RAPM is particularly noteworthy for its ability to capture the value of players who contribute to winning in ways that aren't always reflected in traditional statistics or even other advanced metrics. It does this by focusing solely on how a player impacts the scoreboard, regardless of their individual stat accumulation. RAPM ignores the boxscore.

Key Strengths of RAPM:

• Unselfish Play: Unlike many stats that may incentivize selfish stat accumulation, RAPM rewards players who make their team better, even if it doesn't show up in their individual numbers. Everything counts.
• Capturing Intangibles: RAPM can reflect the impact of hard-to-quantify aspects of the game such as screening, defensive communication, ball movement, and the willingness to pass up good shots for great ones.
• Role-Adjusted Value: It measures a player's impact in their role, which can highlight valuable role players who may be overlooked by other metrics.
• Offensive Rebounding Impact: RAPM tends to properly value contributions like offensive rebounding, which can be undervalued in box score-based models.
• Large Sample Reliability: As the sample size increases, RAPM becomes more reliable and can reveal a player's true impact on the game over time.

Methodology:

RAPM uses ridge regression, a statistical technique that helps address the "multicollinearity" problem arising from players frequently playing together. It analyzes how the score changes when players enter or leave the game, while accounting for the quality of teammates and opponents.

Limitations:

• Requires large sample sizes for accuracy
• Can be noisy in small samples
• Doesn't provide insight into how a player is creating their impact

Variations:

Time Decay RAPM: This variation applies an exponential decay function to a player's entire history, giving more weight to recent performance. It's particularly useful for capturing a player's current level while still considering historical data.

3 Year Recency RAPM: This approach uses the last 3 years of data, with heavier weight on more recent possessions. It balances capturing established performance levels with current trajectory.

Lifetime RAPM: A comprehensive 28-year calculation incorporating an age curve to account for typical player development and decline. It's valuable for historical player comparisons and understanding career arcs.

EPM was developed to address some of the limitations of pure RAPM while maintaining its strengths. It incorporates individual player actions to provide a more granular view of player impact.

Key Features:

• Box Score Integration: EPM uses a statistical plus-minus (SPM) model based on box score stats as a prior for RAPM calculations. This helps stabilize the metric and reduce noise.
• Play-by-Play Data: By incorporating detailed play-by-play data, EPM can capture more nuanced aspects of player performance that might be missed in simple box score aggregations.
• Offensive and Defensive Components: EPM provides separate offensive (OEPM) and defensive (DEPM) ratings, allowing for a more detailed understanding of a player's contributions.
• Predictive Power: The combination of box score and on/off data makes EPM more predictive of future performance than many other metrics.

Methodology:

EPM uses a two-step process:

1. A box score-based regression model is used to estimate player impact.
2. These estimates are then used as priors in a regularized adjusted plus-minus (RAPM) calculation.

Strengths:

• More stable than pure RAPM, especially in smaller samples
• Provides insight into how players are creating their impact
• Balances box score production with on-court impact

Limitations:

• May still undervalue certain non-box score contributions
• Could potentially overvalue high-usage players compared to pure RAPM
• Complexity of the model can make it less transparent than simpler metrics

For more detailed information and current EPM rankings, visit Dunks and Threes.

Key Features:

• Daily Updates: DARKO's daily update feature allows it to quickly adjust to changes in player performance, injuries, or role changes.
• Machine Learning: The system uses advanced machine learning algorithms to process and interpret player data, allowing it to identify patterns and trends that might be missed by simpler models.
• Comprehensive Data Integration: DARKO incorporates box score data, on/off impact statistics, and Time Decay RAPM to create a holistic view of player performance.
• Age and Experience Consideration: The model accounts for factors like player age and experience, allowing it to project future performance and development trajectories.

Methodology:

DARKO uses a Kalman filter, a statistical technique often used in time series analysis and control systems. This allows the model to update its estimates of player skill in real-time as new information becomes available.

Key Metrics:

• DPM (DARKO Plus-Minus): An estimate of a player's overall impact in points per 100 possessions.
• ODPM and DDPM: Separate metrics for offensive and defensive impact.
• Projections: DARKO provides both short-term and long-term projections of player performance.

Strengths:

• DARKO's predictive box score stats are far more predictive than someone's season averages
• Rapid adaptation to player performance changes
• Consideration of player development curves
• Integration of multiple data sources for a comprehensive view

Limitations:

• I think it's a brilliant model with how it handles boxscore stats and RAPM, but I question the box component of the model. Doesn't seem to value playmaking much. The Box O-DPM metric rated Jaylen Brown and Nikola Jokic as similar offensive players in 2023.
• Can overrate high usage players.

DARKO is particularly useful for player evaluation, team construction, identifying undervalued players, and projecting player development. Its daily updates make it valuable for applications like daily fantasy sports.

For up-to-date DARKO projections and more detailed explanations of the methodology, visit DARKO.app.

