Monday, October 15, 2007

What Three Ranking Systems Would Lead You to Expect at the Samsung, Part 2

OK, to pick it up from last post, here are the results from the Samsung World Championship, along with what 5 systems--the Rolex Rankings, the Golfweek/Sagarin Performance Index, and Hound Dog's monthly top 30, and two described below--would lead you to expect.

The first three numbers are from the three major ranking systems, RR as of 10/8/07, GSPI as of 10/8/07, and HD as of 9/10/07. The last two numbers are the player's average ranking among the 3 systems and a points-based formula of my own that combines the RR and the GSPI--assuming "par" to be 72, players get points added to their Rolex average for being "under par" and subtracted for being "over par." Thus, the next-to-last rating gives a sense of how they stack up against their peers, while the last gives a sense of how close to Ochoa they are.

1. Lorena Ochoa 68-67-69-66--270: RR 19.10 [1], GSPI 68.06 [1], HD [1]; #1 [1]/23.04 [1]
2. Mi Hyun Kim 68-70-67-69--274: RR 5.74 [11], GSPI 70.33 [6], HD [5]; #7.33 [5]/7.41 [7]
T3. Jeong Jang 69-68-68-70--275: RR 5.53 [12], GSPI 70.76 [20], HD [13]; #15 [10]/6.77 [9]
T3. Angela Park 67-69-69-70--275: RR 4.12 [25], GSPI 70.36 [8], HD [14]; #15.67 [13]/5.76 [13]
5. Suzann Pettersen 71-69-64-72--276: RR 8.21 [4], GSPI 70.19 [5], HD [3]; #4 [2]/10.02 [2]
T6. Paula Creamer 67-69-71-71--278: RR 6.61 [7], GSPI 69.80 [3], HD [2]; #4 [3]/8.81 [4]
T6. Jee Young Lee 70-70-70-68--278: RR 5.34 [14], GSPI 70.35 [7], HD [9]; #10 [7]/6.99 [8]
8. Stacy Prammanasudh 72-70-70-67--279: RR 5.10 [16], GSPI 70.57 [11], HD [11]; #12.67 [9]/6.53 [11]
9. Angela Stanford 70-66-74-71--281: RR 3.87 [29], GSPI 70.69 [17], HD [15]; #20.33 [15]/5.18 [16]
T10. Se Ri Pak 69-71-70-72--282: RR 6.76 [6], GSPI 70.55 [10], HD [6]; #7.33 [4]/8.21 [5]
T10. Seon Hwa Lee 73-73-66-70--282: RR 4.63 [18], GSPI 70.87 [21], HD [7]; #15.33 [12]/5.76 [12]
T10. Sarah Lee 72-72-69-69--282: RR 3.15 [40], GSPI 70.95 [24], HD [18]; #27.33 [17]/4.20 [18]
13. Morgan Pressel 68-72-72-71--283: RR 6.23 [10], GSPI 70.48 [9], HD [4]; #7.67 [6]/7.75 [6]
14. Cristie Kerr 75-66-70-73--284: RR 8.00 [5], GSPI 70.71 [18], HD [8]; #10.33 [8]/9.29 [3]
15. Maria Hjorth 72-70-71-73--286: RR 3.71 [30], GSPI 71.49 [34], HD [29]; #31 [18]/4.22 [17]
16. Nicole Castrale 73-70-75-72--290: RR 4.26 [24], GSPI 70.63 [12], HD [17]; #17.67 [14]/5.63 [14]
17. Brittany Lincicome 74-70-72-75--291: RR 5.48 [13], GSPI 70.91 [22], HD [10]; #15 [11]/6.57 [10]
18. Ai Miyazato 75-68-76-74--293: RR 5.12 [15], GSPI 71.60 [35], HD [20]; #23.33 [16]/5.52 [15]
19. Michelle Wie 79-79-77-71--306: RR 2.12 [56], GSPI 74.90 [104], HD [NR]; #80 [19]/-.78 [19]
20. Bettina Hauert 76-81-74-76--307: RR 1.10 [118], GSPI 74.76 [227], HD [NR]; #172.5 [20]/-1.66 [20]

So once again, Lorena Ochoa cements her status as the top women's golfer in the world, clinching Player of the Year for the second year in a row. Jeong Jang, Suzann Pettersen, and Angela Park had their chances to hang with Ochoa, but couldn't sustain their good play into the back 9 Sunday, and hence got passed by Mi Hyun Kim, who shot her 4th straight round of 33 or better there.

Kim has to be the biggest surprise of the tournament, coming into it with a back bad enough to get itself an MRI in Korea, no top 10s since the HSBC Women's World Match Play Championship, and a disappointing history at the Samsung. That Jang's wrists held up yet another week is a testament to how tough she is and how great she's been playing over the same stretch. The only reason Park wasn't ranked higher heading into the Samsung is that the Rolex Rankings look back over the past two years of play (which includes a year of struggles on the Futures Tour for her); her finally getting a solid weekend in this year, finishing well ahead of Paula Creamer, with whom she was paired the final 3 rounds, is a sign of even better things to come for the Rookie of the Year. Pettersen has to be disappointed that Ochoa beat her so badly, but she certainly showed why she deserves to be the world #2 this week.

