Motorsports Research Pipeline — Council Ruling

Date: 2026-04-01 Process: Full 5-phase council (Advisory → Anonymization → Peer Review → Chairman Synthesis → Boss Ruling) Advisors: Opus, Sonnet, Gemini 3.1 Pro, Grok 4.20 Reasoning, gpt-oss-120b Winner: Sonnet (2 of 5 peer review votes — tied with Gemini, Opus broke tie) Status: PENDING BOSS RULING on open questions

COUNCIL SUMMARY

Where Advisors Agreed

1. Equipment inequality is THE fundamental difference — F1 car is ~80% of the equation, NASCAR flips this
2. Qualifying is dominant predictor in F1 — 60-70% predictive, pole wins ~40% of races
3. DNF risk (15-20%) is systematically mispriced — markets focus on speed, not reliability
4. Safety car reshuffles field — 70%+ of F1 races have SC, must model probability per circuit
5. 4 separate series need distinct models — F1, NASCAR, IndyCar, MotoGP have fundamentally different dynamics
6. 7+ edge scanners required — Outright, Podium/Top, H2H, Qualifying/Pole, Fastest Lap, Championship, Sprint
7. Practice session long-run data is primary intelligence input before qualifying
8. Weather amplification — rain completely reorders competitive hierarchy
9. Monte Carlo race simulation (100K iterations) for outright/podium/top markets
10. Track type specialization in NASCAR — superspeedway vs short track vs road course require separate models

Where Advisors Disagreed

1. Teammate H2H vs cross-team H2H: Sonnet identified these have fundamentally different variance structures. Others treated all H2H the same. Council verdict: Separate teammate H2H model with team order probability.
2. NASCAR championship modeling: Sonnet noted playoff bracket makes outright market bimodal, not Gaussian. Others used standard MC. Council verdict: Multimodal championship simulation for NASCAR playoffs.
3. Superspeedway modelability: Opus honestly stated superspeedways are "nearly unmodelable" due to drafting pack dynamics. Others tried to model them normally. Council verdict: Reduce edge thresholds and position sizing for superspeedways.
4. Architecture complexity: gpt-oss proposed enterprise stack (Kafka, microservices). Council verdict: SQLite WAL, simple pipeline.

Strongest Arguments (from peer review)

Sonnet wins with the deepest analytical understanding:

• "Four hard truths" of motorsports betting — car dominance, qualifying predictiveness, DNF risk, safety car randomness
• Teammate H2H has different variance structure than cross-team H2H
• NASCAR playoff bracket makes championship outright bimodal/multimodal
• "Free pit window" mechanic as key driver of fastest lap market
• Narrative contamination risk (teams misleading in press conferences)
• Dominant team problem (edge lives in pricing rest of field, not favorite)
• Regime detection for mid-season dominance shifts
• Series-specific modeling throughout (not one-size-fits-all)

Biggest Blind Spot

No one addresses liquidity and bet placement reality — Motorsports markets are thin. Pinnacle limits are low. If you find a 5% edge on NASCAR H2H, max bet might be $200. Line movement speed means 10-minute delays kill edges. No advisor modeled expected dollar value factoring in liquidity, size limits, and account limiting risk.

What Everyone Missed (from peer reviews)

1. Team orders as correlated outcome manipulation — In F1, teams instruct Driver B to let Driver A pass. In NASCAR, manufacturer alliances coordinate drafting. This directly manipulates H2H outcomes and violates independence assumptions in every MC simulation. Need "team directive model" estimating probability/direction of team orders.
2. H2H void risk — Driver DNFs void the H2H bet. Need to adjust conditional probability math for void probability.
3. Constructor development curves — Car performance changes throughout season with upgrades. Need to model mid-season development trajectory, not just static car ratings.
4. Pit crew performance variance — Pit stop execution varies 1-3 seconds per stop, directly affects H2H and position outcomes.
5. Tire allocation strategy — Different tire compound selections affect race strategy trees.

BUILD PLAN

Phase 1: Core Motorsports Data Tables

motorsport_drivers: driver_id, full_name, series (F1/NASCAR_Cup/NASCAR_Xfinity/IndyCar/MotoGP), team_id, nationality, car_number, contract_status, rookie (boolean), active

motorsport_teams: team_id, name, series, manufacturer, budget_tier (A/B/C for F1), car_performance_rating, reliability_rating, pit_crew_rating, updated_at

motorsport_tracks: track_id, name, country, city, track_type (street/permanent/oval/superspeedway/road_course), length_km, turns, surface, altitude_ft, sc_probability_dry, sc_probability_wet, red_flag_probability, drs_zones (F1), banking_degrees (NASCAR)