Key Components:

• Box Score Prior: LEBRON starts with a box score estimate based on boxPIPM (Player Impact Plus-Minus) weightings. This provides a foundation based on individual player statistics.
• Offensive Archetypes: The box score prior is stabilized using Offensive Archetypes, which helps account for different player roles and styles of play.
• Luck-Adjusted RAPM: LEBRON incorporates a luck-adjusted version of Regularized Adjusted Plus-Minus. This helps account for factors outside of a player's control, such as teammate shooting performance.
• Offensive and Defensive Components: LEBRON provides separate offensive (O-LEBRON) and defensive (D-LEBRON) ratings, allowing for a more detailed understanding of a player's impact.

Methodology:

LEBRON uses a multi-step process:

1. Calculate a box score prior based on individual player statistics
2. Adjust this prior using Offensive Archetypes to account for player roles
3. Incorporate on/off court data using luck-adjusted RAPM techniques
4. Combine these components to produce final LEBRON values

Strengths:

• Comprehensive integration of box score and on/off data
• Adjustment for player roles and archetypes
• Luck adjustment to reduce noise from factors outside player control
• Separate offensive and defensive ratings for more nuanced evaluation

Limitations:

• Is the luck adjustment accurate?
• Reliance on accurate role classification for proper context
• Box Prior doesn't use advanced play by play data as far as I can tell, so stats like offensive fouls drawn will not be in the box score prior which is going to result in elite guards being undervalued

LEBRON values are expressed in points per 100 possessions above average, with 0 representing an average player. The metric also includes Wins Added calculations to estimate a player's total impact over a season.

For a more detailed explanation of LEBRON methodology and current player ratings, visit the BBall Index LEBRON Introduction.

Key Metrics:

• Total Contests per 100: The number of shot attempts contested by the player per 100 defensive possessions.
• Contest Dif%: The difference in field goal percentage when the player is contesting versus the shooter's normal percentage.
• Rim Contests per 100: The number of shot attempts contested by the player at the rim per 100 defensive possessions.
• Rim Contest Dif%: The difference in field goal percentage for rim shots when the player is contesting versus the shooter's normal percentage at the rim.
• rRim Acc On: The relative rim field goal percentage when the player is on the court, compared to the league average for that year.
• Rim Acc On/Off: The difference in rim field goal percentage when the player is on the court versus off.
• rRim Freq On: The relative frequency of opponent rim attempts when the player is on the court, compared to the league average for that year.
• Rim Freq On/Off: The difference in frequency of opponent rim attempts when the player is on the court versus off.

Interpretation:

For most of these metrics, negative values are better as they indicate that the player's presence reduces opponent shooting efficiency or frequency of attempts at the rim. The contest metrics (per 100) are exceptions, where higher values indicate more active defensive engagement.

Strengths:

• Provides detailed insights into a player's defensive impact
• Captures aspects of defense not reflected in traditional box score stats
• Allows for comparison of defensive impact across different team contexts

Limitations:

• Does not capture all aspects of defense (e.g., communication, help defense)
• Can be influenced by team defensive schemes
• Requires context of a player's role and defensive assignments for full interpretation

Key Metrics:

• On-Ball Percent: The percentage of time a player has possession of the ball.
• Potential Assists per 100: The number of passes that would be assists if the receiver made the shot, per 100 possessions.
• Pot Ast per 1 min On-Ball: Potential assists created per minute of on-ball time.
• Assists per 100: The number of actual assists per 100 possessions.
• BadPass%: The percentage of potential assists that result in turnovers.
• TOV Adj. Assist EFG%: The effective field goal percentage on shots assisted by the player, adjusted for turnovers.

Interpretation:

Higher values are generally better for most metrics, except BadPass% where lower values indicate better ball security. These metrics together provide a comprehensive view of a player's passing ability, frequency, and efficiency.

Strengths:

• Provides a nuanced view of passing beyond simple assist numbers
• Accounts for both volume and efficiency of playmaking
• Helps identify players who create scoring opportunities, even if teammates don't convert

Limitations:

• Does not account for the difficulty of passes or defensive pressure
• Potential assists can be inflated by players who pass to high-volume shooters
• Context of team offensive system is important for full interpretation

Key Metrics:

• STOP%: The percentage of possessions where the player forces a turnover or blocks a shot recovered by their team, per 100 possessions. It combines steals, offensive fouls drawn, and blocks recovered by the defense.
• Deflections per 100: The number of times a player gets a hand on the ball on defense, per 100 possessions.
• STL per 100: Steals per 100 possessions.
• Off Foul Drawn per 100: Offensive fouls drawn per 100 possessions.
• rFTOV: Relative forced turnovers, comparing the player's forced turnover rate to the league average.
• BLKs Recov per 100: Blocks recovered by the defense per 100 possessions.
• BLKs per 100: Total blocks per 100 possessions.
• BLK Recov %: Percentage of blocks that are recovered by the defense.

Interpretation:

Higher values are generally better for all these metrics. STOP% is a comprehensive metric that captures overall defensive playmaking ability. It represents the percentage of possessions where the player directly causes a change of possession through steals, drawn offensive fouls, or recovered blocks.

Strengths:

• Provides a multi-faceted view of defensive impact
• Captures disruptive defensive plays that may not show up in traditional box scores
• STOP% offers a single number to gauge overall defensive playmaking

Limitations:

• Does not capture all aspects of defense, such as contesting shots or proper positioning
• High values in these metrics can sometimes indicate gambling on defense
• Team defensive schemes can influence individual metrics