Loading match analysis
Pulling the odds, probabilities and the full engine breakdown for this fixture.
Pulling the odds, probabilities and the full engine breakdown for this fixture.
Here's everything the engine sees for this match — odds, probabilities, value, and risk.
“Value edge” is our probability minus the bookmaker's fair probability (margin removed). Positive edge means the price is better than it should be. Current best edge: none found.
| Market | Selection | Odds | Bookmaker | Model | Raw implied | No-vig implied | Margin | No-vig edge | EV |
|---|---|---|---|---|---|---|---|---|---|
| Match winner | Barcelona | 1.70 | Market price | 54% | 59% | 56% | +4.5% | -2.7% | -8.8% |
| Match winner | Draw | 3.90 | Market price | 23% | 26% | 25% | +4.5% | -2.0% | -12.1% |
| Match winner | Sevilla | 5.00 | Market price | 24% | 20% | 19% | +4.5% | +4.7% | +19.1% |
| Goals over/under 2.5 | Over 2.5 Goals | 1.98 | Market price | 62% | 51% | 51% | +0.0% | +12.0% | +23.7% |
| Goals over/under 2.5 | Under 2.5 Goals | 2.02 | Market price | 38% | 50% | 49% | +0.0% | -12.0% | -24.2% |
| Both teams to score | BTTS Yes | 2.06 | Market price | 63% | 49% | 49% | -1.7% | +13.2% | +28.9% |
| Both teams to score | BTTS No | 2.01 | Market price | 37% | 50% | 51% | -1.7% | -13.2% | -24.8% |
OddsPadi estimates Barcelona at 54%, the draw at 23%, and Sevilla at 24%. The current prices do not show a clear positive value edge, so the responsible call is to avoid forcing a pick.
Sports outcomes are uncertain. Predictions are model estimates, not guarantees.
This is the engine's complete working: every check, every doubt, every guardrail. Perfect if you like to see the maths behind the call.
decision-engine-v1 - rules and model reasoning
Decision engine is monitoring BTTS Yes: model 63%, no-vig implied 49%, edge +13.2%, EV +28.9%, fair odds 1.60.
Watchlist thesis requires more evidence before trust: BTTS Yes.
Decision engine is monitoring BTTS Yes: model 63%, no-vig implied 49%, edge +13.2%, EV +28.9%, fair odds 1.60. The research brief says keep on watchlist until required checks refresh odds, context, and review-loop warnings.
Research checks: Refresh bookmaker odds: Recalculate implied probability, bookmaker margin, no-vig probability, value edge, and expected value. Weather check: Refresh the model, rerun the no-vig EV calculation, and re-check abstention gates. Historical learning profile: Import real fixtures and odds, run backtests, then activate learned thresholds only after the corpus is large enough. Similar stored decisions: Keep memory neutral while continuing to store and settle outcomes. Do not raise confidence from training until real historical data is active.
Notebook needs review: 10 operator check(s) remain before trusting BTTS Yes.
BTTS Yes is currently modeled at 63% with +13.2% no-vig edge.
Action: Rerun the sport model after any lineup, injury, weather, odds, or live-state update.
Quoted odds 2.06 imply EV +28.9% after margin removal.
Action: Refresh bookmaker odds and recompute raw implied probability, no-vig probability, edge, and EV.
Open context gaps: Weather check.
Action: Fetch the highest-priority missing provider signals before keeping the same action.
Remove the thesis if refreshed edge or EV is no longer positive; current fair odds 1.60.
Action: Downgrade to avoid when no-vig edge or EV falls to zero or below.
Stress applies unresolved context gaps: Weather check.
Action: Apply the provider context update and rerun actionability before showing the pick.
Belief is strong: BTTS Yes at 63% with +13.2% edge, +28.9% EV, uncertainty 8/100, expires in 10 minutes.
Action: Hide or downgrade the recommendation if the belief expires without a fresh decision run.
Operator checklist: Refresh bookmaker odds: Complete this check, then rerun the decision engine before trusting the current posture. Weather check: Complete this check, then rerun the decision engine before trusting the current posture. Historical learning profile: Complete this check, then rerun the decision engine before trusting the current posture. Similar stored decisions: Complete this check, then rerun the decision engine before trusting the current posture. Do not raise confidence from training until real historical data is active.: Complete this check, then rerun the decision engine before trusting the current posture.
Decision can be shown with normal responsible-use language, while still requiring fresh odds before action.
Data coverage is 32/100: MVP mock/computed inputs are available, but 11 production signal(s) are still missing.
Barcelona vs Sevilla is loaded from the MVP mock provider.
Recent results and ratings are present in the mock fixture; production needs provider-backed historical rows.
Standings snapshots exist in the training schema but are not yet connected to live decisions.
Model computes home/away strength from team rating, league strength, and match context; data quality 90%.
Barcelona: W-W-D-L-W; Sevilla: W-D-L-W-W.
Synthetic injury/news signal slightly discounts the away side until real injury and suspension feeds are connected.
No suspension provider is connected.
Mock context feed sees no major projected lineup disruption; replace with confirmed lineup provider before production.
