+EV Opportunity Tracking

Game	League	Game Start	Next Capture	Captures	Source
Loading scheduled captures...

Event Date	League	Game	Bet	Bookmaker	Initial EV%	CLV%	Time Since Start	Close Capture	CLV Source
Loading pending results...

Found	Event Date	League	Game	Bet	Bookmaker	Initial EV%	CLV%	Status	Result	Units +/-
Loading all results...

Bookmaker	Total Opps	Avg EV%	Wins	Losses	Win Rate	Profit (Units)	Avg CLV%
Click "Refresh Stats" to load bookmaker analytics...

Active Opportunities -- Live right now

Distribution by Time Horizon

By Sport

14-Day Trend

Historical Trends (from daily snapshots)

From:

To:

Select Indicator:

Latest Snapshot

Period Avg: -

Range: -

Change: -

Total Opportunities

High EV (>5%)

Avg EV%

Avg Opportunities/Day

Active Sources

Unique Leagues

Market Mix

0 1X2 0 ML 0 DNB 0 AH 0 Totals 0 Consensus Totals 0 Player Props

System Health & Flags Loading... Current issues only

Loading health indicators...

⚠️ Warning History

Detected	Warning	Details	Affected	Type	Status
Loading warning history...

🔧 Service Reliability Loading...

Loading service health...

Error History (Last 7 Days)

Date	Service	Error Pattern	Count	Severity	Status
Loading error history...

⏰ Automated Agents

🔍 Data Quality Agent

Investigates data anomalies, validates fairlines, checks for cross-game contamination

Schedule: Every 2h at :00

Last run: -

Next run in --:--:--

🔧 SRE Monitoring Agent

Monitors Render services, detects errors/outages, tracks service health patterns

Schedule: Every 2h at :30

Last run: -

Next run in --:--:--

Agents run via macOS launchd every 2 hours. Reports kept for 30 days.

Result Settlement Loading...

Overall Settlement

Soccer

Basketball

Ice Hockey

Missing Results

Daily Settlement Rate (Last 7 Days)

Date	Total	Settled	Rate	Worst Sport
Loading...

Settlement = past games with bet_result populated. Missing results are caused by TheSportsDB API 429 rate limits. Games must be 3+ hours past start time.

Result Linking Pipeline

Game Results (7d)

Linked to Opps

Unlinked Opps

Link Rate

League (unsettled)	Pending	Blocker
Loading...

Linking matches ev_opportunities to game_results by sport + team names + time window. Failures are usually team name mismatches or missing results.

Fairline Calibration by Sport

Sport	Settled	Expected Win%	Actual Win%	Calibration Error	Status
Loading...

Calibration error = Actual Win% - Expected Win%. Negative means underperforming. Values beyond ±3% indicate potential fairline issues.

🤖 Data Quality Investigation Agent Loading...

Autonomous agent (powered by Claude Code) that investigates every indicator, examines every high EV opportunity, and produces detailed reports. Run manually: ./scripts/run_dq_agent.sh

Loading latest report...

📋 Report History

Time	Status	Type
Loading...

📊 Fairline Accuracy Fairline AccuracyMeasures how well fairlines predict outcomes. Compares actual win rates vs expected rates based on fairline probabilities. "Accurate" means within ±3% deviation. Loading...

Validates fairline quality by comparing different sources and detecting anomalies. Healthy fairlines are essential for accurate EV calculations.

Fairline Source Comparison

Source	Count	Avg EV%	Win Rate	Accuracy	Alt EV%	EV Δ
Loading...

Accuracy = Win Rate vs Expected. Alt EV% = avg EV if the other fairline source had been used. EV Δ = difference (positive = current source is more conservative).

Pinnacle vs Consensus Agreement

Sport	Both Present	Avg Diff	High Diff (>5%)	Agreement
Loading...

High disagreement between Pinnacle and Consensus may indicate data quality issues.

🔍 Fairline Anomalies Detected

Anomaly Type	Count	Last 24h	Severity
Loading...

Fairline Coverage by Sport

Sport	Events	Has Pinnacle	Has TFL	Consensus Only	No Fairline	Coverage
Loading...