On the down side, Cristie Kerr and Morgan Pressel are the biggest disappointments of the tournament when you go by the rankings, yet anyone who's been following women's golf knows they've both been struggling over the past several months--Pressel since getting outdueled by Se Ri Pak at the Jamie Farr and Kerr since denying Ochoa her first major at the U.S. Women's Open. (That their struggles continued this week make Kim's finish even more impressive.) By this count, too, Brittany Lincicome's poor play should have been a small surprise at best. But if you go by momentum alone, Maria Hjorth has to be the biggest disappointment of the tournament (certainly to me--I picked her to finish 2nd in this week's PakPicker competition over at the Seoul Sisters golf forum!).

OK, so now that we've looked at the finishes from the perspective of the rankings, let's consider which of the ranking systems produced the best predictions. My system is simple--0 for when a system correctly predicts the finish, 1 for being plus or minus 1 spot, 2 for plus or minus 2 spots, and so on. I also include the percentage of each system's predicted top 10s that actually got top 10s this week.

1. GSPI: 53/80%
2. MH (rankings average): 74/80%
3. RR: 77/70%
T4. HD: 85/70%
T4. MH (formula): 85/70%

So the Golfweek/Sagarin Performance Index, which tracks relative finishes over the past calendar year, won handily this week. If you had been predicting finishes by relative position on the LPGA money list, you would have gotten 87/70%, and if by scoring average 57/90%, so while all the ranking systems did better than the former this week, only GSPI did better than the latter (and then, only in the closeness of the picks, not the top 10 rate). In fairness to Hound Dog, it's hard to tell how much better his system would have worked if he had updated his monthly rankings before the Samsung--maybe he'll let us know in comments!

What's surprising to me about these results is not that the GSPI beat the RR--the purpose of the RR is less to predict future finishes than record who's done best over the past two years, particularly in the majors--but that my formula, which uses the GSPI to adjust the RR, did worse than the RR alone.

Now, one test of the ranking systems tells us very little, and given how much time it took me to do this one, I don't plan on doing it every week. Maybe I'll roll it out before the majors and showdown-type events like the HSBC Women's Champions, the Tournament of Champions, the ADT Championship, and of course the Samsung in 2008 (if I can figure out how to adjust it for larger fields). But since the PakPicker involves other people doing the stats for me, I'll try to stick as closely to the GSPI as I can for the rest of the year, keep my mo/gut/past performance adjustments as small as possible, and see how I do. I may even report back some day....

5 comments:

hound dog said...

My current rankings aren't very different than the ones I posted in September. Pettersen is up to #2 and Park is up to #10 (pending updated stats post-Samsung). Might have helped my standing a little in this case. I'm planning to post an update next week.

Like Rolex, my method is concerned with what a player has done. Of course I take my rankings into account when I'm trying to guess what will happen, but the system wasn't built with that in mind.

The Constructivist said...

thanks!

spyder said...

You know TC this whole attachment to schemas thang seems more designed to help with betting on outcomes than it does for teeing up and hitting the ball. And since i am not a betting person, nor have any self-esteem invested in trying to handicap or construct possible outcomes (other than for fun based on not knowing these sorts of things {good thing you folks don't do baseball with all its statistical indices]), i happily stick to watching the action unfold on the courses.

For the ladies themselves i suspect that most of their concern comes right down to the checks that they cash. And it doesn't appear that rankings or ratings matter all that much when you take endorsements into consideration. Natalie Gulbis has her own TV shows, Paula Creamer gets huge amounts for wearing those pink clothes, and perhaps the greatest travesty is Mz Wie who just has to be the duffer she appears to be to merit millions.

Tiger and the pack played four tournaments for a $10 million payout, and Tiger suggested he wouldn't do it again. Easy for him because that amount (paid by Fed-Ex over a 20 year span wtf?) roughly equals his ball endorsement contract for one year. Ochoa wins seven tournaments and only sees that sort of money from her multi-year deals. Sad and more sad given the whopping Samsung payout of $250k??? That wouldn't even cover Tiger's show up fee.

The Constructivist said...

spyder, you raise great issues as usual and I plan on responding to them at length over the off-season. if I can get to it, I'll have something on assessment in academia over at CitizenSE this wednesday that may serve as a preview.

short answer to your post is that people are motivated by status/dignity/respect/recognition as well as money. also, did you see sloucho's non-post on
>postmodern sport
a little while ago? betting is just one manifestation of the phenomenon he talks about with respect to fantasy football. i'm interested in how the 'ranking' impulse ties to larger questions of value/merit. so I wanted to get some experience trying it out as a fan this season in order to be able to analyze my own practices and reactions.

but you're right that the project is fundamentally absurd. what makes golf so fun is how unpredictable it is!

still, given that I've not been able to watch any golf for over a year, much less play, this golf blogging is one way to continue my interest in the game, and maybe infect my girls with it.

hound dog said...

I do try to predict the outcomes of tournaments on my blog and in the PakPicker at Seoulsisters. I have yet to see a parlay card for the Hana Bank - KOLON Championship, so my life savings are still safe.

I started crunching the numbers because I: 1) wanted to know which players were better than others, and 2) didn't see anybody but Rolex and Sagarin already doing it, and I saw problems with their methods.

Spyder mentions baseball - if I had never read Bill James, you wouldn't be hearing anything from me about golf stats. In fact, my player rating system is formatted in a way Bill used to rate position players (multiple categories, point scales, etc). Since golf stat analysis seems to be wide open (baseball, to contrast, is completely saturated - don't need to get involved there), I think TC and I have found our little niche.