Required before trust: Injuries: Synthetic injury/news signal slightly discounts the away side until real injury and suspension feeds are connected. Suspensions: No suspension provider is connected. Lineups: Mock context feed sees no major projected lineup disruption; replace with confirmed lineup provider before production. News signals: Synthetic injury/news signal slightly discounts the away side until real injury and suspension feeds are connected. Weather: No weather provider is connected for outdoor totals/tempo markets. Fixture for the day: Barcelona vs Sevilla is loaded from the MVP mock provider. League standings: Standings snapshots exist in the training schema but are not yet connected to live decisions. Bookmaker odds: 3 market(s) and 7 selection(s) loaded.
Odds intelligence found 2 actionable value candidate(s) across 3 market(s); best is BTTS Yes.
Model 63%; no-vig 49%; edge +13.2%; EV +28.9%; fair 1.60. BTTS Yes has positive no-vig edge +13.2% and EV +28.9%. Score 0.545. Shortening tolerance 22%.
Model 62%; no-vig 51%; edge +12.0%; EV +23.7%; fair 1.60. Over 2.5 Goals has positive no-vig edge +12.0% and EV +23.7%. Score 0.467. Shortening tolerance 19%.
Model 24%; no-vig 19%; edge +4.7%; EV +19.1%; fair 4.20. Sevilla has positive EV but confidence is low, so it stays on watch. Score -0.116. Shortening tolerance 16%.
Model 38%; no-vig 49%; edge -12.0%; EV -24.2%; fair 2.66. Under 2.5 Goals is priced efficiently or short versus the model after margin removal. Score -0.316. Shortening tolerance 0%.
Match winner looks broadly efficient after margin removal; no selection clears value guardrails.
Goals over/under 2.5 has 1 actionable value candidate(s); best is Over 2.5 Goals at +23.7% EV.
Both teams to score has 1 actionable value candidate(s); best is BTTS Yes at +28.9% EV.
Avoid notes: Barcelona: Barcelona is priced efficiently or short versus the model after margin removal. Draw: Draw is priced efficiently or short versus the model after margin removal. Under 2.5 Goals: Under 2.5 Goals is priced efficiently or short versus the model after margin removal. BTTS No: BTTS No is priced efficiently or short versus the model after margin removal.
Watchlist notes: Sevilla: Sevilla has positive EV but confidence is low, so it stays on watch.
BTTS Yes has a resilient market buffer: current odds 2.06, fair odds 1.60, and 22% shortening tolerance before EV reaches zero.
Edge +13.2%; EV +28.9%. Current quoted odds and model probability before any market move.
Edge +12.4%; EV +25.0%. Small price move against the model thesis.
Edge +11.9%; EV +22.5%. Standard pre-action price stress used by the decision engine.
Edge +10.6%; EV +16.0%. Aggressive market move against the quoted value.
Edge +14.4%; EV +35.4%. Market drifts longer; value may improve but could indicate adverse news.
Movement alerts: Remove or downgrade if odds shorten more than 22% from the current quote. A 5% odds shortening does not fully break the thesis, but still requires a refresh.
Belief is strong: BTTS Yes at 63% with +13.2% edge, +28.9% EV, uncertainty 8/100, expires in 10 minutes.
Impact +13.2%. Model 63%, no-vig 49%, EV +28.9%.
Impact +2.5%. Decision can be shown with normal responsible-use language, while still requiring fresh odds before action.
Impact +1.5%. Case memory found 5 similar stored decisions and did not find enough historical pressure to downgrade.
Impact +2.8%. Barcelona expected goals 2.00, Sevilla expected goals 1.33; top scoreline 1-1.
Invalidates when: Invalidate if BTTS Yes no-vig edge falls to zero or EV turns negative. Invalidate if confirmed lineups, injuries, suspensions, weather, or live events materially oppose the current thesis. Invalidate if bookmaker prices move before the next refresh window.
Probability trace is ready: market prior 49% updated to posterior 60%, with +10.7% posterior edge and +23.7% EV.
49% to 49% (0.0%). Started from bookmaker-margin-adjusted probability 49% before applying model and context evidence.
49% to 60% (+10.8%). Weighted high model probability 63% against the no-vig prior by data quality 90% and bookmaker margin -1.7%.
60% to 60% (0.0%). Applied bounded context shift 0.0% from 2 injury, lineup, weather, news, live, or sport-context signal(s).
60% to 59% (-1.2%). Applied a small pull back toward the no-vig market because 7 selection(s) were market-prior calibrated.
59% to 59% (+0.0%). Data quality 90% does not force a discount; only a tiny reliability nudge is allowed.
59% to 59% (+0.4%). Case memory found 5 similar stored decisions and did not find enough historical pressure to downgrade.
59% to 60% (+0.7%). Decision can be shown with normal responsible-use language, while still requiring fresh odds before action.
60% to 60% (0.0%). No abstention gate triggered, so the posterior is not downgraded here.
Trace conflicts: Model-market disagreement is +13.2% between model probability and no-vig prior. Sevilla availability drag: Synthetic injury/news signal slightly discounts the away side until real injury and suspension feeds are connected.