Confidence & Data Quality

Confidence Score Distribution

Band	Count	%	Bar
Loading...

Data Quality Flag Distribution

Flag	Count	%
Loading...

Cross-Platform Matching Health

Cross-Platform Rate (Last 7 Days)

Opportunities with data from 2+ platforms

Date	Opportunities	Multi-Platform	Rate	Status
Loading...

Data Source Flow (Last 7 Days)

Opportunities by contributing platform (one opp can count for multiple platforms)

Date	OddsJam	TheOddsAPI	BetBurger	Status
Loading...

Data older than 15 minutes is excluded from opportunity detection. Low freshness = missing from cross-platform merging.

Platform	Fresh (<15min) Fresh RecordsRecords where `updated_at` is within last 15 minutes. These are eligible for +EV opportunity detection.	Stale (>60min) Stale RecordsRecords not updated in 60+ minutes. Excluded from processing - may indicate scraper issues or low-volume leagues.	Fresh % Freshness RateFormula: `fresh / total × 100` Target: ≥70% for Good status. Low rates mean opportunities may be missed due to stale data.	Status
Loading...

🔀 Platform Divergence Detection 0

Opportunities where median odds differ >20% between data sources. This may indicate mapping issues, stale data, or mixed market types.

Divergence by Sport

Sport	Divergent	Total	Rate	Status
Loading...

Recent Divergent Opportunities

Event	Sport	EV%	Platforms	Detected
Loading...

🛡️ Market Data Validation 0

Markets rejected from fairline calculation due to invalid data. High rejection rates are expected for Soccer (2-way markets without draw) and incomplete SSE data. This prevents false +EV opportunities.

Individual bookmaker lines

Daily Rejection Trend (Last 7 Days)

Date	Rejections	Opportunities	Total	Rejection Rate
Loading...

Rejections by Reason (24h)

Reason	Count	% of Total	Severity
Loading...

Rejections by Sportsbook (24h)

Sportsbook	Total	Invalid Overround	Both Positive	Extreme Prob	Missing Odds
Loading...

Liquidity Governance Loading... PRD v1.02 Appendix C

Actionable Rate

Suppressed

Avg Confidence (Exch. Included)

LIQUID / THIN exchanges

Avg Confidence (Exch. Excluded)

ILLIQUID / UNKNOWN exchanges

Actionable Rate Trend

Suppression Reasons

Why opportunities are non-actionable

Reason	Count	%
Loading...

Exchange Liquidity by Sport

Classification of exchange volume (LIQUID / THIN / ILLIQUID)

Sport	Liquid	Thin	Illiquid	Unknown
Loading...

League Breakdown (Upcoming Games)

League ↕	Sport	Events ↕	Odds Records ↕	Fairline Cov.	Opps (24h) ↕
Loading...

Mapping Issues

Issue Type	Details	Count	Severity	Suggested Action
Loading...

Daily Opportunities Found

Opportunities by EV Threshold

Platform Coverage Analysis

Coverage by # of Platforms

Platform Combinations

Combination	Count	Avg EV%	%
Loading...

Platform Coverage Over Time

Bookmaker Performance

Top Bookmakers by Opportunities

Bookmaker Statistics

Bookmaker	Opps	Avg EV	≥5% EV	≥10% EV
Loading...

Top 5 Bookmakers Trend (Opportunities per Day)

Top Performing Leagues

Opportunities by League (Top 10)

League Statistics

League	Opps	Avg EV	Win Rate	ROI
Loading...

Top 5 Leagues Trend (Opportunities per Day)

Daily Breakdown

Date	Total	≥1% EV	≥3% EV	≥5% EV	≥10% EV	Avg EV	Leagues	Bookmakers
Loading...

Performance by EV Range

EV Range	Count	Win Rate	Avg CLV	ROI	Total P&L
Loading performance data...

CLV Performance

CLV Range	Count	Win Rate	Avg EV	ROI	Total P&L
Loading CLV data...

📊 League Coverage & Cross-Platform Analysis

Leagues

Total Events

Cross-Platform %

Active Opps

Mapping Gaps

Events vs Yesterday

	League	Platforms	Events	Days Ahead	Cross-Platform	Opportunities	Opp Density	Health
Loading league coverage data...

