Chapter 25: Case Study 2 - Evaluating the Efficiency of NBA Betting Markets

Introduction

Betting markets represent the collective wisdom of thousands of participants, including professional sports bettors, algorithmic trading systems, and recreational gamblers. This case study examines the efficiency of NBA betting markets by analyzing closing line accuracy, identifying potential inefficiencies, and testing whether systematic approaches can generate positive expected value.

Part 1: Market Efficiency Framework

Efficient Market Hypothesis in Sports Betting

The efficient market hypothesis (EMH) adapted to sports betting suggests that: - Closing lines incorporate all available information - Consistent profitable betting is not possible - Any apparent edges are either random or quickly arbitraged away

Weak form efficiency: Lines can't be beaten using historical line/result data Semi-strong efficiency: Lines can't be beaten with any public information Strong efficiency: Lines can't be beaten with any information (including private)

Data and Methodology

Dataset: - 5 seasons of NBA regular season games (6,150 games) - Opening lines, closing lines, and results - Point spreads and totals (over/under)

Metrics: - ATS (Against the Spread) accuracy - ROI (Return on Investment) - CLV (Closing Line Value)

Part 2: Measuring Market Accuracy

Closing Line Performance

The closing line is the market's final consensus prediction:

Spread Accuracy: | Measure | Value | |---------|-------| | Games where favorite covered | 48.2% | | Games where underdog covered | 49.5% | | Push (exact spread) | 2.3% | | Mean Absolute Error | 9.8 points | | RMSE | 12.1 points |

The closing spread has essentially no bias toward favorites or underdogs.

Total (Over/Under) Accuracy: | Measure | Value | |---------|-------| | Overs hit | 49.4% | | Unders hit | 48.8% | | Push | 1.8% | | Mean Absolute Error | 18.2 points | | RMSE | 23.5 points |

Calibration of Implied Probabilities

Converting moneyline odds to implied probabilities and checking calibration:

Implied Win Prob Actual Win % Sample Size
20-30% 26.8% 420
30-40% 35.2% 680
40-50% 44.8% 890
50-60% 54.1% 1050
60-70% 64.5% 920
70-80% 73.8% 650
80-90% 82.1% 380

Markets are well-calibrated, with slight favorite-longshot bias (longshots slightly overvalued).

Part 3: Testing Market Inefficiencies

Test 1: Home Underdog Bias

Hypothesis: Home underdogs may be undervalued due to public preference for favorites.

Results: | Category | ATS Record | Win % | p-value | |----------|------------|-------|---------| | All home underdogs | 512-488 | 51.2% | 0.38 | | Home dogs +3 to +7 | 245-228 | 51.8% | 0.31 | | Home dogs +7.5+ | 142-136 | 51.1% | 0.42 |

Conclusion: No statistically significant edge for home underdogs.

Test 2: Back-to-Back Effects

Hypothesis: Markets may undervalue rest advantage.

Results: | Scenario | ATS Record | Win % | p-value | |----------|------------|-------|---------| | Rested vs B2B | 628-605 | 50.9% | 0.42 | | B2B vs Rested | 595-620 | 49.0% | 0.38 | | Both B2B | 185-192 | 49.1% | 0.40 |

Conclusion: Markets adequately price rest differentials.

Test 3: Line Movement

Hypothesis: Following "sharp" line movement may be profitable.

Method: Bet on sides where line moved 2+ points from open to close.

Results: | Movement Direction | ATS Record | Win % | p-value | |-------------------|------------|-------|---------| | Steam move (follow) | 412-398 | 50.9% | 0.41 | | Fade public (contrarian) | 388-395 | 49.6% | 0.45 |

Conclusion: Simple line movement strategies don't yield consistent edges.

Test 4: Totals Market Inefficiencies

Hypothesis: High totals may be overvalued due to recreational preferences.

Results: | Total Range | Under Record | Under % | p-value | |-------------|--------------|---------|---------| | 200-210 | 182-178 | 50.6% | 0.44 | | 210-220 | 548-532 | 50.7% | 0.38 | | 220-230 | 685-662 | 50.9% | 0.35 | | 230+ | 352-338 | 51.0% | 0.36 |

Conclusion: Slight lean toward unders in high-total games, but not statistically significant.

Part 4: Advanced Efficiency Tests

Closing Line Value Analysis

CLV measures whether you can consistently get better lines than the close:

Methodology: 1. Simulate betting at opening lines 2. Track how often opening line beats closing line 3. Measure implied edge

Results: - Opening lines beaten by close: 52.3% of the time - Average line movement: 0.8 points toward "correct" side - Implication: Market becomes more efficient as game approaches

Predictive Model Comparison

Comparing a sophisticated model to closing lines:

Model specifications: - Elo + efficiency differentials + situational factors - Trained on 3 seasons, tested on 2

Results: | Metric | Model | Closing Line | |--------|-------|--------------| | Accuracy | 66.4% | 67.1% | | Brier Score | 0.214 | 0.209 | | ATS vs Close | 50.2% | - | | ATS vs Open | 51.1% | - |

Conclusion: Sophisticated models can match but not beat closing line accuracy.

Information Incorporation Speed

Testing how quickly markets incorporate new information:

Injury Announcement Study: 1. Tracked 50 significant injury announcements 2. Measured line movement before/after announcement 3. Analyzed time to full adjustment

Results: - Average movement on injury news: 2.8 points - Time to 90% adjustment: 8 minutes - Early bettors captured 1.2 points average CLV

Implication: Information is incorporated extremely quickly, but not instantaneously.

Part 5: Why Markets Are Efficient

The Efficient Ecosystem

  1. Professional bettors: Sophisticated models, quick execution
  2. Sportsbooks: Adjust lines based on betting patterns
  3. Competition: Multiple books create arbitrage opportunities
  4. Information flow: Injury reports, lineup news spread instantly

The Vig Problem

Even with a slight edge, the vig makes consistent profit difficult:

Example: - True ATS edge: 52% - Vig: -110 (52.4% break-even) - Expected value: 52% - 52.4% = -0.4% per bet

A 52% edge is not enough to overcome standard vig.

What Would Be Needed

To be profitable at -110: - 53% ATS: +0.6% ROI - 54% ATS: +1.6% ROI - 55% ATS: +2.6% ROI

Over 1000 bets at $100, a 54% bettor expects $1,600 profit - but variance is high: - Standard deviation: ~$1,500 - 95% CI: [-$1,400, +$4,600]

Part 6: Potential Edge Sources

Where Edges Might Exist

  1. Speed: Acting on information before lines adjust
  2. Private information: Connections to teams, injury insiders
  3. Micro-markets: Player props, less liquid markets
  4. Live betting: In-game markets with more noise
  5. Promotional value: Sign-up bonuses, odds boosts

What Doesn't Work

  1. Simple systems: Home dogs, fade public, etc.
  2. Historical patterns: Markets have adapted
  3. Public models: If you can build it, so can sharps
  4. Past results: No persistence in ATS performance

Part 7: Practical Implications

For Analysts

  1. Use closing lines as the benchmark
  2. Don't expect to beat efficient markets consistently
  3. Focus on prediction quality, not betting profits
  4. Markets are a valuable information source

For Recreational Bettors

  1. Expect negative expected value
  2. Treat betting as entertainment cost
  3. Don't chase losses
  4. Use bankroll management

For Researchers

  1. Markets aggregate information efficiently
  2. Closing lines are strong predictive baselines
  3. Inefficiencies are small and fleeting
  4. Market prices contain information about true probabilities

Conclusion

This analysis confirms that NBA betting markets are highly efficient. Closing lines are well-calibrated, simple systems don't generate consistent edges, and sophisticated models struggle to outperform market consensus. While small inefficiencies may exist - particularly for those with speed advantages or private information - the average analyst or bettor should not expect to profit consistently.

The efficiency of betting markets makes them valuable tools for basketball analysis. Closing spreads represent the market's best estimate of true team strength differentials, and implied probabilities are well-calibrated estimates of win likelihood. Rather than trying to beat the market, analysts can use market prices as inputs to their own models and benchmarks for evaluation.

Discussion Questions

  1. If markets are efficient, why do professional sports bettors exist?

  2. How might the legalization of sports betting in more states affect market efficiency?

  3. What ethical considerations exist when analyzing betting market efficiency?

  4. How could teams potentially use betting market information in their own analysis?

  5. Why might player prop markets be less efficient than game spreads?