Trace safeguards: Posterior probability is clamped between 2% and 98%. The probability trace cannot upgrade the final action beyond deterministic guardrails. Fresh odds, lineups, injuries, live events, and stored outcomes can still invalidate the posterior. This is public audit math, not hidden chain-of-thought or a guarantee of the match result.
Attribution is supportive: BTTS Yes value edge is the strongest driver, with value score 100/100 and risk score 36/100.
Probability impact +13.2%. BTTS Yes has no-vig edge +13.2% and EV +28.9%.
Probability impact +10.8%. Weighted high model probability 63% against the no-vig prior by data quality 90% and bookmaker margin -1.7%.
Decision can be shown with normal responsible-use language, while still requiring fresh odds before action.
Data coverage is 32/100: MVP mock/computed inputs are available, but 11 production signal(s) are still missing.
Injuries: Synthetic injury/news signal slightly discounts the away side until real injury and suspension feeds are connected.
Suspensions: No suspension provider is connected.
Missing-data drag: Injuries, Suspensions, Lineups, News signals, Weather.
The final action is mainly supported by model-vs-market edge, posterior probability, odds intelligence, and reliability checks.
Uncertainty needs watchlist treatment at 49/100; primary uncertainty is Data coverage uncertainty.
Data coverage is 32/100: MVP mock/computed inputs are available, but 11 production signal(s) are still missing.
Mitigation: Injuries: Synthetic injury/news signal slightly discounts the away side until real injury and suspension feeds are connected.
Model and no-vig market differ by +13.2%.
Mitigation: Refresh bookmaker odds, compare closing price, and rerun no-vig probability before trusting the edge.
Missing context: Weather check.
Mitigation: Fetch or verify Weather check before raising confidence.
Model uncertainty is low; model data quality is 90%.
Mitigation: Improve model inputs with provider-backed history, form, team/player availability, and settled calibration.
BTTS Yes has a resilient market buffer: current odds 2.06, fair odds 1.60, and 22% shortening tolerance before EV reaches zero.
Mitigation: Remove or downgrade if odds shorten more than 22% from the current quote.
Robustness is 98/100: 6/6 stress tests preserve the current action. Review loop kept monitor after repair checks: Do not raise confidence from training until real historical data is active. Refresh bookmaker odds: Recalculate implied probability, bookmaker margin, no-vig probability, value edge, and expected value.
Mitigation: Refresh bookmaker odds: Recalculate implied probability, bookmaker margin, no-vig probability, value edge, and expected value.
Decision impact: Keep the decision monitored until the primary uncertainty bucket is reduced.
Decision boundary is at-risk; nearest flip is Context-shock tolerance: +10.2% clearance.
Current 60%; threshold 49%; margin +11.5%. Posterior probability is 60% versus break-even 49%; no-vig market floor is 49%.
Current 2.06; threshold 1.67; margin +0.39 odds. Quoted odds 2.06 must stay at or above posterior fair odds 1.67.
Current +10.7%; threshold 0.0%; margin +10.7%. Current edge is +10.7%; the hard floor is positive edge.
Current +23.7%; threshold 0.0%; margin +23.7%. Current expected value is +23.7%; EV at or below zero removes value.
Current 79/100; threshold 24/100; margin +55 pts. Decision score is 79/100; lean-value consideration starts at 24, while strong value starts at 42 with high confidence.
Current 90/100; threshold 62/100; margin +28 pts. Model data quality is 90% and coverage audit score is 32/100; below 62/100 hard-blocks trust.
Current 49/100; threshold 66/100; margin +17 pts. Uncertainty score is 49/100; 66/100 or higher is high-risk unless mitigated.
Current +10.2%; threshold 0.0%; margin +10.2%. Worst-case stress still leaves minimum edge/EV margin at +10.2%.
Current 22%; threshold 3%; margin +19.4%. The price can shorten about 22% before value disappears; below 3% is execution-sensitive.
Must stay true: BTTS Yes posterior probability stays above break-even 49%. Quoted odds stay at or above posterior fair odds 1.67. No-vig edge and expected value stay positive. Decision score stays at or above 24 and no hard abstention gate triggers. Model data quality stays at or above 62/100. Uncertainty stays below 66/100 and context-shock stress keeps value above zero.
Flip triggers: No active boundary pressure; refresh odds and context before public display.
AI protocol needs data: 5/7 questions answered, 3 watch check(s), and 2 missing tool request(s).
BTTS Yes has model probability 63%, no-vig probability 49%, edge +13.2%, and EV +28.9%.
Follow-up: Refresh bookmaker odds before public display.
BTTS Yes has a resilient market buffer: current odds 2.06, fair odds 1.60, and 22% shortening tolerance before EV reaches zero.
Follow-up: Refresh odds, recompute no-vig probability, and downgrade if the latest quote crosses the fair-odds or EV threshold.
Context-shock tolerance: +10.2% clearance.
Follow-up: Use watchlist posture until data coverage or uncertainty improves.
Injuries: Synthetic injury/news signal slightly discounts the away side until real injury and suspension feeds are connected.
