PGA's new putting stat: Putts Gained Per Round

[chriskzoo]

Not sure if you'll be able to see this if you're not a WSJ subscriber, but interesting new stat the PGA is going to start using:

http://online.wsj.com/article/SB1000...142473884.html

There is a lot of number crunching involved, but essentially what it does is take every putt up to 100 feet (down to the inch) and create a "par" for that putt based on the average tour player (based on everything from the difficulty of the green, strength of the field playing that week, etc). So if a player is faced with a 10 foot putt that has a calculated valuce of 1.82 (e.g. the average player in the field taking that putt needs 1.82 strokesd to hole it) and a player gets down in 1, they will have "gained" 0.82 putts for that hole. Get down in 2, and they will have lost 0.18 putts.

Using this new metric, in 2009 the best putters were:

1. Luke Donald
2. Tiger Woods
3. Ben Curtis
4. Bryce Molder
5. Brad Faxon
6. Jim Furyk

Surprisingly the "#1" putter last year, Steve Stricker, ranked 69th (mainly due to playing on a lot of "easy" greens" and Phil Mickelson ranked a lowly 149th.

They are also creating a similar stat for "off green strokes saved per round" (think "up and down" I guess) and using 2009 data Tiger was a full stroke better than anyone else on Tour - which makes sense with his ridiculous short game and putting combination.

In any event, despite being impossible for the lay golfer to calculate, it seems a much more accurate way of determining who the "best putter" really is, as other stats like putts per green in regulation are affected by the length of your first putt (owing to your approach shot) and putts per round, which can be depressed by missing a lot of greens and having a good chipping game.

[B-Con]

This is good. Measuring average putts per green heavily favors those who don't get GIR, measuring average putts per GIR rewards those with good approach accuracy. Hopefully this

From the article, it looks like -- to generalize -- they rated greens in a similar way they rated courses for the handicap system.

But I'm disappointed that it looks like it's impossible for an individual to keep track of for them-self, so this won't help any of our own stat keeping.

[edit]
The article mentions that a ton of stats that were made available. Anyone know where to find the detailed putting stats?

[adig]

very interesting. i would have never thought of something like this. not surprised to see my guy sergio at 133rd....

[iacas]

Agreed, this is great - it's something the PGA Tour can do that we, unfortunately, cannot within an application like Scorecard.

Will be interesting to see how this state shakes out.

[chriskzoo]

But I'm disappointed that it looks like it's impossible for an individual to keep track of for them-self, so this won't help any of our own stat keeping.

It would be impossible and impractical to have an individual track this sort of stat. Sure, they could measure the difficulty of the greens easily enough, but this has so many other variables that are field specific that there's not way you could apply that to an amateur. Part of it is even measuring putts down to the inch - how would an amateur even attempt that without affecting play?

[B-Con]

From my above post, I meant to say, "Hopefully this will be the statistic that allows putting ability to be compared between individuals." Not sure how I deleted the last part.

It would be impossible and impractical to have an individual track this sort of stat. Sure, they could measure the difficulty of the greens easily enough, but this has so many other variables that are field specific that there's not way you could apply that to an amateur. Part of it is even measuring putts down to the inch - how would an amateur even attempt that without affecting play?

Right. So it's not going to help any of us. We're considered detailed if we even estimate the distance of our first putt.

[Diddy]

I could see the layperson using some sort of over-generalized method of keeping this statistic that doesn't take into account the difficulty of the course/green. I obviously haven't given this much thought, but wouldn't something like this work:

Putts between 2' and 5', average strokes = 1.5 (i.e. you make 50% of them and never three-putt)
Putts between 5' and 10', average strokes = 1.75
Putts between 10' and 25', average strokes = 2.0
Putts between 26' and 50', average strokes = 2.25
Putts over 50', average strokes = 2.5

This is obviously rudimentary, but I think it gives a general enough gauge of what the normal hack does on the putting surface. I'm sure there's a study out there that gives more precise average values for "Average Joe" golfers, or maybe you can just use pro stats and judge yourself against the best. Anyway, this whole PGPR stat seems really interesting, and with some tinkering I might try it out this year and see what numbers I come up with.

