Skip to main content

​
Overview

Advanced metrics provide deeper insight into player and team performance through calculated statistics, play-type analysis, and adjusted metrics that account for context and competition.

​
Play-Type Analysis

​
playtype.csv

Player-level play-type statistics showing how players score in different offensive situations. File: playtype.csv / playtype_p.csv
​
PLAYER_ID
integer
required
Unique player identifier
​
Player
string
required
Player name
​
Team
string
Team abbreviation
​
TEAM_ID
integer
Team identifier
​
TEAM_NAME
string
Full team name
​
year
integer
required
Season year
​
playtype
string
required
Type of offensive play:
  • Isolation: 1-on-1 isolation plays
  • Transition: Fast break opportunities
  • PRBallHandler: Pick and roll ball handler
  • PRRollMan: Pick and roll roll man
  • Postup: Post-up plays
  • Spotup: Spot-up shooting
  • Handoff: Handoff plays
  • Cut: Cutting to the basket
  • OffScreen: Coming off screens
  • Putbacks: Offensive rebound putbacks
  • Misc: Miscellaneous plays
​
Position
string
Player position (G, F, C, or combinations)
​
GP
integer
Games played
​
% Time
float
Percentage of offensive possessions used in this play type
​
Poss
integer
Total possessions of this play type
​
Points
integer
Total points scored on this play type
​
PPP
float
Points per possession - primary efficiency metricElite: > 1.10 PPP
Good: 0.95-1.10 PPP
Average: 0.85-0.95 PPP
Below Average: < 0.85 PPP
​
FGm
integer
Field goals missed (alternative spelling in some datasets)
​
FGM
integer
Field goals made
​
FGA
integer
Field goals attempted
​
FG%
float
Field goal percentage
​
aFG%
float
Adjusted field goal percentage (accounts for 3-point value)
​
%TO
float
Turnover percentage - portion of possessions ending in turnover
​
%FT
float
Free throw percentage - portion of possessions resulting in free throws
​
%SF
float
Shooting foul percentage - portion of possessions drawing shooting fouls
​
%Score
float
Scoring frequency - portion of possessions resulting in points
​
FTFreq%
float
Free throw frequency (duplicate field)
​
TOVFreq%
float
Turnover frequency (duplicate field)
​
SFFreq%
float
Shooting foul frequency (duplicate field)
​
And OneFreq%
float
And-one frequency - made basket with foul drawn
​
ScoreFreq%
float
Scoring frequency (duplicate field)
​
Percentile
integer
Percentile rank among all players for this play type (0-100)Based on PPP efficiency relative to league average
Example: 
Position,Player,Team,GP,% Time,Poss,Points,PPP,FGm,FGM,FGA,FG%,aFG%,%TO,%FT,%SF,%Score,year,playtype,...
G,Stephen Curry,GSW,79,12.4,412,478,1.16,142,178,320,55.6,58.9,11.2,8.5,7.3,62.1,2015,Transition,...

​
Regularized Adjusted Plus-Minus (RAPM)

​
rapm.csv

Advanced plus-minus metrics that estimate player impact while controlling for teammates, opponents, and game context. File: rapm.csv
​
playerId
integer
required
Unique player identifier
​
playerName
string
required
Player name
​
season
string
required
Season in format “YYYY-YY” (e.g., “2014-15”)
​
primaryKey
string
required
Unique key combining player ID and season (e.g., “101106_2009-10”)

​
RAPM Metrics

​
RAPM
float
Overall Regularized Adjusted Plus-MinusNet points per 100 possessions contributed by player relative to average, adjusted for teammates and opponents.Interpretation:
  • Elite: +3.0 or higher
  • Good: +1.5 to +3.0
  • Average: -1.0 to +1.5
  • Below Average: < -1.0
​
RAPM_Rank
integer
League rank by RAPM (lower is better)
​
RAPM__Off
float
Offensive RAPM - points added per 100 possessions on offense
​
RAPM__Off_Rank
integer
League rank by offensive RAPM
​
RAPM__Def
float
Defensive RAPM - points prevented per 100 possessions on defenseNote: Positive values indicate better defense (fewer points allowed)
​
RAPM__Def_Rank
integer
League rank by defensive RAPM
​
RAPM__intercept
float
Regression intercept for RAPM model

​
Luck-Adjusted RAPM (LA_RAPM)

​
LA_RAPM
float
Luck-adjusted RAPM - adjusts for shooting variance and randomness
​
LA_RAPM_Rank
integer
League rank by luck-adjusted RAPM
​
LA_RAPM__Off
float
Offensive luck-adjusted RAPM
​
LA_RAPM__Off_Rank
integer
Offensive LA_RAPM rank
​
LA_RAPM__Def
float
Defensive luck-adjusted RAPM
​
LA_RAPM__Def_Rank
integer
Defensive LA_RAPM rank
​
LA_RAPM__intercept
float
Regression intercept for LA_RAPM model