Follow-up: Injuries: Synthetic injury/news signal slightly discounts the away side until real injury and suspension feeds are connected.
Actionability is 89/100: BTTS Yes can be shown as an inspectable value candidate after the listed refresh checks. Review loop kept monitor after repair checks: Do not raise confidence from training until real historical data is active. Refresh bookmaker odds: Recalculate implied probability, bookmaker margin, no-vig probability, value edge, and expected value.
Follow-up: Fresh odds must keep no-vig edge and expected value positive.
No 10-year public historical evidence is attached to this decision run. Run the public historical evidence proof before using history to discipline raw model edges.
Follow-up: Keep historical discipline attached to the audit trail.
Refresh odds, recompute no-vig probability, and downgrade if the latest quote crosses the fair-odds or EV threshold.
Unlocks: Recalculates implied probability, no-vig edge, EV, boundary margins, and closing-line-value target.
Injuries: Synthetic injury/news signal slightly discounts the away side until real injury and suspension feeds are connected.
Unlocks: Reduces data/context uncertainty and reruns the actionability and review-loop gates.
Keep ready for in-play rechecks.
Unlocks: Updates live-state model inputs and hard in-play abstention gates.
Latest real-data backtest is available for shadow comparison only; live guardrails require an active model-bound calibration promotion.
Unlocks: Tunes learned minimum edge, data-quality weight, market-adjustment weight, and calibration thresholds.
Attach public historical evidence or a persisted provider-backed learning profile.
Unlocks: Confirms whether provider-enriched historical results beat market consensus before any raw edge can be promoted.
Reviewer guardrails: Use only supplied evidence IDs and model artifacts. Do not invent injuries, lineups, weather, news, odds, scores, or private facts. Do not expose hidden chain-of-thought; return public audit notes only. Do not upgrade a local avoid or monitor action into a stronger recommendation. Downgrade or abstain when safety gates, boundary breaches, or unsupported material claims remain.
Reasoning graph is contested: 7 supporting node(s), 5 watch node(s), and 0 blocker node(s).
Decide whether Barcelona vs Sevilla can show BTTS Yes as a responsible value candidate.
Probability trace is ready: market prior 49% updated to posterior 60%, with +10.7% posterior edge and +23.7% EV.
Evidence: probability-trace-summary
Odds intelligence found 2 actionable value candidate(s) across 3 market(s); best is BTTS Yes.
Evidence: odds-intelligence-summary, market-movement-summary
Data coverage is 32/100: MVP mock/computed inputs are available, but 11 production signal(s) are still missing.
Evidence: data-coverage-summary
Uncertainty needs watchlist treatment at 49/100; primary uncertainty is Data coverage uncertainty.
Evidence: uncertainty-summary
Decision boundary is at-risk; nearest flip is Context-shock tolerance: +10.2% clearance.
Evidence: decision-boundary-summary
Attribution is supportive: BTTS Yes value edge is the strongest driver, with value score 100/100 and risk score 36/100.
Evidence: attribution-summary
Actionability is 89/100: BTTS Yes can be shown as an inspectable value candidate after the listed refresh checks.
Evidence: review-loop-summary, attribution-summary
Strongest path: objective -> model-probability -> market-value -> boundary -> attribution -> actionability -> final-action.
Blocking path: objective -> data-coverage -> uncertainty -> learning-memory -> tool-requests -> final-action -> final-action.
Tool orchestration needs tools: 4/12 task(s) ready, 4 high-priority config gap(s), next task Refresh bookmaker odds and no-vig probabilities, readiness 48/100.
Provider: mockSportsDataProvider. Depends on: none.
Barcelona vs Sevilla is loaded from the MVP mock provider.
Decision impact: Without the fixture, the decision stays avoid and every downstream task is blocked.
Provider: mockSportsDataProvider. Depends on: fixtures-today.
Recent results and ratings are present in the mock fixture; production needs provider-backed historical rows.
Decision impact: Weak history keeps model strength, form weighting, and learned thresholds conservative.
Provider: missing-provider. Depends on: fixtures-today, historical-results.
Standings snapshots exist in the training schema but are not yet connected to live decisions.
Decision impact: Missing standings lower data quality and increase context uncertainty.
Provider: mockSportsDataProvider + deterministic-model. Depends on: fixtures-today, historical-results.
Recent form: Barcelona: W-W-D-L-W; Sevilla: W-D-L-W-W.
Decision impact: If form or home/away inputs change, the model probability and value edge are recalculated.
Provider: Sport-specific context providers. Depends on: fixtures-today.
Injuries: Synthetic injury/news signal slightly discounts the away side until real injury and suspension feeds are connected.
Decision impact: Material team news can downgrade or flip the action before kickoff.
Provider: Bookmaker odds provider. Depends on: fixtures-today.
Refresh odds, recompute no-vig probability, and downgrade if the latest quote crosses the fair-odds or EV threshold.
Decision impact: A price move can erase the edge, change the safer alternative, or move the pick to avoid.
Provider: Live score and event provider. Depends on: fixtures-today, odds-refresh.
Wait until kickoff or in-play mode before live score and event polling becomes active.