[iacas]

I could see the layperson using some sort of over-generalized method of keeping this statistic that doesn't take into account the difficulty of the course/green. I obviously haven't given this much thought, but wouldn't something like this work:

I like that. And Scorecard does keep track of those stats (you can click a radio button to see the actual values too):

None

[Shanks A Million]

Agreed, this is great - it's something the PGA Tour can do that we, unfortunately, cannot within an application like Scorecard.

I think you could write the algorithm easy, but who can compile all that data? Think about it, it would make Nasa's data stream from it's missions seem minuscule. My father designed the computers that do that for the Space Shuttle (preflight to T-31), and they are not all that powerful, they lasted a long time (some may still be in use). They were basic realtime computers that collect data from thousands of sensors and gauges, and turn it into display, sound alarms, etc.

The concept is the same. Take a whole bunch of measurements, then design an algorithm to crunch it all into some meaningful number. They already have the info, so it was merely a matter of writing the algorithm, right?

[B-Con]

I think you could write the algorithm easy, but who can compile all that data? Think about it, it would make Nasa's data stream from it's missions seem minuscule. My father designed the computers that do that for the Space Shuttle (preflight to T-31), and they are not all that powerful, they lasted a long time (some may still be in use). They were basic realtime computers that collect data from thousands of sensors and gauges, and turn it into display, sound alarms, etc.

Have the released the algorithm publically? Reverse engineering it would not be fun.

Coming from my question above, is the compiled data publicly availible? If it and the algorithm are, implementation would not be overly complex. But it still wouldn't include individual ratings for the greens the individual plays, which, as I understand the article, is like playing golf on a course without a course or slope rating and trying to count your score toward your handicap. Without individual green ratings I think we can only get about half of the benefit from such a method. Which still isn't that bad, Diddy above made a decent suggestion, but it won't compare to the new official PGA-sanctioned statistic.

[chriskzoo]

Have the released the algorithm publically? Reverse engineering it would not be fun.

It's more complex than even just knowing the green ratings because they take into account the strength of the field playing as well (and probably other factors). So a 10 foot putt at Pebble Beach during the AT&T; Pro Am might have an average value of 1.80, but during the US Open the same exact putt might be a 1.77 due to the stronger field (who, in turn, are probably assumed to be better putters).

It's not really a useful stat for any golfer to utilize - it's only meant as a means of comparing competing golfers to each other.

[Shanks A Million]

It's more complex than even just knowing the green ratings because they take into account the strength of the field playing as well (and probably other factors). So a 10 foot putt at Pebble Beach during the AT&T; Pro Am might have an average value of 1.80, but during the US Open the same exact putt might be a 1.77 due to the stronger field (who, in turn, are probably assumed to be better putters).

Even the most complex calculation could be built into an algorithm. The rankings of players in the field by OWGR ranking points could be used, or the ranking itself could be taken as a reciprocal average, and divided by your own ranking or something... So, Tiger and Stricker would be (1/2 + 1/1) / 2 = 0.75 And Phil's ranking would be 3, so 1/3 x .75 = +.25

So Phil's putt counts as X - .25 If, however, you had #100 and 150 in the world with Tiger, it would be: (1/100 + 1/150 + 1/1) / 3 = 0.3358 Now Phil's putt is a X - .11 If he was expected to get down in 1 on each, he'd get "charged" .75 a stroke with the first field, and .89 of a stroke with the second. So, it's complex, but a computer can do even mind boggling equations in microseconds. Some dude somewhere could probably write this whole thing out and make it work. Rest assured, it would take about a thousand lines of equation.

