Let's review the Hall of Fame Probability calculator from our previous post. The calculator, used by Basketball-Reference.com, attempts to determine the probability of a given NBA player making the Naismith Hall of Fame given the player's statistics and accomplishments. I've always been interested in extending this calculator to the WNBA, and now I've finally been able to do it. pilight provided the MVP shares data and made some suggestions that I've incorporated into the final product.
Of course, the WNBA and the NBA are not the same. Extending WNBA data to an NBA metric required overcoming some problems. As we go through the metric, we'll explain the glitches and the resolutions to those problems.
Another problem - not really a problem, if you think about it - is that Hall of Fame selection for basketball depends upon factors other than one's pro career, regardless if one examines Naismith HOF or Women's Basketball HOF criteria. Contributions in college ball and the international career of a player can be considered whereas the HOF probability calculator ignores each of those. In the end, the number yielded by the HOF probability calculator is the probability that a player would make the (in our case) a hypothetical women's pro basketball HOF solely based on her professional career statistics. In real life, a player could have a probability significantly less than 100 percent but might be assured HOF selection given a strong college career or a coaching career - the HOF considers all contributions made by a player to the sport.
With that in mind, let's look at each of the factors in the linear metric used by Basketball-Reference.com and check them for relevance.
Player statistics: Player statistics are probably the most important part of any metric to determine player greatness - how good could a player be if her greatness doesn't show up in a box score? There must be dozens of metrics out there to evaluate player greatness; which one should be used?
The answer is the "tried and true" metric of points per game/rebounds per game/assists per game, usually represented by numbers separated by slashes - 17.2/10.1/1.8. Any one who has worked with basketball metrics knows that numbers such as points per game can be misleading - point production depends on offensive pace. Your average player will have more points per game playing for the 2009 Phoenix Mercury than for the 2009 Indiana Fever. Statistics must be weighed by circumstances.
Unfortunately, the average person doesn't weigh by circumstances. Bill James said that in the end a player's statistics are all that she has. Take say, Katie Smith. Everyone who has seen her knows that Smith's good. Smith, however, is getting long in the tooth and will probably retire in a couple of years, and all that women's basketball fans will have are memories of great games she played - and of course, her stats.
However, people die. The cohort of fans that saw Katie Smith in a WNBA game will begin to progressively shrink after she retires, and in several decades will dwindle to the point where no one living has actually seen her play. (See: Cobb, Ty.) All we'll have are a few tantalizing film clips, and the bulk of the visual record will remain unviewed. As time passes, Smith's stat line - the only thing that will not change about her - will gain greater and greater prominence.
The problem is that out of the statline, the only stats most people will care about are PPG/RPG/APG, deprived of context. What I admire about the HOF probability calculator is that it weights on the things that most fans consider to be important and not on the things statisticians consider to be important.
Do we have to adjust these stats to make the metric work? Yes. The metric depends on a 48 minute game; the WNBA has always played for only 40 minutes. Stats have to be ballooned to 48 minutes by multiplying by 6/5. There is another problem in that the WNBA has not always used a 24 second clock - this problem is handled in another part of the metric. A final problem is that due to a set of indeterminate factors, assists are not as common in the WNBA as they are in the NBA. That part of the metric, sadly, can't be fixed and those players like Lindsay Whalen who can deal the dimes get shortchanged.
Most Valuable Player Shares: We've already written about MVP Shares. A review of the idea: one looks at Most Valuable Player voting for any given year. The top voter getter in any year - the MVP - is given a full MVP share of "one (1)" for the year, and the runners up are given a fraction of a share based on how close they came to the player who was chosen MVP. A player's MVP shares are added across her career.
This part of the metric determines what the working press thought of the player's career. Part of what makes a player a member of the Hall of Fame is good press. Players that made an impression on a city's beat reporters and HOF balloters get credit in the metric.
All-Star Selections: The metric also positively weights players who for one reason or another were fan favorites. Creators of Hall of Fame metrics - of any sport - disdain All-Star selections, not wanting to turn their respective Halls of Fame into popularity contests. This doesn't make much sense, since in a manner of speaking being a member of the Hall of Fame is exactly like winning a popularity contest - the only difference is in the reasons given for being chosen Prom Queen, so to speak. A better way to put it is that fans of the sport fear that the Hall of Fame will be open to charismatic (but mediocre) players.
Seeing as how the Naismith HOF has few women - players or otherwise - in their HOF and that the Women's Basketball HOF is just getting started, this isn't much of a concern...yet. Filling a Hall of Fame is always difficult. You don't want to be like the Japanese Baseball Hall of Fame where gaining membership is rather difficult, but you don't want to open the floodgates either where fans begin to talk about "weak Hall of Famers". You want the balance to be "just right".
On the other hand, you have to give the fans credit for something. Most of the fans selections are spot on, but there are a few perpetual All-Stars, particularly in women's basketball where the players aren't well known and a visitor might cast an All-Star ballot for the few names he or she recognizes. However, players like that had to be doing something right to begin with to become so well-known.
There are two problems in counting All-Star Game selections. The first is that the WNBA had one year - 2008 - where it held no All-Star Game. Another problem is that the WNBA has only been holding All-Star games since 1999 but the league has been around since 1997.
pilight suggested that players who won medals in the Olympic Games should be given credit for an All-Star selection. I agreed, provided that the medal earned was bracketed by ABL/WNBA years on either side of their Olympic appearance. Lauren Jackson gets some honorary All-Star selections (but nothing earned before 2001). Beijing Becky Hammon gets an honorary All-Star selection for playing with the Russians.
Technically, this method could be extended to the 1998 FIBA games to create some pseudo-All Stars. However, I don't have the data that I need. Besides, the ABL players had All-Star games before 1999; let's give those players an advantage for once.
Height: Height is the only part of the metric with a negative coefficient - the taller you are, the more credit the metric takes away. It isn't because tall players aren't any good. The reason is that if you have player who is 5-foot-6 and a player who is 6-foot-6 and their stats are exactly alike, you assume that the smaller player is better because she did more with less.
The problem is that WNBA are (obviously) not as tall as NBA players. I've thought about the height problem a lot. I found somewhere on line where it stated that female height is distributed the same way as male height - the probability that some female is 20 percent less than a female of average height is exactly the same as the probability that some male is 20 percent less than a male of average height. The only difference in both problems? The average heights.
My conclusion was that the average ABL/WNBA player was 6'1" and the average NBA player was 6'7". I therefore granted each WNBA player an increase of six inches in height so that the NBA metric would work. pilight states that the difference in average heights is closer to seven inches that it is to six but I'm keeping the six inch difference for now.
Last year played: After the 2005 season, the WNBA made two changes to the game - they abandoned the halves system in favor of quarters, and they dumped the 30-second college clock for the 24-second pro clock. The transition from 2005 to 2006 provides a useful benchmark between the two styles of play.
Furthermore, if a player's pro career ended before 2005, there's a good chance that she got cheated out of some of her productive years. If she retired at age 35 in 2005, she was born in 1970 and by the time pro ball started again in the US she was 26 or 27.
Therefore, if a player's career ended in 2005 or before, the player was granted a bonus in the metric. If your last year of play is 2005 or earlier, you get the bonus. If the year is later, you don't get the bonus. It's strictly an either-or proposition. If the league is still around in 2020, I might move the "switch" to give a bonus to any player whose career ended in 2009 or earlier - but for now, we stick with 2005.
Rings: It's the old argument: "yeah, Chamique Holdsclaw is a great player, but how many rings does she have?" The idea is a simple one: great players win championships. They excel in the post-season, they elevate their teams, whatever you want to call it. How can one be a Hall of Famer without the hardware to prove it?
It's simple - players get extra weight in the metric for championship rings, even if they were just the water-carriers on championship teams. (Ask Bill James about all the New York Giants players from the 1920s that ended up in the Hall of Fame.) Out of the ten players who got a perfect score in the metric, only two of those players don't have a ring - and for one of those players, her career ended before the start of the 2006 season. Houston Comets fans will rejoice; New York Liberty fans not so much.
The hard part is determining who got a ring - who was on the roster at the end of the season - and who wasn't. I did the best I could with that.
(* * *)
Now, we take all of these numbers, throw them into a blender, and have them spell "mother". The final number yielded is a number between 0 and 100. This number represents the probability that the player could get in a hypothetical Hall of Fame based solely on their professional career.