Decision impact: Live goals, cards, retirements, or tempo shocks can invalidate pre-match value.
Provider: missing-provider. Depends on: fixtures-today.
No weather provider is connected for outdoor totals/tempo markets.
Decision impact: Severe weather can move totals, BTTS, fatigue, or model uncertainty.
Blocking tasks: historical-results, standings-table, context-availability, odds-refresh.
Execution order: fixtures-today -> historical-results -> standings-table -> recent-form-home-away -> context-availability -> odds-refresh -> live-state-events -> weather-context -> historical-training -> historical-discipline -> decision-memory -> openai-review.
Tool execution audit is blocked: Load team/player historical results, Load league standings, Fetch injuries, suspensions, lineups, and news, Refresh bookmaker odds and no-vig probabilities must run before the current decision can be trusted.
Load today's fixture executed from mockSportsDataProvider with 1 observed record(s).
Outputs: fixture, kickoff, teams, league
Load team/player historical results is blocked: Recent results and ratings are present in the mock fixture; production needs provider-backed historical rows.
Outputs: team-history, player-history, long-form
Load league standings is blocked: Standings snapshots exist in the training schema but are not yet connected to live decisions.
Outputs: league-standings, motivation-context
Compute recent form and home/away profile executed from mockSportsDataProvider + deterministic-model with 12 observed record(s).
Outputs: recent-form, home-away-profile, team-strength
Fetch injuries, suspensions, lineups, and news is blocked: Injuries: Synthetic injury/news signal slightly discounts the away side until real injury and suspension feeds are connected.
Outputs: injuries, suspensions, lineups, news
Refresh bookmaker odds and no-vig probabilities is blocked: Refresh odds, recompute no-vig probability, and downgrade if the latest quote crosses the fair-odds or EV threshold.
Outputs: raw-implied-probability, no-vig-probability, expected-value, value-edge
Fetch live score and match events is waiting: Wait until kickoff or in-play mode before live score and event polling becomes active.
Outputs: live-score, match-events, in-play-state
Fetch weather where relevant is blocked: No weather provider is connected for outdoor totals/tempo markets.
Outputs: weather, wind-rain-temperature
Next run: Load team/player historical results: Recent results and ratings are present in the mock fixture; production needs provider-backed historical rows.
Control policy blocks public display: 1 blocker gate(s), primary blocker Tool execution.
Directive: Block public display and collect the required evidence first.
Next best action: Load team/player historical results: Recent results and ratings are present in the mock fixture; production needs provider-backed historical rows.
BTTS Yes is the current selected market with edge +13.2% and EV +28.9%.
BTTS Yes has a resilient market buffer: current odds 2.06, fair odds 1.60, and 22% shortening tolerance before EV reaches zero.
Data coverage is 32/100: MVP mock/computed inputs are available, but 11 production signal(s) are still missing.
Required: Injuries: Synthetic injury/news signal slightly discounts the away side until real injury and suspension feeds are connected.
Tool execution audit is blocked: Load team/player historical results, Load league standings, Fetch injuries, suspensions, lineups, and news, Refresh bookmaker odds and no-vig probabilities must run before the current decision can be trusted. Tool orchestration needs tools: 4/12 task(s) ready, 4 high-priority config gap(s), next task Refresh bookmaker odds and no-vig probabilities, readiness 48/100.
Required: Load team/player historical results: Recent results and ratings are present in the mock fixture; production needs provider-backed historical rows.
Decision boundary is at-risk; nearest flip is Context-shock tolerance: +10.2% clearance.
Required: Use watchlist posture until data coverage or uncertainty improves.
No 10-year public historical evidence is attached to this decision run. Run the public historical evidence proof before using history to discipline raw model edges.
AI protocol needs data: 5/7 questions answered, 3 watch check(s), and 2 missing tool request(s).
Required: Injuries: Synthetic injury/news signal slightly discounts the away side until real injury and suspension feeds are connected.
Reasoning graph is contested: 7 supporting node(s), 5 watch node(s), and 0 blocker node(s).
Required: data-coverage
Allowed: persist blocked audit only; collect required data; rerun the decision engine
Forbidden: publish as value candidate; show as actionable; upgrade by AI review; invent missing data
Monitoring is active with medium priority; review every 10 minutes and keep refresh bookmaker odds first.
Current edge is +13.2% and EV is +28.9%; a price move can erase value.
Action: Recalculate implied probability, bookmaker margin, no-vig probability, value edge, and expected value.
This missing signal can move the probability estimate or downgrade confidence before kickoff.
Action: Refresh the model, rerun the no-vig EV calculation, and re-check abstention gates.
Without enough real historical fixtures and odds, thresholds stay conservative defaults.
Action: Import real fixtures and odds, run backtests, then activate learned thresholds only after the corpus is large enough.
Case memory found 5 similar stored decisions and did not find enough historical pressure to downgrade.
Action: Keep memory neutral while continuing to store and settle outcomes.
Stop when: Stop if BTTS Yes no-vig edge falls to zero or expected value turns negative. Invalidate if BTTS Yes no-vig edge falls to zero or EV turns negative. Invalidate if confirmed lineups, injuries, suspensions, weather, or live events materially oppose the current thesis. Invalidate if bookmaker prices move before the next refresh window.
