Monday, July 28, 2008

Player Rating System

I noticed the other day that my original post on this subject was written in September 2006. My rating system has been tweaked a time or two since then and I’ve gotten much more comfortable with its results, so I felt like an update was necessary – especially for you folks who haven’t been reading my stuff that long.

The system takes five statistical categories that are
all available from – victories, money earned, scoring average, Top 10 percentage and missed cuts – and combines them into one number which can go as high as 100 or as low as zero. The point scales for each category slide down to the level of about 50 players, so I believe it delivers accurate rankings at least to position #40. I have an advanced ranking system which attempts to rate players farther down the scale (I’ll unveil it someday – promise!) but for my Top 30, the simpler version works to my satisfaction. I confine the method to the numbers from the current calendar year. My early-season rankings combine the player’s score from the previous season with their current score in various degrees, and by the middle of the season the prior year’s influence gets weeded out.

23 points are allocated to money list position, 22 points to scoring average, 15 for Top 10% and 12 to missed cuts. Each victory is worth four points (majors count for eight) up to a maximum of 28 points. Each missed cut costs a player two points – miss six and you get zero for that category. Zero points is also the score if they fall out of the Top 40 on the money list, average 72.7 or higher, or fail to finish in the Top 10 12% of the time. In 2007, Lorena Ochoa scored 98. She maxed out three of the categories and just missed on the scoring average and Top 10% scales. Suzann Pettersen scored 80 – five wins (one a major) gave her 24 victory points, plus 22 for second on the money list, 17 for her scoring average of 70.86, 11 for a Top 10% of .458, and six points because she missed three cuts.

The system is geared toward evaluating a season of 20-30 starts. As a player plays fewer events, it becomes mathematically easier to avoid missing multiple cuts and maintain a high scoring average and Top 10% but their overall total suffers by having less opportunity to win tournaments and earn money. Players with an abnormally low number of starts and excellent results (Michelle Wie 2006, Stacy Lewis 2008) can skew the rankings somewhat. I prefer to live with that known flaw and subjectively adjust my posted Top 30.

I’ve used this method to look at
previous LPGA seasons and retroactively rank those players. The scoring average scale needed adjusting more and more the further back I went – if I hadn’t done that, prior to 1968 only Mickey Wright would have ever gotten even a point or two for her scoring ability. For the most part, my system has agreed with the official Player of the Year. Notable exceptions are 1994 (Laura Davies should have won over Beth Daniel) and 1993 (Brandie Burton over Betsy King). My Rookie of the Year selections tend to disagree a little more often, mainly because the Rolex rookie point system favors a player who gets in more events. Grace Park failed to win the 2000 ROY (Dorothy Delasin did) for that very reason.

When I first developed this method, I remember thinking “that came together so easily, there must be something wrong with it”. Other than its intentional design limitation of only reaching to 40-50 players and the low-number-of-starts issue I mentioned, it has served me well enough that I haven’t tweaked it at all in the last 18 months. It seems to have the proper balance of rewarding good, consistent play and penalizing the opposite. When a player goes on a Top 20 streak of four or five weeks, they trend upwards. When one misses two or three straight cuts, they drop like a stone. Players who have won tournaments tend to be inside the Top 20, unless they have also missed a bunch of cuts and haven’t done much else to support a lofty ranking (ala Louise Friberg and Leta Lindley in 2008).

There is one other limitation to my system. It only uses data compiled in LPGA events. Ji-Yai Shin is probably among the ten best players in the world but I don’t believe there is a reliable way to compare players from different tours given the small amount of interconnected data. My system currently ranks Shin at #24 based on her LPGA numbers (I moved her up two positions in my Top 30 because of her KLPGA record). It is possible to analyze how, for example, Sophie Gustafson, Suzann Pettersen and Gwladys Nocera play here and on the LET and how Shin plays here and on the KLPGA and use those results to make generalizations about the abilities of the other players on those tours. But those rounds of cross-pollination are a drop in the bucket. The LPGA season consists of 150+ players racking up over 10,000 rounds a year and the rounds played on the LET, KLPGA, JLPGA and others number at least 30,000 more. Any method that takes a slice of data smaller than 1% of the total and uses that tiny slice to make major adjustments to the significance of the other 99…well, that’s just bad math.


svenson said...

Thanks for the overview of your system. I wonder if a simple statistic, such as average final round (Sunday) score would hold much meaning other than to generate discussion. I've always wondered if there is a set of players that score better than others on Sunday.

PS on the LPGA TV coverage commentary: Notice how every player discussed on TV happens to be a good "ball striker"?

The Constructivist said...

Hey, HD, we both must have ratings on the brain. I just finished my JLPGA ranking--sadly lacking due to yours going unupdated for awhile--and am starting work on the LPGA's Junior Mints!

Thanks for the new insights into how your system works!

The Constructivist said...

Ah, here it is!

The Constructivist said...

Hey, HD, you catch Rebecca Hudson's complaints about the Rolex Rankings slighting LET players? Any thoughts on that? THere are way more JLPGA players in the RR top 100 than LET. Are the Japanese (and Korean) players on that tour that much better?

Unknown said...