​
Component Metrics

RAPM broken down by four factors (eFG%, turnover rate, offensive rebound rate, free throw rate):
​
RA_EFG
float
RAPM component for effective field goal percentage
​
RA_EFG__Off
float
Offensive eFG% RAPM component
​
RA_EFG__Def
float
Defensive eFG% RAPM component
​
RA_FTR
float
RAPM component for free throw rate
​
RA_FTR__Off
float
Offensive FTR RAPM component
​
RA_FTR__Def
float
Defensive FTR RAPM component
​
RA_ORBD
float
RAPM component for offensive rebound differential
​
RA_ORBD__Off
float
Offensive rebound RAPM component
​
RA_ORBD__Def
float
Defensive rebound RAPM component
​
RA_TOV
float
RAPM component for turnover rate
​
RA_TOV__Off
float
Offensive turnover RAPM component
​
RA_TOV__Def
float
Defensive turnover RAPM component (turnovers forced)
All component metrics include _Rank and _intercept fields following the same pattern as overall RAPM. Example: 
playerId,playerName,RAPM,RAPM_Rank,RAPM__Off,RAPM__Def,...,season,primaryKey
101106,"Andrew Bogut",3.06,18,0.49,2.57,...,2009-10,101106_2009-10

​
Calculated Shooting Metrics

​
True Shooting Percentage (TS%)

Found in: scoring.csv, totals.csv, passing.csv, team_shotzone.csv Formula: PTS / (2 * (FGA + 0.44 * FTA)) Interpretation:
  • Elite: > 60%
  • Good: 55-60%
  • Average: 52-55%
  • Below Average: < 52%
Accounts for:
  • Value of 3-point shots (worth 1.5x a 2-pointer)
  • Value of free throws (efficient scoring)
  • And-one free throws (the 0.44 coefficient)

​
Effective Field Goal Percentage (eFG%)

Found in: player_shooting.csv, tracking.csv, team_shooting.csv, team_shotzone.csv Formula: (FGM + 0.5 * 3PM) / FGA Interpretation:
  • Elite: > 55%
  • Good: 52-55%
  • Average: 48-52%
  • Below Average: < 48%
Adjusts field goal percentage to account for the added value of 3-point shots.

​
Points Per Possession (PPP)

Found in: playtype.csv, teamplay.csv Formula: Points / Possessions Interpretation:
  • Elite: > 1.10
  • Good: 0.95-1.10
  • Average: 0.85-0.95
  • Below Average: < 0.85
Primary metric for evaluating efficiency of specific play types.

​
Usage Examples

​
Finding Elite Pick and Roll Ball Handlers

import pandas as pd

# Load play-type data
playtype = pd.read_csv('playtype.csv')

# Filter for PnR ball handlers with significant volume
pr_handlers = playtype[
    (playtype['playtype'] == 'PRBallHandler') &
    (playtype['year'] == 2023) &
    (playtype['Poss'] >= 200)  # Minimum possessions
].sort_values('PPP', ascending=False)

print(pr_handlers[['Player', 'Team', 'Poss', 'PPP', 'Percentile', '%Score', '%TO']].head(15))

​
Analyzing RAPM Two-Way Players

# Load RAPM data
rapm = pd.read_csv('rapm.csv')

# Filter for 2022-23 season
rapm_2023 = rapm[rapm['season'] == '2022-23']

# Find elite two-way players (top 50 in both offense and defense)
two_way = rapm_2023[
    (rapm_2023['RAPM__Off_Rank'] <= 50) &
    (rapm_2023['RAPM__Def_Rank'] <= 50)
].sort_values('RAPM', ascending=False)

print(two_way[['playerName', 'RAPM', 'RAPM__Off', 'RAPM__Def', 'RAPM_Rank']])

​
Comparing Play-Type Efficiency by Position

# Load play-type data
playtype = pd.read_csv('playtype.csv')

# Focus on isolation plays
iso = playtype[
    (playtype['playtype'] == 'Isolation') &
    (playtype['year'] == 2023) &
    (playtype['Poss'] >= 100)
]

# Group by position
iso_by_pos = iso.groupby('Position').agg({
    'PPP': 'mean',
    'FG%': 'mean',
    '%TO': 'mean',
    'Poss': 'sum'
}).round(3)

print(iso_by_pos.sort_values('PPP', ascending=False))