OddsJam TheOddsAPI BetBurger | 🟢 Healthy (2+ platforms, fresh) 🟡 Warning (1 platform or stale) 🔴 Critical (no data) ⚠️ Mapping gaps | ↑ More vs yesterday ↓ Less vs yesterday

🗺️ Team Mapping Health

Match Rate

-- mapped / -- seen

Unmapped Teams

blocking event matching

Leagues with Gaps

Stale Mappings

not used in 90+ days

⚠️ Top Unmapped Teams (Last 7 Days)

Team	League	Platform	#
Loading...

📊 League Coverage Gaps (Worst First)

League	Platform	Coverage	%
Loading...

Pinnacle Odds Divergence Monitor

Multi-Source Events

Avg Divergence

Max Divergence

PIN/TOA Exact Match

OJ Mkt Mismatch

PIN Direct Events

PIN Direct

TheOddsAPI w/ PIN

OddsJam w/ PIN

Event	Sport	PIN H/A	TOA H/A	OJ H/A	Max Div%	Src
Loading...

Compares Pinnacle odds from PIN direct, TheOddsAPI, and OddsJam. PIN/TOA should be ~0% (same CDN). OJ soccer uses draw_no_bet (not comparable to h2h).

Spider Run Health (Last 24h)

Run Status Distribution

Current Cycle

Cycle Duration Trend

Scraping Freshness by Platform

Platform	Last Scraped	Age	Status
Loading...

Stale Fairline Backlog

Stale Events

Total Upcoming

Stale Rate

Recent Spider Runs

Platform	Started	Duration	Status
Loading...

League	Total Opps	With Results	Without Results	Coverage %	Result Sources	Mapping Status	Overall Status	Action
Loading results tracker data...

Opportunity League Name	Possible ESPN Match	Similarity	Action
Checking for mismatches...

Source	Type	Configured Mappings	Unique Leagues	Results Obtained	Result Leagues	Coverage %	Status
Loading...

Indicators Reference

Expected Value (EV)

Raw EV%	`(decimal_odds × fair_probability - 1) × 100` Edge before any risk adjustment. Source: Pinnacle/Consensus fairline vs bookmaker odds.
Adjusted EV%	`raw_ev × haircut_factor` Risk-adjusted EV using empirical calibration. Haircut varies by segment (sport × market_type × market_state). Source: `segment_calibration` table, updated daily.
Current Best EV	Highest EV across all recreational bookmakers for the selected devig method. Excludes sharps, exchanges, prediction markets, and cross-validation flagged bookmakers. Refreshed every 30s.

Fairlines

Pinnacle Fairline	Devigged probabilities from Pinnacle's odds using the selected method (additive, multiplicative, power, shin, worst_case). Pinnacle is the primary sharp bookmaker — their prices reflect the most efficient market.
Consensus Fairline	Weighted median of devigged probabilities from all non-Pinnacle sharp bookmakers (CEP = Consensus Excluding Pinnacle). Only available when 2+ sharps have odds. Used as cross-validation, not primary reference.
Devig Methods	Additive: subtract equal margin from each outcome. Multiplicative: scale proportionally. Power: favorite-longshot bias correction. Shin: insider trading model. Worst Case: maximum margin removal (most conservative).

Market State

Classification of the opportunity's data quality structure. Drives haircut factors and EV thresholds.

STRONG SHARP	3+ sharp bookmakers, native market. Best data quality.
MULTI SHARP	2 sharp bookmakers, native market. Good independent price validation.
THIN SHARP	1 sharp bookmaker (typically Pinnacle only). Higher variance, still actionable.
DERIVED STRONG	Derived market (DNB, double chance) with 2+ sharps. Fairline derived from 3-way odds.
DERIVED THIN	Derived market with 1 sharp. Weakest structure — requires high EV threshold (5%+).

Note: Exchanges (Betfair, Matchbook) are NOT counted as sharp bookmakers. They provide an independent price signal via the Exchange Agreement indicator.

Confidence Score (0-100)

Penalty-based model starting at 100. Measures signal reliability, NOT edge magnitude. Recomputed every 30s refresh.

