Value Add: Calculating Each Player's Impact on Point Spread
The two most common questions I receive when updating ratings for college basketball players regard: 1) Sports Information Directors who want to explain the value of their star player and 2) people who want to understand how much the absence of a key players should impact the point spread of an upcoming game. While ESPN and CNN/SI have done a great job of explaining Value Add Basketball, the basic explanation is that a player’s Value Add Rating equals how much the point spread would change if he misses a game.
As an example, let’s take the game last season pitting eventual national champion Louisville against the worst team they played the entire season, the Kangaroos of Missouri-Kansas City. When you click on Value Add basketball for 2006-2014 seasons and enter “2013 Act” and Louisville and then Kansas City in the search boxes, you can add all of the Value Adds for each player on each team broken down by Offensive Value Add, Defensive Value Add and PG/Per you get the following totals to plug into the Value Add Predictor table:
||Ave. 2013 Team
|Base (pick same number to use for both)
|Teams Offense (VA never gives negatives)
|MINUS Opponents PG/Per (-2.5 for full team)
|Predicted Final (not adjusted for tempo)
You start with a base number for both teams which just has to be the same. You might want to put in 45 for a slow match-up between grind it out teams like when Miami plays Clemson, or 70 for a game between fast-paced teams in an Iowa vs. Indiana match-up, but it gives a base score for if both teams were comprised of equally solid High-Major bench players.
If you add all of the Offensive Value Add figures for Louisville players last season you can estimate that they combine to score about 21.69 points per game MORE than good replacements, so that is the next figure (60+21.69=81.69). UMKC’s players combine to let opposing teams score about 10.03 points MORE than a set of solid replacement defenders would allow to up Louisville’s predicted score to 91.72. Finally the adjustment for perimeter defense and point guard play will always be -2.5 if all players are healthy. Add them up and Louisville is predicted to beat UMKC by a score of roughly 89-41.
The actual score when they played was a little higher scoring 99-47.
HOW MANY POINTS IS EACH PLAYER WORTH? The impact of each player has proven to be more than I originally expected when introducing Value Add, indicating I was actually cutting the true value in half while explaining the system. With a couple of years of results it appears that a point of Value Add is worth about 1 point given average tempo. Typically an All-American player would have a Value Add of at least 9.0, which means his team would lose about nine points on their margin of victory/defeat if he had to miss a game.
However, a great player lost from a low major team will be hurt much more because his replacement will be much worse than the bench player in a major conference. For example, in 2008 Stephen Curry had a Value Add of 9.20 on the main Value Add page when he led Davidson to stunning wins over Gonzaga (82-76), Georgetown (74-70) and Wisconsin (73-56) before losing 59-57 to Kansas in the Elite 8. However, because Curry played for a Low-Major team if he had been injured he would have been replaced by a Low-Major bench player. You can click on the Low Major page to see that Curry was actually worth 15.16 points extra points to Davidson over what a Low-Major bench player would have likely produced, so Davidson without him would have likely lost to Gonzaga by nine or Georgetown by 11.
HOW SHOULD YOU CALCULATE THE PREDICTED SCORE IF KEY PLAYERS ARE MISSING? If you want to know the likely spread of a game, it is best to start with one of three computer rankings that predicts scores by clicking on the pages of statistician Ken Pomeroy, Jeff Sagarin or Ken Massey – and then adjust the final score based on any key player(s) that are missing.
Louisville is expected to win by 17 points at Rutgers Saturday based on www.kenpom.com. If you suddenly learned that Louisville star Russ Smith was not going to play Saturday, you could subtract his Value Add of 6.08 and conclude that Louisville was really only 11 points better than Rutgers in a road game. However, early in the season stars from big conferences are usually a little low because they have not played as many minutes as normal in blowout wins in November and December.
Rather than just clicking on the 2014 Value Add page this early in the season, I would recommend you click on the 2006-2014 Value Add table where you would see that Russ Smith’s 2013 Value Add ended at 8.51, so a better prediction would be that Louisville minus Russ Smith was just eight points better than Rutgers (17 point margin calculated by Pomeroy minus 9 points for what Smith is worth). Typically teams are close to their predicted scores about four out of six games, doing much better one in six games and much worse one in six games.
Follow John Pudner on Twitter @jpudner