Let's walk through the case of Player X. (Maybe you can even guess who Player X is.) Ms. X played 216 games, all in the WNBA so we don't have to worry about ABL rings or MVP shares. Her averages per game were 13.4 points per game, 4.5 rebounds per game and 1.2 assists per game. (I've rounded the exact numbers, and exact numbers are used in the calcuation.) Multiplying each of these by 6/5 gives us 16.1 PPG, 5.4 RPG and 1.4 APG - these are what her numbers would be if she was an NBA player.
What about MVP shares? She got 303 votes for MVP in 1998, but Cynthia Cooper got 426, so that's 303/426 = 0.71 shares. Add another 0.01 share for the three MVP votes she received in 1999. She never got any other votes, so she has a total of 0.72 MVP shares.
All-Star Games? Only one. She was tagged in 1999 but was never called an All-Star again.
Her listed height was 6'3", or 75 inches. We add six inches to that to make her an NBA player, giving her 81 on our inches metric.
Was her last season before 2005? Her last season was in 2003, so she gets the "2005 and before bonus" which is equal to one (1).
Any championship rings? Not a one. She gets a "zero" in the rings department.
Now let's add everything:
The sum is used as an exponent in the formula e^x/(1+e^x). We calculate e^0.203/(1+e^0.203) and get 0.551, which is a probability of 55 percent. In other words, given this player's professional career, the chances that that player would end up in the HOF based solely on her pro years is 55 percent.
There are ten players on the list of perfect scores - each player got 100 percent in the metric, rounding up. One of these players is already in the Women's Basketball Hall of Fame. Barring insanity in Knoxville, Tennessee, the other nine soon will be.
In alphabetical order, those players are:
Tamika Catchings: She's the only player in the Perfect Ten who doesn't have a ring and is still playing. She got very close in 2009, though. Clearly one of the great-uns. A freshman on the undefeated 1998 Tennessee NCAA Championship team. Also a member of the All-Decade Team.
Cynthia Cooper: Coooooop. Four rings. What's so amazing about Cooper is that she lost a great hunk of her pro career and ended up with four rings in spite of it all. She's the only member of the Perfect Ten that's already in the Hall of Fame. All-Decade Team member.
Yolanda Griffith: Those Sacramento Monarchs fans out there will Stand Up and Testify. She didn't have a college career at a Big Name University, but Griffith won an MVP award in WNBA and two Olympic gold medals Someday, Yo will be in Knoxville. Yo is also an All-Decade Team member.
Lauren Jackson: The Aussie. Every year, fans in Seattle hold their breath in hopes that LJ will return. Even if she moves to the Outback and never touches American soil again - she's done enough. She's won a lot of silver for the Australian team and she is also an All-Decade Team member.
Lisa Leslie: If she were Leroy Leslie, she would have averaged a double-double in the NBA. Some say she's a primadonna, but she sure belted out beautiful arias in the WNBA. A good argument could be made that Leslie is the Best Player Ever. All-Decade team member.
Katie Smith: I think of her as the Lou Gehrig of the WNBA. The Iron Woman. Rings in two different pro leagues. The most games played out of the Perfect Ten. All-Decade team member.
Sheryl Swoopes: If you say Coop, you must say Swoop(es). Game Texas Tech an NCAA championship in 1993. First women's basketball player to have a shoe named after her. She kept playing for as long as she could, and she still wants to play. We might not have seen the last of Swoopes. All-Decade team member.
Diana Taurasi: For those who argue that Leslie is the best ever, there are voices from Phoenix saying "Dee ain't finished yet." The youngest player on the list. Led Connecticut to three straight NCAA championships. When she steps on the floor, fans say "that's the best player here" even if they know nothing about women's basketball - or, for that matter, basketball. All-Decade team honorable mention.
Tina Thompson: When the sun goes cold and the earth is a lifeless rock floating freely in space, Katie Smith and Tina Thompson will still be playing one-on-one with the last basketball in existence. Second to Lisa Leslie in all-time points scored. All-Decade team member.
Natalie Williams: I never knew how good Natalie Williams was. It sounds like I missed a lot. First woman to be named an All-American in basketball and volleyball. Great internatioal player. All around great athlete. ABL superstar.
All of the above received 100 percent on the Hall of Fame probability calculator. All of these women could get into a Hall of Fame based on their American pro career alone. But with names like this, you don't need a calculator to tell you they're great.
Coming next: those players who fell short of 100 percent. Like, for example, that nameless player who got 99 percent. But she made it in anyway.