[B-Con]

Using a method similar to your numbers, I might come up with this:

[zeg]

B-Con, I don't think that's the correct way to count it. From the description in the article, if you have a 50-foot putt with a value of 2.7 and 3-putt, you would lose 0.3 putts. In their example, a 1-putt on a green worth 1.82 gains 0.82 putts while a 2-putt loses 0.18, with no concern for the rating of the second putt. This makes more sense to me, since in your approach you could get some wacky results. For example, if you had a 30 foot putt worth 2.0 putts and after a few too many beers, you putted in the wrong direction, leaving yourself another 30 foot putt also worth 2.0. Miraculously, you sink it. By your method, you get 0 for the first stroke (since you 2-putted in the end) and then gain a stroke on the second since you one-putted, yielding 1 stroke gained. That's silly, since you 2-putted. So you only need to know the # of putts from a particular distance, adjusted for the particular green (and the field).

So to be exact, you do need more data than odds of making a putt from a given distance---you need the distribution of number of putts so you can calculate the average. But as has been pointed out, outside of serious tournaments, it's unlikely you can meaningfully calculate exactly the statistic their using, but an approximation that captures its intent could be extremely useful. If they even publish the putts-to-go table (i.e., the scores before adjustment for particular green and field), that'd be an enormous benefit. For most greens, I'd bet the adjustments are relatively minor, so you could just run with it, perhaps binned to a more reasonable size for the average golfer to measure (since most of us don't measure every putt to the inch).

[chriskzoo]

Zeg - I think you're right, once the ball lands on the green, the "average putts" is determined one time for that green from that spot. So if I have a putt that's a 1.82 and putt to 3 feet, a putt which is valued at 1.05, if I 2 putt, I end up +0.18 (2-1.82) and not +0.13 ((2-1.82) + (1-1.05)).

I could be wrong, though - I can see calculating it either way. In fact, doing it the latter way is probably more accurate, since it measures the value of EVERY putt.

[iacas]

B-Con, I don't think that's the correct way to count it. From the description in the article, if you have a 50-foot putt with a value of 2.7 and 3-putt, you would lose 0.3 putts. In their example, a 1-putt on a green worth 1.82 gains 0.82 putts while a 2-putt loses 0.18, with no concern for the rating of the second putt. This makes more sense to me, since in your approach you could get some wacky results.

I agree, you only count the first putt. Which, conveniently, is all we care about in Scorecard (but that's the other thread).

It's "total putts from here" where "here" is the first time you're on the green. Adding to the wacky data like the example zeg gave, if you miss a three-footer that has a rating of 1.02 but make the three-footer coming back, that's 0.98 strokes lost, not 0.02 * 2.

[ApocG10]

Its definately a better system for the PGA, where every putt is measured to the inch, and every shot is logged.

[B-Con]

You guys are right. I was confused by two different values mentioned in the original paper (of which I've now read a solid half) and the article.

First, the article itself is a bit confusing:

With this baseline established, the model compares the results of each putt a pro takes to the expected putts-to-go average on that particular green by a hypothetical average field.

The phrase "each putt a pro takes" seems very close. However, the next sentence says:

If a player holes a 15- footer whose value is 1.82, he gains .82 strokes on the field. If he needs two putts, he loses .18 strokes on the field

It says nothing about the second putt's value and still talks about the final score for the field. However, it doesn't say that there is nothing else contributing from that hole to the final score, it just says that the initial putt contributes to the final score.

So, I went back to the source. The paper both lays out the process for choosing the model and proposes the "putts-gained-per-round" statistic. PGA just adopted what the paper proposed. The paper discusses more than just putting, it discusses stroke measurement on the whole. Putting is just a specific instance of their method. They have what is called "stroke-value", which is the "strokes-gained" for a given shot. If you add all those together, you get the total value for the hole In short, I thought you calculated the actual performance from each shot and averaged them, in reality you calculate the calculated values from each shot and average that. When you sum up all values from each putt it is the same as just comparing the number of putts to the starting location on the green, because the calculated value of a putt is decided by the "strokes-to-go" from the starting and ending position. When you add and subtract all these, all but the first strokes-to-go value and total number of strokes get canceled out. So you are summing up all the values of the shots not all the actual performances from the shots. That's what had confused me. Carry on, I was just confused. All you need to know is how far the first putt was and how many putts you took.