Escalate when: If two high-priority monitoring tasks remain unresolved at the next review, downgrade the action to monitor or avoid. If bookmaker movement removes the value edge, remove the recommendation and rerun the committee. If confirmed team news, weather, surface, or live-event data opposes the thesis, rerun the model before showing the pick. Do not raise confidence from historical learning until the real-data profile is active.
Actionability is 89/100: BTTS Yes can be shown as an inspectable value candidate after the listed refresh checks.
BTTS Yes has edge +13.2% and EV +28.9%.
Confidence is high and risk is low.
Data quality is 90%.
Missing signals: Weather check.
Belief is strong: BTTS Yes at 63% with +13.2% edge, +28.9% EV, uncertainty 8/100, expires in 10 minutes.
Before action: Do not raise confidence from training until real historical data is active. Refresh bookmaker odds: Recalculate implied probability, bookmaker margin, no-vig probability, value edge, and expected value. Weather check: Refresh the model, rerun the no-vig EV calculation, and re-check abstention gates. Historical learning profile: Import real fixtures and odds, run backtests, then activate learned thresholds only after the corpus is large enough.
Treat the output as statistical analysis, not certainty. Do not use this audit as staking, bankroll, or financial advice. Refresh odds and context before relying on any displayed edge. Avoid acting when the monitoring plan is blocked, expired, or unresolved.
Review loop kept monitor after repair checks: Do not raise confidence from training until real historical data is active. Refresh bookmaker odds: Recalculate implied probability, bookmaker margin, no-vig probability, value edge, and expected value.
Evidence: BTTS Yes is the leading thesis because the model price beats the no-vig market by +13.2% with +28.9% EV per unit. Belief is strong: BTTS Yes at 63% with +13.2% edge, +28.9% EV, uncertainty 8/100, expires in 10 minutes.
Evidence: Historical learning profile: Latest real-data backtest is available for shadow comparison only; live guardrails require an active model-bound calibration promotion. The public-action invariant downgraded the committee candidate from consider to monitor. Decision committee recommends consider with unanimous consensus: stable decision: surface BTTS Yes as inspectable value, but require fresh odds and context checks before action.
Required: Historical learning profile: Latest real-data backtest is available for shadow comparison only; live guardrails require an active model-bound calibration promotion.
Evidence: Monitoring is active with medium priority; review every 10 minutes and keep refresh bookmaker odds first. Refresh bookmaker odds: Current edge is +13.2% and EV is +28.9%; a price move can erase value. Weather check: This missing signal can move the probability estimate or downgrade confidence before kickoff. Historical learning profile: Without enough real historical fixtures and odds, thresholds stay conservative defaults.
Required: Fetch missing context, refresh probabilities, and rerun actionability.
Evidence: Do not raise confidence from training until real historical data is active. Refresh bookmaker odds: Recalculate implied probability, bookmaker margin, no-vig probability, value edge, and expected value. Weather check: Refresh the model, rerun the no-vig EV calculation, and re-check abstention gates. Historical learning profile: Import real fixtures and odds, run backtests, then activate learned thresholds only after the corpus is large enough.
Required: Do not raise confidence from training until real historical data is active.
Evidence: Actionability is 89/100: BTTS Yes can be shown as an inspectable value candidate after the listed refresh checks. Decision can be shown with normal responsible-use language, while still requiring fresh odds before action. Risk shift: raise; confidence shift: keep.
Repairs: Do not raise confidence from training until real historical data is active. Refresh bookmaker odds: Recalculate implied probability, bookmaker margin, no-vig probability, value edge, and expected value. Weather check: Refresh the model, rerun the no-vig EV calculation, and re-check abstention gates. Historical learning profile: Import real fixtures and odds, run backtests, then activate learned thresholds only after the corpus is large enough.
Unresolved: Historical learning profile: Latest real-data backtest is available for shadow comparison only; live guardrails require an active model-bound calibration promotion. Missing signal: Weather check.
Release criteria: Fresh odds must keep no-vig edge and expected value positive. Belief state must be unexpired before the decision remains visible. Monitoring plan must be active or explicitly cleared. Actionability must stay actionable or the product should downgrade to watch/avoid. Do not raise confidence from training until real historical data is active. Refresh bookmaker odds: Recalculate implied probability, bookmaker margin, no-vig probability, value edge, and expected value. Weather check: Refresh the model, rerun the no-vig EV calculation, and re-check abstention gates.
Robustness is 98/100: 6/6 stress tests preserve the current action.
Edge +11.4%; EV +25.2%. A 5% odds shortening projects a smaller no-vig edge for BTTS Yes.
Repair: Refresh bookmaker odds and remove the pick if edge or EV is no longer positive.
Edge +11.2%; EV +24.8%. Stress applies unresolved context gaps: Weather check.
Repair: Fetch lineups, injuries, suspensions, weather, and news before keeping the same action.
Edge +11.7%; EV +25.8%. Data-quality stress reflects current data score 90% and provider uncertainty.
