Regularized Adjusted Plus Minus (RAPM) is, by most statistical estimations, the best predictive “single number” statistic. While metrics such as Player Efficiency Rating (PER) and Win Shares (WS) are more commonly referenced, RAPM outperforms both, as well as Wins Produced, in retrodictive studies used to measure the ability to predict future outcomes.
Using RAPM as a base, I projected the Golden State Warriors’ 2013-14 record. The individual RAPM data for the 2012-13 season is available here. However, as common sense would suggest, players improve and decline over time, and last season’s RAPM will not necessarily be an accurate predictor of this season’s.
To adjust for the player development and decline, I relied on the RAPM versus age provided on the RAPM source site.
Here is the projected RAPM per 100 possessions for each Warrior next season. I did not assign a value to Nemanja Nedovic, both because I do not expect him to receive significant playing time and because I could not find a reliable way to convert foreign based stats into a projected RAPM.
Next, I estimated the minutes per game for each player to create percent of team minutes played. These projections were simply based of intuition, not statistical methods. These projections are not inscrutable, and may prove to be inaccurate. However, I think they reflect a reasonable rotation.
To convert RAPM to a wins metric, I set the replacement level RAPM at -2.5. Both -2.5 and -3.5 are commonly used as replacement level. To generate Wins Above Replacement, I multiplied each players’ RAPM above replacement by the percentage of possible minutes they are expected to play by 2.54–2.54 is a “rough” estimate used to replace the correlation between point margin and season wins. The commonly used equation is Net Rating times 2.54 plus 41 equals season wins.
By this method, the Warriors are expected to generate 47.65 wins above replacement.
However, this does not reflect season wins, merely wins above the amount expected from a team constructed of players with a -2.5 RAPM. A team playing players of a -2.5 RAPM would have a net margin of -12.5, because 5 -2.5 players would play together the entire game. A team with a -12.5 net rating is expected, by the previously mentioned equation, to win around 9.25 games. Thus, by this estimation, the Warriors are projected to win 56.9 games next season.
This, however, is based on a minutes projection that ignores the significant possibility of injuries. While the per game figures may prove to be accurate, an injury forcing a missed game would reduce aggregate wins above replacement for a player, reducing team wins.
Nearly every team will suffer through injury issues. The Warriors’ heavy reliance on Andrew Bogut and Stephen Curry, first and third in projected wins above replacement, means that injuries to them could drastically affect their win total, not a stunning revelation.
For comparison, an ESPN poll of True Hoop Network writers projects the Warriors to win only 50 games. So, 57 is a very optimistic number, a best-case health scenario, but shows the potential of the Warriors’ roster.