Historical Marginal Win Score calculation process

If you read this blog you know that my latest project is an attempt to attribute wins and losses to players who participated in the NBA in the 1960s. People have asked me how I am calculating Marginal Win Scores for players from the 1960s when I the NBA did not then keep some of the key statistics necessary to the MWS formula, namely individual steals, turnovers, and blocks, and all Opponent Statistics.

Essentially, I do so by using the statistics of today to estimate the missing statistics of yesterday.  Here is my procedure.

First, I “position” each player according to likely defensive assignments. I do this by considering a mix of height and weight. I have found this task rather more difficult for players of the ’60s. The ’60s are filled with players who would be too small to play their designated positions today, or who are very close in height and weight, which makes it difficult to position them. For instance, Jerry West was 6’2” and 175 pounds in his playing days.  West was generally considered a point guard. His backcourt mate in the late 1960s was Gail Goodrich, who stood 6’1” and weighed 170 pounds.  Goodrich was generally considered a shooting guard. But, if the two were matched against, say, the Jones boys from Boston, with SG Sam Jones being 6’4” and 200 lbs and PG KC being 6’1” and 170, I am almost certain West would guard Sam Jones and Goodrich would guard KC. Thus if the two are on the court I am inclined to position West as a SG, but its such a tough call.

Alright, once I have positioned the players, I then estimate their steals, turnovers, and blocks.  I estimate these numbers based upon: historical per minute averages for the position played, physical profile, and a crude “similarity scoring system”.  It works fairly well.  Despite the substantially larger number of possessions per game in the 1960s, I am nevertheless arriving at NBA Team Win Score averages that are consistently in line with the 43.75 average we see today. The key is the turnovers. I believe the turnover numbers in the 1960s were most likely staggering. Throughout the known statistical era, in both the ABA and NBA, turnovers have equaled roughly 13.7% of the following player total: FGAs + .5FTAs + Assists + Turnovers. If we apply this formula to Wilt Chamberlain’s known statistics from the 1965-66 season, for instance, we arrive at the near record turnover total of 426. Such a total would be unthinkably bad today.  In fact, it would surpass the modern record of 422 set by George McGinnis. Nevertheless, I believe that total and others I calculated accurately reflect the likely number of turnovers produced per “at risk” possession in the faster paced game of yesterday.  In order to believe otherwise, you have to believe that the players of the 1960s were somehow able to play at a breakneck speed and still produce a lower percentage of turnovers per possession than the percentage produced by every other subsequent era. There’s no chance that can be true.  In fact, if I were betting I would take the “over” on my turnover estimates, as mindboggling as some of them are to the modern eye.  Go back on Youtube and look at the ballhandling of the 60s. Its rudimentary.

Once I have positioned the players and estimated their missing individual numbers, I calculate each player’s Win Score per 48, and from that data I then calculate positional Win Score averages for the entire NBA.  I then calculate the average Win Score distribution per position. In general, the 60s featured higher Win Scores for centers and small forwards, and lower Win Scores for point guards and shooting guards. This is due to the lack of a three point shot, and the faster pace of the game.

Once all that is done, I calculate each team’s “Team Win Score” average and then, based upon the team’s Pythagorean win total for the season, I estimate what their Opponent Win Score average must have been (Team MWS win totals and Team Pythagorean win totals are strongly correlated). Once I have done that, I then distribute the Opponent Win Score to each position using the positional average posted by the rest of the NBA. I then slightly tier the averages within the given position.

After that is done, I distribute Opponent Win Score to each player and calculate their Wins and Losses in the same manner I do for a player of today.

To gauge accuracy, I used the Historical Marginal Win Score process to calculate wins and losses from several teams of the 2010-11 season. I compared the results I got against the results using known data. The results were surprisingly close. The error factor was about 11%.

Thus I am confident the metric produces accurate pictures of player production in the 1960s.  By doing so, it opens the door on an important era in the game of basketball that frankly I knew little about. That’s why I have undertaken my project.

I will post 1965-66 numbers tomorrow.


One Response to “Historical Marginal Win Score calculation process”

  1. jbrett Says:

    Ty, I’m no expert, but your process seems sensible to me. Regarding the astronomical turnover numbers, I’m thinking of a comment Keith Jackson made–yes, that Keith Jackson–during the broadcast of the 1972 All-Star Game: “Jerry West is averaging just under ten steals per game this season.” Ten? TEN?

    I’m curious where he got the number, obviously, and I’m willing to believe my all-time favorite player was a defensive wizard beyond measure. But I’m mainly thinking that every steal for one player is a turnover for someone else. If that stat–which no one even blinked at during the broadcast, by the way–was at all accurate, then turnovers HAD to be off the charts.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: