OPTIQALOPTIQAL
Back to NBA Models
NBA Totals

NBA Totals Model

Game total points predictions using pace-adjusted team efficiency metrics, contextual factors, and statistical edge detection.

Methodology

The NBA Totals model projects expected total points by analyzing offensive and defensive efficiency ratings normalized per 100 possessions. This pace-adjusted approach allows accurate comparison across different tempo matchups.

// Core Projection Formula
Expected_Total = (Team_A_Points + Team_B_Points) × Pace_Factor
Where:
Team_A_Points = f(A.ORtg, B.DRtg, Context)
Team_B_Points = f(B.ORtg, A.DRtg, Context)
Pace_Factor = Expected_Possessions / 100

Efficiency Metrics

  • • ORtg: Points per 100 possessions (offense)
  • • DRtg: Points allowed per 100 possessions
  • • League Avg: ~114 ORtg/DRtg
  • • Std Dev: ~11 points per game

Confidence Tiers

  • • MAX (2.0u): Edge ≥ 1.5 SD
  • • STRONG (1.5u): Edge ≥ 1.0 SD
  • • STANDARD (1.0u): Edge ≥ 0.7 SD
  • • Odds floor: -200 maximum

Key Factors

Pace & Tempo

Expected possessions per game based on each team's historical pace. Fast teams (100+ pace) create higher variance totals.

Rest Differential

Back-to-back games reduce pace by 3-4 points. Rest advantage is a significant factor in total projections.

Home Court

Home teams average 2-3 points advantage. This affects both team totals and overall pace expectations.

Injury Impact

Star player absences affect pace and efficiency. Adjustments based on usage rate and historical team data.

Frequently Asked Questions

A team averaging 115 points at 105 possessions is fundamentally different from one scoring 115 at 95 possessions. Pace normalization (per-100 possessions) lets us compare true offensive and defensive efficiency independent of tempo, which is crucial for accurate total projections.
The model uses a multi-step process: (1) Calculate each team's offensive rating (ORtg) and defensive rating (DRtg) per 100 possessions, (2) Estimate expected points for each team based on matchup (Team A ORtg vs Team B DRtg), (3) Estimate expected pace for the game, (4) Multiply efficiency by pace to get raw totals, (5) Apply contextual adjustments for rest, injuries, and venue.
Back-to-back games typically reduce a team's pace by 3-4 points. We model rest advantage/disadvantage based on historical performance differentials. A rested team facing a B2B opponent has a measurable edge that affects both pace and efficiency.
The model incorporates home court advantage (typically 2-3 points), rest days differential, recent form (last 10 games weighted), head-to-head history, and travel factors. Injury impacts are assessed based on the missing player's usage rate and team historical data without them.
Edge represents the difference between our projected total and the book's line, measured in standard deviations. An edge of 1.0 SD means our projection is one standard deviation away from the book line. Higher absolute edge values indicate stronger disagreement with the market.
Beyond -200, the risk/reward structure becomes unfavorable. You're risking $200 to win $100, requiring extremely high precision. Our model maintains profitability by avoiding heavy juice situations where the vig erodes expected value.
Confidence tiers (MAX, STRONG, STANDARD) are based on a combination of: expected value (EV), edge percentage, and standard deviation distance. MAX tier (2.0u) requires the highest thresholds across all three metrics. Only picks that pass minimum thresholds are surfaced.
Picks are generated and locked daily at 10:00 AM ET. This timing ensures we have the latest odds data and injury reports while giving you time to act before lines move. The model limits output to the top picks by edge strength each day.
CLV measures how much the line moved toward your position after you locked. If you took Over 224.5 and it closed at 226, you captured 1.5 points of CLV. It's the strongest predictor of long-term profitability and validates model accuracy.
Games with high pace variance (both teams play fast or have erratic tempos) have wider standard deviations. The model accounts for this by requiring larger edges in high-variance matchups to maintain the same confidence tier.