Chapter 19: Key Takeaways - Lineup Optimization

Core Concepts Summary

1. Lineup Net Rating Fundamentals

• Net Rating = Offensive Rating - Defensive Rating
• Offensive Rating = (Points Scored / Possessions) x 100
• Defensive Rating = (Points Allowed / Possessions) x 100
• Possessions = FGA + 0.44 x FTA - OREB + TOV
• Modern NBA league average Net Rating: 0.0 (by definition)
• Elite lineups: +10 or better; Poor lineups: -10 or worse

2. Combination Analysis Hierarchy

3. Sample Size Challenge

4. Stagger Principles

• Ensure at least one star on court at all times
• Overlap stars in high-leverage situations (closings)
• Minimum rest between stints: 2-3 minutes
• Maximum continuous playing time: 10-12 minutes

Essential Formulas

Net Rating Calculation

Net Rating = (Points For - Points Against) / Possessions x 100

Standard Error = 11 x 100 / sqrt(Possessions)

95% Confidence Interval

CI = Net Rating +/- 1.96 x Standard Error

Bayesian Posterior Estimate

Posterior Mean = (Prior Variance x Observed + Obs Variance x Prior Mean) /
                 (Prior Variance + Obs Variance)

Shrinkage = Prior Variance / (Prior Variance + Obs Variance)

Two-Man Synergy Score

Synergy = Net Rating (Both On) - Net Rating (A On, B Off)

Spacing Score

Spacing Score = (Reliable Shooters / 5) x 50 +
                (Weighted 3PT% / 0.40) x 50

Lineup Possessions

Possessions = FGA + 0.44 x FTA - OREB + TOV

Implementation Checklist

Setting Up Lineup Analysis Pipeline

• [ ] Data Collection
• [ ] Acquire play-by-play with lineup tracking
• [ ] Gather possession-level statistics
• [ ] Collect player-minute associations

• [ ] Track game state context (score, time)

• [ ] Basic Metrics Calculation

• [ ] Calculate lineup possessions
• [ ] Compute Offensive and Defensive Rating
• [ ] Generate Net Rating for all lineups

• [ ] Filter by minimum sample thresholds

• [ ] Combination Analysis

• [ ] Identify all two-man combinations
• [ ] Calculate three-man core performances
• [ ] Compute synergy scores

• [ ] Rank combinations by Net Rating

• [ ] Advanced Analysis

• [ ] Apply Bayesian regularization
• [ ] Calculate confidence intervals
• [ ] Adjust for opponent quality

• [ ] Luck-adjust for shooting variance

• [ ] Optimization

• [ ] Define player minute constraints
• [ ] Build rotation simulation
• [ ] Evaluate stagger strategies

• [ ] Identify optimal closing lineups

• [ ] Visualization

• [ ] Rotation charts
• [ ] Net Rating comparison graphs
• [ ] Stagger coverage heatmaps
• [ ] Confidence interval displays

Common Pitfalls to Avoid

1. Over-Trusting Small Sample Net Ratings

Problem: Treating a +15 Net Rating in 80 possessions as reliable Solution: Calculate standard error (~12 points); require 200+ possessions minimum

2. Ignoring Context

Problem: Comparing lineups that played against different competition Solution: Adjust for opponent quality or analyze within similar contexts

3. Fixed Closing Lineup Thinking

Problem: Assuming one lineup is optimal for all late-game situations Solution: Vary closers based on score differential, matchups, game state

4. Neglecting Stagger Benefits

Problem: Playing all best players together maximizes immediate impact but creates weak non-star minutes Solution: Stagger to maintain consistent quality throughout game

5. Confusing Correlation with Causation

Problem: Assuming a high Net Rating lineup "causes" good performance Solution: Consider selection effects (when lineup is deployed), opponent context

Quick Reference Tables

Net Rating Interpretation Guide

Minimum Sample Thresholds

Lineup Construction Skill Balance

Lineup Archetype Guide

Closing Lineup

Purpose: High-stakes late-game situations Key Traits: - Best available players - Elite free throw shooting - Ball security (low turnover rate) - Defensive versatility - Multiple shot creators

Transition Lineup

Purpose: Maximize pace and fast breaks Key Traits: - Speed and conditioning - Quick decision-making - Good outlet passing - Less half-court structure needed

Defensive Lineup

Purpose: Protect leads, stop opponents Key Traits: - Elite perimeter defenders - Rim protection - Switchability - Rebounding - Willing to sacrifice some offense

Development Lineup

Purpose: Give young players experience Key Traits: - Mix experienced and young players - Deployed in non-critical situations - Focus on process over results

Rest Lineup

Purpose: Preserve star players Key Traits: - Stars resting - Deep bench players - Often used in blowouts - Maintain competitive level

Application Scenarios

Scenario 1: Evaluating a Trade for Lineup Fit

1. Identify current best lineups
2. Project new player into those combinations
3. Calculate expected spacing change
4. Assess defensive versatility impact
5. Model stagger possibilities with new player
6. Compare to alternatives

Scenario 2: In-Game Rotation Decisions

1. Monitor fatigue levels
2. Check upcoming opponent lineups
3. Calculate optimal substitution timing
4. Ensure star coverage continuity
5. Adjust for game score and time

Scenario 3: Building a Closing Unit

1. Identify situation (protect lead, chase deficit, tie)
2. Prioritize relevant skills (FT%, defense, shot creation)
3. Check matchup considerations
4. Select five players meeting criteria
5. Have backup options for specific adjustments

Scenario 4: Season-Long Rotation Planning

1. Set player minute targets
2. Design stagger schedules
3. Plan for back-to-back games
4. Build in development minutes
5. Create injury contingency lineups

Key Insight Summary

1. Sample size is the fundamental challenge: Most lineup data is too limited for confident conclusions

2. Two-man analysis is more reliable than five-man: Larger samples enable better inference

3. Staggering maximizes star value: Always having a star on court spreads quality across all minutes

4. Closing lineups should be situational: Protect lead = defense/ball security; Chase deficit = shooting/creation

5. Spacing multiplies offensive efficiency: Five capable shooters create non-linear advantages

6. Defensive versatility enables switching: Modern offenses hunt mismatches; versatility neutralizes this

7. Bayesian approaches handle small samples: Shrink extreme observations toward reasonable priors

8. Context matters: Same lineup performs differently vs. different opponents and in different situations

9. System design reduces lineup variance: Good systems make many lineup combinations viable

10. Optimization is continuous: Lineups that work today may not work tomorrow; adaptation is essential

Tools and Resources

Python Libraries

• nba_api: Access NBA lineup data
• pandas: Data manipulation
• numpy: Statistical calculations
• scipy: Optimization algorithms
• sklearn: Bayesian regression

Key Data Sources

• NBA Stats API (official)
• Basketball-Reference (historical)
• Cleaning the Glass (luck-adjusted)
• Second Spectrum (tracking data)

Recommended Workflows

1. Daily: Monitor recent lineup performance
2. Weekly: Update rolling Net Ratings
3. Monthly: Reassess rotation patterns
4. Quarterly: Full lineup optimization review