motorsport_race_results: result_id, race_id, driver_id, grid_position, finish_position, status (finished/DNF_mechanical/DNF_crash/DSQ), laps_completed, total_laps, fastest_lap, fastest_lap_time, pit_stops, tire_strategy (JSON), points_scored

motorsport_qualifying: quali_id, race_id, driver_id, session (Q1/Q2/Q3 for F1, round for NASCAR), position, time, gap_to_pole, penalties_applied, grid_penalty_reason

motorsport_practice: practice_id, race_id, session (FP1/FP2/FP3), driver_id, best_time, long_run_pace, tire_compound, laps_completed, fuel_corrected_pace

motorsport_reliability: driver_id, team_id, season, races_started, mechanical_dnfs, crash_dnfs, dnf_rate, component_usage (JSON — engines, gearboxes for F1)

motorsport_weather: race_id, session, forecast_time, temp_f, track_temp_f, wind_speed_mph, rain_probability, rain_intensity, conditions (dry/damp/wet)

motorsport_safety_car: race_id, sc_number, type (VSC/SC/red_flag), lap_deployed, lap_ended, cause, positions_affected

motorsport_championships: season, series, driver_id, points, position, races_remaining, title_probability

Phase 2: Series-Specific Models

Phase 3: Distribution Models

• Race simulation: 100K iterations per race
• Per-iteration: Simulate lap-by-lap with position changes based on pace differential, pit strategy trees, random SC/red flag events, DNF probability per lap
• SC model: Bernoulli per lap with circuit-specific probability × weather adjustment
• DNF model: Exponential hazard per driver/team with reliability rating
• Team orders model: Probability of team directive based on championship gap, contract status, race position

Phase 4: 7 Edge Scanners

Phase 5: Matchup Card Format

RACE: [Grand Prix/Race Name] | [Track] | [Series] | [Date]
TRACK: [Type] | [Length] | [Turns] | SC Prob (dry): [%] | SC Prob (wet): [%]
WEATHER: [Temp]°F | Track: [°F] | Wind: [mph] | Rain: [%]

DRIVER: [Name] | CAR: [#] | TEAM: [Name] ([Tier])
  Grid: [position] (Quali: [time] | Gap: [+sec])
  Car Performance Rating: [val/10]
  Season: [W-P-T5] in [races] starts
  Track History: [starts] starts | Best: [pos] | Avg: [pos]
  DNF Rate: [%] (Mechanical: [%] | Crash: [%])
  Practice Pace: FP1 [pos] | FP2 [pos] | FP3 [pos]
  Long Run Pace: [rank in field] on [compound]
  Tire Strategy (Predicted): [strategy]
  Championship: [points] ([position]) | Title Prob: [%]

[Repeat for field or key drivers]

TEAM DYNAMICS:
  Teammate: [Name] | Grid: [pos] | Team Order Probability: [%]
  Constructor Standings: [points] ([position])

STRATEGY NOTES:
  Likely pit windows: [lap ranges]
  DRS zones: [count] | Overtaking difficulty: [high/medium/low]

INTELLIGENCE:
  [CRITICAL/MODERATE/CONTEXT findings]

Phase 6: Dashboard

• Race board: Upcoming races with grid, weather, edge counts per market
• Driver drill-down: Full card + all market edges + season trends
• Grid analysis: Qualifying results with race prediction conversion
• Weather center: Forecast timeline, rain probability, condition changes
• SC/Red Flag tracker: Historical SC rates per circuit, live SC probability
• Championship tracker: Title probabilities updated each race
• H2H board: All H2H matchups with edges, teammate vs cross-team split
• Reliability monitor: DNF rates per team, component usage tracking
• Edge alerts: By magnitude, filterable by series and market type
• P&L tracker: By series, market type, edge bucket, Brier scores

Phase 7: Kill Switch

• Red flag: Pause all live scanners for affected session
• Race cancellation/postponement: Void all pre-race positions
• Weather delay: Re-run weather model, update SC probabilities
• Driver WD: Void that driver's H2H, re-run outright MC with N-1 field
• Superspeedway circuit breaker: Reduce max position size to 50% of normal (near-unmodelable pack dynamics)

OPEN QUESTIONS FOR BOSS RULING

1. Series scope: All 4 series (F1/NASCAR/IndyCar/MotoGP) at launch? Or F1 + NASCAR first?
2. Practice data sourcing: Live timing data requires scraping or paid feeds. Worth the cost?
3. Team order model: Build probability model for team directives? Critical for F1 H2H.
4. NASCAR playoff bimodal model: Build separate championship MC for NASCAR's unique playoff format?
5. Superspeedway modeling: Accept lower confidence/sizing, or skip superspeedway markets entirely?
6. Sprint race scanner: Worth building separate scanner for F1/MotoGP sprint format?
7. Liquidity assessment: Should pipeline estimate bet size limits before flagging edges?

COUNCIL METADATA