Sharp Depth	0 to -60	Non-Pinnacle sharp bookmakers contributing to the fairline for this specific line.
Sharp Diversity	0 to -25	Number of independent corporate pricing groups. 1 group = all sharp data from same entity.
Dispersion	0 to -30	Max probability spread among CEP sharp books. High dispersion = sharps disagree on fair price.
Market Derivation	0 to -30	`-30 / sharp_count` for derived markets (DNB, double chance). 3 sharps = -10, 1 sharp = -30.
Exchange Liquidity	0 to -10	Penalty when an exchange in the CEP has THIN liquidity classification.
Exchange Agreement	-5 to +5	Comparison of exchange devigged probability vs Pinnacle. Within 2% = confirms (+5), >5% = contradicts (-5). Source: Betfair, Matchbook.
Prediction Market	-5 to +8	Agreement between prediction market (Polymarket/Kalshi) and sharp fairline. Only when liquidity ≥ $1K. Weight scales with liquidity ($1K = ~1%, $100K+ = full). Source: Polymarket, Kalshi.
Prediction Mkt Penalty	0 to -10	Applied when the opportunity bookmaker IS a prediction market. Different market dynamics.
Data Quality	0 to -50	Penalties for: impossible_ev (>100%), low_sharp_dnb, odds/fairline mismatch, cross-market inconsistency.
Market Consensus Density	-15 to +10	Other recreational bookmakers within 5% of opportunity odds (excluding self). 5+ = +10 (strong confirmation), 0 = -15 (isolated price).
Hold Comparison	0 to -15	Bookmaker overround vs Pinnacle overround. Large difference suggests aggressive margin extraction.
Longshot Risk	0 to -20	Penalty for low-probability outcomes (fair prob < 15%). Small errors in probability estimation create disproportionate EV on longshots.
Soft Tail Inflation	0 to -15	Detects when recreational bookmakers systematically price longshots higher than sharps. Only for odds > 3.0 decimal.
Bookmaker Volatility	DB-driven	Per-bookmaker penalty for known volatile/unreliable odds providers (e.g., GGBet = -15). Stored in `canonical_bookmakers.confidence_penalty`.

Market Agreement & Divergence

Phase 1 scoring signals that measure structural support for the edge. Computed per-bookmaker during EV detection.

Market Agreement	Sharp-weighted cluster analysis. Measures how many sharp bookmakers support the fairline side. High agreement = many sharps agree = efficient market. Score 0–1.
Market Divergence (v3)	Z-score based. Market-specific classification: Moneyline/AH/Totals (monotonic): higher z = better. Bands: WEAK (<1.5), MODEST (1.5–2), GOOD (2–3), STRONG (3+). 3-way/DNB (non-monotonic): sweet spot at z 2–3. Bands: WEAK (<1.5), STRONG (1.5–3), EXTREME_CAUTION (3+).
Divergence Multiplier	Applied to per-bookmaker EV based on divergence band. STRONG/GOOD = 1.0x, MODEST = 0.9x, WEAK = 0.75x, EXTREME_CAUTION = 0.5x.

Opportunity Classification

Edge classification based on market agreement, divergence, and sharp book count. Data-calibrated 2026-03-26 — empirical CLV/ROI per class shown.

HIGH_VALUE	BET	EV ≥ 15% at any confidence level. +22.9% ROI across 732 settled bets. Large detected edge — profitable regardless of confidence score.
PRIME	BET	Confidence ≥ 80, EV ≥ 2%. +3-4% ROI, best risk-adjusted segment. High fairline quality + meaningful edge = most reliable opportunities.
STANDARD	BET	Confidence 50-79, EV ≥ 2%. +1-3% ROI. Solid fairline quality with decent edge.
VALUE	BET	Confidence < 50, EV ≥ 2%. +1% ROI, profitable but volatile. Lower confidence (thin fairline source) but still positive expectation from settlement data.
THIN_EV	PASS	EV < 2%. -1.83% ROI across 5,434 bets. Edge too thin to survive vig and line movement. Only consistently losing segment.
NO_EDGE	PASS	0 sharp books or negative/zero EV. No reliable fairline to evaluate against. -14.4% ROI when 0 sharps.

