Lab
Research record. Patterns are hypotheses until a challenger validates them out of sample.
Lottery LabLeg count, dogs, time spread, near misses, failure position, counterfactuals›Premium Payout LabPayout bands, efficiency, calibration by band, drawdown›Research QueueWhat runs next and why — validation is never starved›Models & DataProps, opportunity & joint models, capability matrix, quarantine, sport plugins›Portfolio & RiskShared exposure, drift, recipe decay, incidents, disagreement, lineage, stress›Research HealthEfficiency, hypotheses, saturation, coverage balance, data gaps, family scorecards›System HealthGO / NO-GO, deployment, workers, backups, alerts, risk controls›
Real-data research pending until provider keys are configured. Everything below is the research record.
Research record (composer, summaries, recipes, experiments, patterns)
Composer runs
Constraint failures and optimizer comparisons from the Build screen. Every run is replayable.
NO COMPOSER RUNS YET
Daily research summaries
Review only — one day never changes models or recipes.
NO SUMMARIES YET
Created daily after settlement.
Performance by origin
Control, experimental, challenger and production research are never mixed.
INSUFFICIENT EVIDENCE
No settled paper tickets yet.
Recipe performance
Sorted by sample size. Small samples are not conclusions.
Controls & challengers
- BALANCED@V1CONTROLUNVALIDATED
- HIGHEST_PROBABILITY@V1CONTROLUNVALIDATED
- HIGH_ODDS@V1CONTROLUNVALIDATED
- HIGH_ODDS_ONE_DOG@V1CONTROLUNVALIDATED
- LOTTERY@V1CONTROLUNVALIDATED
- LOTTERY_ONE_DOG@V1CONTROLUNVALIDATED
- LOTTERY_ONE_DOG_PROPS@V1CONTROLUNVALIDATED
- LOTTERY_PROPS@V1CONTROLUNVALIDATED
- PREMIUM_PAYOUT@V1CONTROLUNVALIDATED
- PREMIUM_PAYOUT_ONE_DOG@V1CONTROLUNVALIDATED
- PROOF_TWO_HEAVY_FAVORITES@V1EXPERIMENTALUNVALIDATED
- VALUE@V1CONTROLUNVALIDATED
Experiments
Stopping rules are fixed before results. Promotion needs the gate and an approver.
NO EXPERIMENTS YET
Failure & winning patterns
Evidence for hypotheses — never automatic rules.
NO PATTERNS YET