The APBRmetrics community has come a long way in a short time, and sometimes it is easy to forget the magnitude of that progress. I remember just five years ago, during the 2003-04 season, posting by hand leaderboards in various advanced statistics because they could not be found anywhere on the Internet. That memory makes me take a little extra pride in officially introducing Basketball Prospectus' NBA statistics pages.
Surfing over to the statistics section, you will find a complete index of pages for each player and team. These pages are also linked in every Basketball Prospectus column on the first reference to the player or team. (For example, see LeBron James or the Cleveland Cavaliers.) Also, whenever these stats are used within a column, you can mouse over the abbreviation to see a definition and explanation.
Many of the statistics you'll find on these pages are exclusive to Basketball Prospectus. The player pages include numbers derived from my WARP rating system as well as Bradford Doolittle's NBAPET rating system. Meanwhile, the team pages include the statistics Bradford uses each week to create the Prospectus Hoops List rankings. Above and beyond these, there are other player stats, like two-point percentage, you simply won't find anywhere else.
At its heart, Basketball Prospectus is about the numbers. While Bradford and I have been making use of the statistics on these pages in our work over the last year and a half, now they are only a click away for all of our readers at any time. I hope they will become a regular part of your trips to Basketball Prospectus. Also, be on the lookout for future enhancements, including multiple seasons of player statistics and leaderboards.
READING THE PLAYER STATS PAGES
The Primary Statistics line should be pretty self-explanatory. The one stat worth noting is the +/- stat at the far right. This is a player's raw plus-minus, which is how much his team has outscored opponents by or been outscored by with the player on the floor.
The Advanced Value Statistics line includes the most important player metrics as well as numbers from the WARP rating system. Effective field-goal percentage and True Shooting Percentage are the two best ways to represent a player's shooting efficiency (TS% factoring in trips to the free-throw line, while eFG% does not); in concert with a player's Usage rate, they give a good picture of a player's value as a scorer. Rebound percentage is the best measure of rebounding, and Pass rating is a junk stat used to measure a player's ability as a passer based on both frequency of assists and assist-to-turnover rate.
The last five numbers are all components of WARP. In concept, the rating system seeks to create a team of the player plus four average teammates--similar to the Marginal Lineup Value used by our Baseball Prospectus colleagues. The presented Offensive and Defensive Ratings are for this "team," and Bias expresses how much better the team is on offense or defense. Those ratings are used to create a win percentage for the imaginary team, and this is used to determine how many wins above a replacement-level contributor the player has created for his team over the course of the season.
The Advanced Skill Statistics line offers a variety of metrics, many of them found only on Basketball Prospectus, that seek to measure player skills. Assist percentage, steal percentage, block percentage and foul percentage are all calculated per team possessions to remove the impact of pace. There is also a set of what I term player tendency stats--the percentage of a player's possessions that are used on two-point and three-point shots, trips to the free-throw line and turnovers. Lastly, there is block against percentage, the frequency at which a player's own shots are rejected that can be a meaningful indicator for players who struggle to get shots off inside against shot-blockers (see Love, Kevin).
The last table on the player page lists an assortment of data collected from the NBAPET system, which is compiled-from-game by game box scores.
The first three columns reflect a player's offensive usage and production: oPOSS is an estimate of the raw number of possessions used on the offensive end by each player. The formula is the same as the one used to calculate team possessions (see below) except that in each game, the total number of individual possessions is brought into balance with the team possessions using a multiplier. PC is an estimate of points created by the player. Points created in NBAPET tabulates only "positive" contributions, ie., making a shot, getting to the foul line, grabbing a rebound, etc. The value of each category varies somewhat from year to year because the value of a possession changes a little each season. Points lost (PL) is an estimate of offensive efficiency, depending on the league-wide value of a possession. (PC-(oPOSS * league Pts/Poss)). Thus, the lower the points-lost total, the more efficient a player is. Zero is exactly average. Soon to be added to the table will be PC100, or points created per 100 possessions (PC / oPOSS * 100).