Repair: Improve provider coverage or downgrade confidence until real data fills the gap.
Edge +11.7%; EV +25.8%. Monitoring state is active; stale belief should reduce trust in the edge.
Repair: Rerun the belief state and monitoring plan before showing the candidate again.
Edge +10.7%; EV +23.8%. Review loop status is repaired; unresolved repairs should be priced into the decision.
Repair: Do not raise confidence from training until real historical data is active.
Edge +10.2%; EV +22.7%. Actionability status is actionable with score 89/100.
Repair: Do not raise confidence from training until real historical data is active.
Hedges: Both Teams To Score: BTTS Yes at model 63%. Goals: Over 2.5 Goals at model 62%. Double chance: Barcelona or Draw at model 76%.
Required rechecks: Refresh bookmaker odds: Recalculate implied probability, bookmaker margin, no-vig probability, value edge, and expected value. Weather check: Refresh the model, rerun the no-vig EV calculation, and re-check abstention gates. Fresh odds must keep no-vig edge and expected value positive. Belief state must be unexpired before the decision remains visible.
Evaluation plan keeps this on watch: settle the final result and inspect whether missing context or price movement would have changed the action.
Settle whether BTTS Yes won for Barcelona vs Sevilla.
Capture closing odds for BTTS Yes and compare them with the quoted 2.06.
Record whether monitoring tasks changed the thesis before kickoff: Refresh bookmaker odds, Weather check, Historical learning profile.
Link the settled outcome to this decision run so confidence, health, Brier score, ROI, and CLV can be measured.
Success: BTTS Yes settles as correct for the chosen market. Closing-line value is at least positive or the closing no-vig probability confirms the edge. Settled outcome improves calibration for high-confidence low-risk decisions. No unresolved review-loop release criterion would have blocked the pick at kickoff.
Failure: BTTS Yes loses or pushes against the selected market settlement rules. Closing odds move against the thesis enough to erase the pre-match value edge. A required recheck was missed: Refresh bookmaker odds: Recalculate implied probability, bookmaker margin, no-vig probability, value edge, and expected value. The outcome joins similar stored cases that later discount this pattern.
Learning questions: Was 63% model probability calibrated against the binary settlement result? Did the no-vig market probability of 49% underprice the selection at decision time? Did closing odds validate the edge or expose stale market data? Did the unresolved monitoring or review-loop checks predict the final risk?
Post-match actions: Store the settled outcome through the decision outcome endpoint with the linked decision_run_id. Recompute calibration by confidence and decision health after settlement. Compare quoted odds with closing odds for closing-line value. Keep learned thresholds inactive until enough real historical fixtures and odds are imported.
Case memory found 5 similar stored decisions and did not find enough historical pressure to downgrade.
epl-001; score 79; reliability 100/100; same market both_teams_to_score, model probability distance, same selection label
epl-001; score 79; reliability 100/100; same market both_teams_to_score, model probability distance, same selection label
epl-001; score 79; reliability 100/100; same market both_teams_to_score, model probability distance, same selection label
Latest real-data backtest is available for shadow comparison only; live guardrails require an active model-bound calibration promotion.
Barcelona vs Sevilla; data quality 90%; missing signals 1.
Model produced 2.00-1.33 expected goals and winner total 1.000.
BTTS Yes is the current best edge at +13.2%.
BTTS Yes is classified as low risk.
0 watch/conflict checks and 0 triggered abstention gates.
Weighted decision score 79; the agent will only surface the selection if guardrails agree.
The public-action invariant downgraded the committee candidate from consider to monitor. Decision committee recommends consider with unanimous consensus: stable decision: surface BTTS Yes as inspectable value, but require fresh odds and context checks before action.
BTTS Yes is mathematically live because model probability is 63% against no-vig 49%.
Challenge: Weather check: Weather is not connected yet; relevant for outdoor football totals and tempo. Sevilla availability drag: Synthetic injury/news signal slightly discounts the away side until real injury and suspension feeds are connected. Source: mock-context-feed; confidence 54%.
Checks: Recompute probabilities after fresh odds and context updates.
The price still has to survive market scrutiny; odds shortening would change the edge on BTTS Yes.
Checks: Refresh bookmaker odds before surfacing a final edge.
Context is incomplete for Barcelona vs Sevilla; 1 missing signals remain.
Challenge: Weather check A moderate adverse injury, suspension, or lineup shock is applied to the selected side.
Checks: Check lineup, injury, suspension, weather, live-state, and news signals near start time.
Reliability is 100/100 with stable health and trust calibration action.
Checks: Honor triggered abstention gates before any public recommendation.
Case memory found 5 similar stored decisions and did not find enough historical pressure to downgrade.
Challenge: Compared against 40 stored football decisions. Average similar-case reliability is 100/100.
Checks: Store this decision and settle outcomes so memory can learn from real results.
stable decision: surface BTTS Yes as inspectable value, but require fresh odds and context checks before action.
Challenge: Weather check Fresh bookmaker price Historical learning profile
Checks: Downgrade to monitor or avoid if adverse lineup, injury, weather, live-state, or data-quality signals oppose the selected side. Recalculate immediately; if the no-vig edge falls below zero or EV turns negative, remove BTTS Yes from value picks.
