{"request_id":"a6f5f670-4b22-4f96-9e73-642c73b592e7","inference_contract":"/v1/meta/inference","description":"ST-IM is an inference layer built on decades of Stock Trends classification history. Its value is not only point-in-time prediction, but probabilistic interpretation of current signals against long-horizon historical behavior. ST-IM outputs are not guarantees, price targets, investment advice, or direct buy/sell commands.","data_provenance":{"historical_coverage_start_year":1980,"approximate_observation_count":"16M+","update_frequency":"weekly","classification_framework":"Stock Trends trend classification methodology","semantic_continuity":"Stock Trends indicators use a consistent classification doctrine across decades of observations.","native_signal_domains":["trend classification","relative performance","relative performance direction","volume activity","market breadth","sector leadership","regime structure"],"research_value":["long-horizon signal validation","regime analysis","sector rotation research","portfolio construction research","causal and probabilistic market analysis","agentic market-intelligence workflows"],"important_limits":["Historical depth supports research and probabilistic interpretation but does not guarantee future performance.","Stock Trends outputs are not investment advice, price targets, or direct buy/sell commands.","Externally named market events are not native Stock Trends labels unless separately curated."]},"inference_provider":{"provider_id":"stim","provider_name":"Stock Trends Inference Model","provider_role":"current_baseline_inference_provider","not_final_intelligence_layer":true,"architecture_source":"docs/STOCK_TRENDS_COGNITION_ARCHITECTURE.md"},"base_period_mean_returns_pct":{"x4wk":0.0,"x13wk":2.19,"x40wk":6.45},"forecast_horizons":[{"id":"x4wk","weeks":4,"base_period_mean_return_pct":0.0},{"id":"x13wk","weeks":13,"base_period_mean_return_pct":2.19},{"id":"x40wk","weeks":40,"base_period_mean_return_pct":6.45}],"columns":{"x4wk1/x13wk1/x40wk1":"Lower confidence interval bound for mean return (%)","x4wk2/x13wk2/x40wk2":"Upper confidence interval bound for mean return (%)","x4wk/x13wk/x40wk":"Midpoint / mean estimate (%)","x4wksd/x13wksd/x40wksd":"Std dev of assumed normal distribution (%)"},"classification_role":"The Stock Trends classification system converts raw weekly market behavior into structured, repeatable factor states. These states create historical populations from which forward-return distributions can be estimated and compared.","randomness_assumptions":["Markets are noisy, uncertain, and partly random.","ST-IM does not eliminate randomness; it estimates conditional historical tendencies.","Individual outcomes can diverge materially from the estimated distribution."],"distribution_framing":{"assumption":"normal_approximation","central_limit_theorem_intuition":"Large historical populations of similarly classified observations can support distribution-level reasoning even though individual securities remain uncertain.","probability_formula":"probability_outperform = 1 - normal_cdf((base_mean - stim_mean) / standard_deviation)"},"probability_interpretation":["Interpret ST-IM probabilities as conditional historical tendencies.","Do not interpret probabilities as guarantees, price targets, or direct buy/sell commands.","Compare each ST-IM mean and lower bound to the matching base-period mean before drawing conclusions.","Treat stale or missing ST-IM records as historical fallback context."],"stim_select":{"role":"strict_filtering_framework","description":"Identifies instruments whose probability and confidence profiles appear favorable relative to broad/random baseline distributions.","criteria":{"x4wk1":{"operator":">","threshold_pct":0.0},"x13wk1":{"operator":">","threshold_pct":2.19},"x40wk1":{"operator":">","threshold_pct":6.45},"prob13wk":{"operator":">=","threshold":0.55},"all_criteria_required":true},"ranking":"prob13wk descending"},"strengths":["Repeatable classification-driven inference","Distribution-level comparison across large historical populations","Useful ranking and screening context","Explainable baseline for future provider comparison","Portfolio workflow support through repeated decision-making"],"portfolio_applications":["ranking","screening","allocation_review","regime-aware interpretation","market-structure analysis","repeated decision-making under uncertainty"],"limitations":["regime_shifts","non_stationarity","sample_size_weakness","tail_events","liquidity_shocks","news_shocks","uncertainty_in_individual_stock_outcomes"],"future_provider_compatibility":"ST-IM is the current baseline inference provider. Future Causal AI providers should integrate through /v1/meta/inference rather than forcing all cognition surfaces into ST-IM-specific fields.","missing_data_note":"If a symbol has no ST-IM estimate for a weekdate, the provider could not estimate reliably due to insufficient samples."}