdie putts than Donald, simply because his GIR was so much better. Stricker is #2 in the world because he gains more putts against the field than anybody except #1.

And "relative difficulty of the greens" seems to ignore the fact that some putts are more difficult than others on the same green. If we're going to rate relative difficulty of the greens, we should also include relative difficulty of the individual putt -- information that should be in that mountain of ShotLink data.

I'm all for more stats that tell us how player performances under similar circumstances compare, but I'm not convinced this stat will do that. I'd be far more interested to know who makes more birdie putts, or who makes more par-saving putts... and what distances they make them from.

The Constructivist said...

Good questions, Mike. My sense is that they are evaluating each putt on its own merits (and place on the green), but I agree that taking into account whether it's an eagle putt, birdie putt, par saver, or double avoider makes a big difference!

Why don't you contact the authors of the MIT piece everyone's talking about and get answers for yourself? I'm a little caught up in the politics of funding higher ed right now to pursue this myself.

Mike said...

TC, I found the URL for the paper and downloaded it... but it's 50 pages and I'm not so sure I'll even understand it! But here's where you can get it:


There's a link labeled "One-Click Download" at the top of the page. It'll give you the 50-page PDF.