The next three columns are an attempt to quantify a player's defensive contributions. In each box score, NBAPET "charges" each player for the offensive production from his most likely counterpart. In general, starting point guards are matched up with opposing starting point guards, etc., though an effort is made to capture cross-matchups. Also, these figures are affected by playing time. Thus if Dwight Howard is matched with Kendrick Perkins and Howard plays 39 minutes and Perkins 21, Perkins' defensive totals will reflect just 21 minutes' worth of Howard's production. The final 18 minutes are distributed among Orlando's reserve interior defenders. dPOSS is an estimate of a player's defensive possessions used on the season. Points against (PA) would be more accurately described as "points created against" and uses the same methodology as points created. Points saved (PS) is the counterpart to points lost. Thus, it is an aggregate total reflecting a player's defensive efficiency as compared to league average. The higher the total, the better. Zero is exactly average. Soon to be added to the table will be PA100, or points allowed per 100 possessions (PA / dPOSS * 100).
Two more defensive metrics follow. First is dMULT, or defensive multiplier. This measures how effectively a player has limited his counterparts' efficiency on a per-possession basis compared to their season rate. Thus, if Ron Artest has a dMULT of .950, that says Artest has held his counterparts to 95% of their normal production on a per possession basis. dQUAL (defensive counterpart quality) is a strength-of-opponent measure. It calculates the composite season points created per possession figures for all of a player's box score counterparts and compares it to the league average. A dQUAL total of 1.08 means that the player's counterparts have been eight percent better than league average.
Finally, the last three columns reflect the meat of the NBAPET player rating system. The aggregate offensive and defensive points created, saved and lost are combined to calculate the number of wins a player is responsible for. The raw total is WP, or wins produced. WP82 is wins produced prorated for the full 82-game season. However a player's availability is reflected in this total. Thus if a player has missed 50% of his team's games, his WP82 figure will reflect this--it isn't "wins produced per 82 games." WP3K, however, removes all playing time and availability differences. It's simply wins produced per 3000 minutes.
READING THE TEAM STATS PAGES
As mentioned, the first table on the team statistics page contains the information used to compile the power rating included in the weekly Hoops List, as well as some additional data collected from NBAPET. W82 is the team's 82-game win pace (current winning percentage * 82). pW is the Pythagorean win pace: ((Tm PTS^14 / (Tm PTS^14 + Opp PTS^14))) * 82. LUCK is the difference between W82 and pW and thus is an estimate of how many wins a team is on pace to outperform or underperform its point differential over a full season. Hwin, Hloss, Awin, Aloss are self-explanatory. pOPP% reflects the aggregate winning percentage of a team's past opponents. fOPP% is the same figure for a team's future opponents. SOSrnk is the team's league ranking in pOPP%. The following column, aWIN%, adjusts a team's actual winning percentage based on point differential, home/road performance and strength of schedule. The power rating (POW) is aWIN% * 82, with its league rank listed in the column to the right. In between, is clW and clL, or close wins and close losses. These include any game a team has played that was decided by five or fewer points.
The Team Overview line offers the most critical information about a team's performance at a glance, including Dean Oliver's Four Factors. Here and throughout the page, defensive stats are expressed as "Cleveland opponents." The first cut of team analysis is Offensive and Defensive Ratings, points scored and allowed per 100 possessions. Note that Basketball Prospectus uses the formula .96*(FGA + (.44*FTA) - OR + TO) to estimate (NBA) possessions. This is different from the formulas used by Basketball-Reference.com and KnickerBlogger.net. John Hollinger of ESPN Insider uses the same formula, but without the .96 multiplier that accounts for team offensive rebounds (primarily when a missed shot is knocked out of bounds by a defensive team).
The Four Factors were designed by Oliver to measure the critical parts of team performance--shooting (eFG%), getting to the free-throw line (FTA/FGA), rebounding (OR%/DR%) and avoiding turnovers (TO%). Naturally, these same factors also apply at the defensive end, so there are in a sense really eight factors. Of these, shooting is the most important for a team to control, and about 40 percent of a team's performance is explained by its shooting and that of its opponents. Turnovers (25%) are second in importance, followed by rebounding (20%) and free throws (15%). Using the Four Factors and team rankings, it is easy to spotlight areas of improvement or need for a team.
The Team Per Game line offers--you guessed it!--per-game statistics.
The Team Advanced line features a number of additional metrics to track team performance, including team True Shooting Percentage and two-point percentage which offer further detail above and beyond the Four Factors.
Kevin Pelton is an author of Basketball Prospectus.
You can contact Kevin by clicking here or click here to see Kevin's other articles.