1. Value Over Replacement Player
The system is similar to baseball WAR (Wins Above Replacement), meaning if a team is a 2-point favorite in a game before learning a player with a Value Add Rating of 4.0 cannot play and must be replaced by a borderline replacement player, then you subtract his rating and the team would be a 2-point underdog.
The first difference between Value Add and WAR are than WAR represents WINS above replacement over the course of a SEASON while Value Add measures POINTS above replacement in a typical GAME. The second difference is that in baseball a player is replaced by one player (a shortstop is replaced by a shortstop who also takes a spot in the line-up), the impact of losing one player has a domino effect of perhaps a new "ninth player" getting a few minutes, but also of other players having to play more and thus becoming less efficient and drawing more defensive attention as calculated in great detail in this Big Apple Buckets piece.
2. The four main ratings
Value Add v5.0 Pts/Game - The most important figure is the Actual Value Add figure generated from the 5th version (v5.0) of the system. This gives the actual point impact a player has on a game, so a team that won 73-71 with a player with a Value Add of 4.00 would best estimated to lose 71-73 if a replacement player played instead of him. This final figure is a factor of the three figures that follow, but also accounts for a player getting less possessions because he is on a very good team and factors like this. In earlier years, this figure had to be adjusted based on the position a player played (the former PGPer) because point guards and perimeter defenders rated so much lower because they faced more risk of turnovers and could be muscled up and prevented from getting near the basket. Luckily the "freedom of movement rules" have allowed guards to move freely and there is now virtually no statistical difference between the average ratings of guards, forwards and centers eliminating a cumbersome adjustment.
AdjO - this is the figure of the player's impact on the identical rating that www.kenpom.com calculates for teams. So if www.kenpom.com gives a team a rating of 110.0 points per 100 trips down the court and a player's AdjO is 4.0, then the www.kenpom.com rating without him would be estimated at 107.0 (110.0 minus 4.00).
AdjD - the same is true of the defensive figure, where the bigger negative indicates great defense. So if www.kenpom.com calculates a player's defense at 94.0 AdjD per 100 trips down the court, a player's AdjD of 4.0 would be very bad, meaning the team would actually be a much better 90.0 AdjD with him replaced. However, a -4.00 AdjD for a player would be very good, meaning without the player the defense would start allowing 98.0 per 100 trips rather than just 94.0 if the player could not play.
AdjEM - The total of the two previous numbers equals the overall impact on www.kenpom.com team rating with the same name, which he gets by SUBTRACTING AdjD from AdjO as we do for the player. This is actually the Value Add 3.0 Rating, however it is in terms of 100 trips down the court rather than the lower number that results from the 70 or so trips during a game. While the AdjEM allowed a ball hog on a bad team to ring up an incredibly high rating by taking many of the shots, Version 5.0 calculates how many points the player would actually add in GAME for a strong team, so the Value Add number is much lower than the AdjEM.
3. Detailed Science Behind Value Add
If you want to truly follow the math, read on.
The science behind Value Add Basketball is laid out in Cracked Sidewalks pieces beginning in May 2011 and now. A massive effort to upgrade the system and pinpoint the impact each player had on his team's Offensive and Defensive Efficiency Ratings culminated at the end of 2015 with Value Add Version 3.0, 1000 lines of code, a series of improvements detailed here, and a simple explanation with illustrative tables to make the impact of each player easy to understand.
Others such as the Basketball Prospectus took the formula and suggested additional calculations based of the system, and Version 4.0 was introduced at the outset of the 2016-17, but honestly was the one version that did not work. We went back to the drawing board, and reach incredible breakthroughs in Version 5.0, which results in a current ratings that pinpoints everything we can measure on a court. Those improvements are detailed here.
John is now the Executive Director of Take Back Our Republic, a group working to find conservative solutions to the problem of money in politics. Anyone who wants to join that group simply needs to go to that page and type in their email and zipcode.