​
True Shooting vs. Effective Field Goal Analysis

# Load scoring and totals data
scoring = pd.read_csv('scoring.csv')
totals = pd.read_csv('totals.csv')

# Merge datasets
merged = pd.merge(
    scoring[['nba_id', 'Player', 'year', 'TS%', 'PTS']],
    totals[['nba_id', 'year', 'FGA', 'FTA']],
    on=['nba_id', 'year']
)

# Calculate FTA rate
merged['FTA_rate'] = merged['FTA'] / merged['FGA']

# Players with high TS% but low FTA rate (pure shooters)
filtered_2023 = merged[
    (merged['year'] == 2023) &
    (merged['PTS'] >= 500)  # Min scoring threshold
]

# Find efficient scorers who don't rely on free throws
efficient_shooters = filtered_2023[
    (filtered_2023['TS%'] >= 60) &
    (filtered_2023['FTA_rate'] < 0.3)
].sort_values('TS%', ascending=False)

print(efficient_shooters[['Player', 'TS%', 'FTA_rate', 'PTS']])

​
RAPM Component Analysis

# Load RAPM data
rapm = pd.read_csv('rapm.csv')

# Focus on recent season
rapm_recent = rapm[rapm['season'] == '2022-23']

# Analyze component contributions
rapm_recent['shooting_impact'] = rapm_recent['RA_EFG__Off'] - rapm_recent['RA_EFG__Def']
rapm_recent['turnover_impact'] = rapm_recent['RA_TOV__Off'] - rapm_recent['RA_TOV__Def']
rapm_recent['rebound_impact'] = rapm_recent['RA_ORBD__Off'] - rapm_recent['RA_ORBD__Def']
rapm_recent['ft_impact'] = rapm_recent['RA_FTR__Off'] - rapm_recent['RA_FTR__Def']

# Find players who excel at specific skills
top_shooters = rapm_recent.nlargest(10, 'shooting_impact')[['playerName', 'shooting_impact', 'RAPM']]
top_playmakers = rapm_recent.nsmallest(10, 'turnover_impact')[['playerName', 'turnover_impact', 'RAPM']]

print("Top Shooting Impact:")
print(top_shooters)
print("\nTop Playmaking Impact (low turnovers):")
print(top_playmakers)

​
Play-Type Specialization Index

# Load play-type data
playtype = pd.read_csv('playtype.csv')

# Calculate specialization for 2023
pt_2023 = playtype[playtype['year'] == 2023]

# For each player, find their most efficient play type (min 50 poss)
player_best = pt_2023[
    pt_2023['Poss'] >= 50
].sort_values('PPP', ascending=False).groupby('PLAYER_ID').first().reset_index()

# Players who are elite specialists (>90th percentile)
specialists = player_best[
    player_best['Percentile'] >= 90
].sort_values('PPP', ascending=False)

print(specialists[['Player', 'playtype', 'PPP', 'Percentile', 'Poss']])

​
Integration with Core Schemas

​
Joining Advanced Metrics with Player Stats

# Combine RAPM with basic stats
scoring = pd.read_csv('scoring.csv')
rapm = pd.read_csv('rapm.csv')

# Extract year from season string
rapm['year'] = rapm['season'].str[:4].astype(int)

# Merge
advanced = pd.merge(
    scoring,
    rapm[['playerId', 'year', 'RAPM', 'RAPM__Off', 'RAPM__Def']],
    left_on=['nba_id', 'year'],
    right_on=['playerId', 'year'],
    how='inner'
)

print(advanced[['Player', 'year', 'PTS', 'TS%', 'RAPM']].head())

​
Comprehensive Player Profile

# Create complete player profile combining multiple sources
def get_player_profile(player_id, season):
    scoring = pd.read_csv('scoring.csv')
    playtype = pd.read_csv('playtype.csv')
    rapm = pd.read_csv('rapm.csv')
    hustle = pd.read_csv('hustle.csv')
    
    # Basic stats
    basic = scoring[
        (scoring['nba_id'] == player_id) &
        (scoring['year'] == season)
    ]
    
    # Play types
    plays = playtype[
        (playtype['PLAYER_ID'] == player_id) &
        (playtype['year'] == season)
    ].sort_values('Poss', ascending=False)
    
    # RAPM
    season_str = f"{season}-{str(season+1)[-2:]}"
    impact = rapm[rapm['season'] == season_str]
    
    # Hustle
    hustle_stats = hustle[
        (hustle['PLAYER_ID'] == player_id) &
        (hustle['year'] == season)
    ]
    
    return {
        'basic': basic,
        'play_types': plays,
        'impact': impact,
        'hustle': hustle_stats
    }

# Example usage
profile = get_player_profile(201939, 2023)  # Stephen Curry
print("Play Type Breakdown:")
print(profile['play_types'][['playtype', 'Poss', 'PPP', 'Percentile']])

Build docs developers (and LLMs) love

Get started for freeTalk to us