Key insight: the classifier measures signal quality, not edge quality. High signal quality (STRONG) often means efficient markets with small edges. Low signal quality (THIN, FALSE) can still be profitable. Edge Score in the Combo Finder normalizes for this.

Edge Score (0–100) — Combo Finder

Composite score for comparing combos across all dimensions. Normalizes for odds tier inflation so a +5% CLV favorites combo and a +15% CLV longshots combo are comparable. Used in the Combo Finder (ROI Statistics tab).

Validation (35 pts)	Historical backtest agreement + ROI. Largest component — a combo without historical validation can never rank highest. Validated + hist ROI > 0 = 35. Validated + hist ROI ≤ 0 = 20. Directional + hist ROI > 0 = 25. Directional + hist ROI ≤ 0 = 10. Low Sample + hist ROI > 0 = 15. No Data = 8. Disagree = 0–5. Both Negative = 0.
Consistency (30 pts)	`+CLV% × 30`. How often you beat the closing line. A combo at 80% +CLV scores 24/30.
Edge Quality (20 pts)	`min(CLV / EV, 2.0) / 2.0 × 20`. Did the detected edge hold at close? CLV/EV ratio normalizes across odds tiers. Ratio of 1.0 = edge held perfectly. Capped at 2.0 to prevent outlier inflation.
Sample Size (15 pts)	`min(ln(n) / ln(200), 1.0) × 15`. Log-scaled, maxes out at n=200. n=20 → 8.5 pts, n=50 → 11.1, n=100 → 13.0, n=200 → 15.0. A 6.5-point spread between smallest and largest samples.

Color coding: 80+ strong edge, 60–79 promising, 40–59 weak, <40 avoid.

Combo Validation Badges

Historical backtest (585K simulated opps from 6-month TheOddsAPI data) compared to live results. Backtest CLV is from sparse hourly snapshots — directional guide only, not precise.

Validated	Same CLV direction (both positive) + historical +CLV% ≥ 50% + gap within 10x. Strongest signal that the edge is real and persistent.
Directional	Same direction but weak: historical +CLV% < 50% or gap > 10x. Edge might be real but historical data doesn't strongly confirm it.
Disagree	Opposite directions — live says profitable, historical says not (or vice versa). Treat with caution.
Both Neg	Both live and historical CLV are negative. Strong signal to avoid this combo.
Low Sample	Historical match exists but below the Backtest min-bets threshold. Not enough historical data to validate.
No Hist	No matching historical combo at all. Common for markets/sports not covered in the 6-month TheOddsAPI dataset.

Fairline Source Tiers

Fairlines are calculated using a priority cascade. Higher tiers are preferred when available.

Tier 1: Pinnacle	Devigged Pinnacle odds. Most accurate fairline. Confidence cap: 100.
Tier 2: Consensus EV	Weighted median of 2+ non-Pinnacle sharp bookmakers. Used when Pinnacle has no odds for the event. Confidence cap: 100.
Tier 3: Betfair Exchange	Devigged Betfair odds when liquidity is LIQUID or UNKNOWN. Independent price from exchange orderbook. Confidence cap: 55.
Tier 4: Robinhood	Prediction market probabilities from Robinhood. Different market dynamics — use as last resort. Confidence cap: 45.
No Source	No sharp/exchange/prediction market data available. Event cannot be priced — no opportunities generated.

Coverage improvement: Pinnacle alone covers ~40% of events. Adding Betfair (+15%) and Robinhood (+3%) brings total coverage to ~56%.

Stake Recommendation

Raw Kelly%	`(b × p - q) / b` where b = decimal_odds - 1, p = fair_prob, q = 1 - p. Optimal bankroll fraction assuming the fairline is exact.
Adjusted Kelly%	`raw_kelly × (confidence / 100)`. Scales stake by model trust. Confidence 0 = no bet regardless of raw Kelly.
Recommendation	No Bet: adj. kelly ≤ 0. Minimum: < 0.5%. Small: < 2%. Standard: < 5%. Aggressive: ≥ 5%.