Open disagreements: Final public action is monitor; the pre-invariant committee candidate was consider.
BTTS Yes carries model probability 63%, no-vig implied 49%, odds 2.06, and EV +28.9%.
Impact: Supports showing the pick as inspectable value.
The selection still depends on a live bookmaker price. Market movement scenario projects consider at score 72.
Impact: Requires odds refresh before the recommendation can be trusted.
1 missing context signals remain, led by Weather check.
Impact: Controls whether the engine considers, monitors, or abstains after late provider data arrives.
Base case decision score is 79; calibration says trust with 100/100 reliability.
Impact: stable decision: surface BTTS Yes as inspectable value, but require fresh odds and context checks before action.
Case memory found 5 similar stored decisions and did not find enough historical pressure to downgrade.
Impact: Memory stays neutral.
This missing signal can move the probability estimate or downgrade confidence before kickoff.
If confirmed: Refresh the model, rerun the no-vig EV calculation, and re-check abstention gates.
A shorter price can erase the +13.2% edge and +28.9% EV.
If confirmed: Recalculate implied probability, bookmaker margin, no-vig probability, value edge, and EV.
Without enough real historical fixtures and odds, thresholds stay conservative defaults.
If confirmed: Import real fixtures and odds, run backtests, then activate learned thresholds only after the corpus is large enough.
Case memory found 5 similar stored decisions and did not find enough historical pressure to downgrade.
If confirmed: Keep memory neutral while continuing to store and settle outcomes.
Bad-data path: Downgrade to monitor or avoid if adverse lineup, injury, weather, live-state, or data-quality signals oppose the selected side. Market-move path: Recalculate immediately; if the no-vig edge falls below zero or EV turns negative, remove BTTS Yes from value picks.
Context layer applied residual probability effects after reviewing 2 structured signals.
Mock context feed sees no major projected lineup disruption; replace with confirmed lineup provider before production.
Synthetic injury/news signal slightly discounts the away side until real injury and suspension feeds are connected.
Positive no-vig edge of +13.2% versus margin-adjusted implied probability.
Expected return is +28.9% per unit at decimal odds 2.06.
high confidence after edge, probability, and data-quality checks.
Data quality is 90%.
low risk based on confidence and odds level.
Lineups, injury/suspension news, and weather are not connected yet.
Pre-match state avoids in-play recalibration risk.
Latest real-data backtest is available for shadow comparison only; live guardrails require an active model-bound calibration promotion.
Case memory found 5 similar stored decisions and did not find enough historical pressure to downgrade.
Run the public historical evidence proof before using history to discipline raw model edges.
Home/draw/away probabilities sum to 1.000.
Confidence is high with 90% data quality.
Missing context does not contradict the current action.
BTTS Yes is low risk while action is consider.
Fixture is not live, so pre-match modeling is coherent.
Barcelona expected goals 2.00, Sevilla expected goals 1.33; top scoreline 1-1.
BTTS Yes has 63% model probability versus 49% no-vig implied probability. Raw implied is 49% and market margin is -1.7%. Expected value is +28.9% per unit.
Barcelona: W-W-D-L-W; Sevilla: W-D-L-W-W.
Data quality is 90%; low quality downgrades confidence.
Weather is not connected yet; relevant for outdoor football totals and tempo.
Provider football context xG: Poisson expected goals used deterministic team strength, form, and xG proxies because provider-backed football context was not attached.
Context layer applied residual probability effects after reviewing 2 structured signals. Side shift home 0.0%, away -1.0%; total shift 0.0%.
Mock context feed sees no major projected lineup disruption; replace with confirmed lineup provider before production. Source: mock-context-feed; confidence 58%.
Synthetic injury/news signal slightly discounts the away side until real injury and suspension feeds are connected. Source: mock-context-feed; confidence 54%.
Model 63% - fair odds 1.60
Model 62% - fair odds 1.60
Model 38% - fair odds 2.66
Model 37% - fair odds 2.67
Model 76% - fair odds 1.31 - needs bookmaker market
Current model, odds, data-quality, and missing-context state.
Projected edge would be +11.9%.
A moderate adverse injury, suspension, or lineup shock is applied to the selected side.
Lineups, injury/suspension news, and weather arrive and support the existing model signal.
BTTS Yes clears the positive-edge filter at +13.2% and EV +28.9%.
Data quality is 90%; below 62% the agent abstains.
BTTS Yes risk is low; high-risk picks need a wider edge before consideration.
Pre-match fixture; in-play event model is not required yet.
Missing context is noted but does not force abstention at the current edge/data-quality level.
Fixture has no future-season seed or synthetic market flag.
No real-data training profile is active; default minimum-edge guardrail is used.
Case memory did not find enough similar stored pressure to force abstention.
No public historical discipline evidence is attached, so this gate stays neutral.
Verdict: no clear value
The agent does not see enough separation between OddsPadi probabilities and no-vig market probabilities for Barcelona vs Sevilla.
Model football-poisson-v2; data quality 90%.