Ranking Score (0–100) — Per-Opportunity

Composite score for real-time opportunity prioritization in the Active Opportunities table. Different from Edge Score (which is for combos/aggregates).

Adjusted EV (40%)	0–40 points. 20% adjusted EV = max 40 points.
Confidence (30%)	0–30 points. Confidence score × 0.3.
Adjusted Kelly (20%)	0–20 points. 5% adjusted kelly = max 20 points.
Market Quality (10%)	STRONG_SHARP=10, MULTI=8, THIN=5, DERIVED_STRONG=4, DERIVED_THIN=1.

Calibration System

Haircut Factor	`avg_clv / avg_raw_ev` per segment. What fraction of displayed edge historically materializes. Floored at 0.10, capped at 1.00. Segments: sport × market_type × market_state. Updated daily at 6 AM UTC.
Min EV Threshold	Lowest EV band where positive CLV rate ≥ 50%. Per segment. Prevents detection of opportunities in segments where low EV historically loses money.
Bayesian Shrinkage	`(segment_val × n + parent_val × k) / (n + k)`, k=50. Small segments blend toward parent (market_state level) to prevent overfitting. 30+ samples required for segment-specific calibration.
Close Quality Score	0-1 scale. `time_score (0-0.5)`: proximity to start (T-0 = 0.5, T-10h = 0). `sharp_score (0-0.5)`: sharp books at close (3+ = 0.5). Higher quality CLV data is preferred in calibration.

Closing Line Value (CLV)

CLV%	`(closing_fairline - initial_implied_prob) / initial_implied_prob × 100`. Measures whether the initial bet had edge relative to the closing price. Positive CLV = the line moved in your favor. Visible in Results & Analysis tab.
Edge Decay	`initial_ev - closing_ev`. How much edge eroded between detection and close. High decay = the market corrected quickly.

Data Quality Signals

Price Outlier	Bookmaker odds > 2 standard deviations from the market cluster mean. Warning shown in modal when the opportunity bookmaker is the outlier.
Cross-Validation	For parameterized markets (AH, spread): detects when a bookmaker's odds appear identically under a different market type or param on the same event. Indicates platform labeling error. Flagged bookmakers excluded from Best EV.
Platform Divergence	Median odds differ by > 20% between platforms for the same event. Indicates stale data or market type mixing. Opportunity suppressed.
Impossible EV	EV > 100%. Indicates mapping error or data corruption. Opportunity suppressed.

Bookmaker Categories

Recreational	Standard sportsbooks (DraftKings, FanDuel, bet365, etc.). These are the actionable bookmakers — opportunities are detected against their odds. Shown in main table.
★ Sharp	Price-making bookmakers (Pinnacle, Marathon, Circa, etc.). Their devigged odds form the fairline. Not actionable — shown as reference in modal only.
↔ Exchange	Betting exchanges (Betfair, Matchbook). Provide independent price discovery. Not counted as sharp books. Exchange Agreement signal in confidence. Shown as reference in modal.
Prediction Mkt	Belief-based markets (Polymarket, Kalshi). Different probability semantics — not devigged. Provide Prediction Market Agreement signal when liquidity ≥ $1K. Never actionable.

Data Sources

OddsJam REST	Batch odds fetch every 20 minutes. Covers all major sportsbooks + prediction markets. Source of most bookmaker odds.
OddsJam SSE	Real-time odds stream. Updates existing records with live odds changes. Low latency.
TheOddsAPI	Scrapy spider, priority-tiered schedule. Primary source for Pinnacle odds and European bookmakers.
Prediction Mkt Liquidity	USD liquidity per outcome from Kalshi (weight 1.0) and Polymarket (weight 0.85). Used for the $1K liquidity gate on prediction signals.

Per-opportunity indicators recompute every 30-second refresh cycle. Calibration updates daily at 6 AM UTC. Edge Score and combo validation update on-demand via Combo Finder. Classification recalibrated 2026-03-26 based on 15K+ live opportunities. Formulas reference PRD v2.2.1, CLV Calibration PRD v1.0, and Quality Scoring PRD v1.0.

📊 +EV Opportunity Tracking 🔄 Refresh

⏱ Scheduled Close Captures (0)

Bookmaker Performance Analytics