Evidence & Validation

Evaluating the Stock Trends API

Use Stock Trends when you need classified market states, probabilistic forward-return context, mature outcome evidence, portfolio return history, and agent-ready workflow metadata — not just quotes and HLCV.

Stock Trends is context infrastructure that augments your own investment process rather than replacing it, so the evidence that matters is evidence you can inspect yourself. Three separate pillars are documented below, each exposed as endpoints you can query directly.

This page provides framework evaluation evidence for developers, AI agents, researchers, and prospective customers. It documents the historical foundation, ST-IM realized outcome data, Monte Carlo process analysis, portfolio evidence endpoints, and agent research workflows. These materials support evaluation and research — they are not guarantees or price targets.

View ST-IM outcomes endpoint Understand ST-IM API documentation
Three pillars

Three Separate Kinds of Evidence

Stock Trends evidence falls into three distinct families. They have different histories, different methodologies, and different limitations, and they answer different questions. Read them separately — combining them into a single performance figure would misrepresent all three.

  • 1. Historical classification provenance A consistent weekly classification record since 1980, with 16M+ structured observations. Establishes comparability across market cycles — it is a data foundation, not a performance result. Jump to section
  • 2. Realized statistical outcome evidence Mature ST-IM Select forward returns at 4, 13, and 40 weeks, measured against base-period means, plus Monte Carlo analysis of repeated process application. Distributional evidence, not projections. Jump to section
  • 3. Strategy and model-portfolio evidence Declared strategy rules, model-portfolio return histories, closed-position records, and portfolio-to-strategy provenance — all inspectable as endpoints. Process-level evidence, not audited accounts. Jump to section

These three pillars are the evidence architecture. Other records referenced on this page — the Stock Trends weekly publication history and the Monte Carlo process simulations — are supporting material that sits underneath a pillar rather than standing as a fourth or fifth kind of evidence. Long history does not eliminate uncertainty, transaction costs, market risk, or regime change. Each pillar below states its own scope and limitations.

Why Stock Trends

Stock Trends API vs. a Generic Market Data API

The Stock Trends API is not a data pipe. It layers a decades-tested classification and inference framework on top of market data, making it practical to build evidence-grounded agent workflows without assembling the framework yourself.

Capability Generic market data API Stock Trends API
Quotes / HLCV Yes Yes
Trend classifications (Bullish/Bearish/Weak) No Yes
Relative strength vs. S&P 500 No Yes
Volume behavior context No Yes
Forward outcome evidence (mature realized data) No Yes
Portfolio return history No Yes
Portfolio strategy provenance No Yes
Agent workflow metadata (OpenAPI, llms.txt, /ai/tools) No Yes
Machine payment No x402 and MPP supported
Historical foundation

Built on Decades of Weekly Market Observations

1980+ coverage

Stock Trends is built on decades of weekly market observations, not short-lived signal backtests. The classification record extends back to 1980 with continuous weekly coverage across NYSE, NASDAQ, AMEX, TSX, and index data, and 16M+ structured weekly observations encoded through a consistent Stock Trends doctrine.

This weekly cadence supports intermediate-term market and portfolio research. It captures trend formation, regime transitions, relative strength shifts, and volume behavior across many distinct market cycles.

  • Classification history since 1980
  • 16M+ weekly observations
  • NYSE, NASDAQ, AMEX, TSX, and index coverage
  • Weekly classification cadence
  • Consistent signal semantics across decades
  • Trend classification, relative strength vs. S&P 500, volume behavior, market breadth, and sector leadership — all accessible through the API
  • Historical depth is research provenance — not investment advice, not guaranteed future performance

Stock Trends outputs are designed to improve the distribution of investment outcomes. They are not buy/sell commands, price targets, or guarantees of future results.

Framework

The Stock Trends Classification Framework

The Stock Trends framework transforms weekly price and volume data into structured, named market states. This classification layer makes long-horizon statistical research possible: it creates repeatable, comparable observations across decades and market cycles.

Core framework elements exposed through the API:

  • Trend classification — Bullish, Bearish, Weak Bullish, Weak Bearish, and crossover states
  • Relative strength vs. the S&P 500 — above or below benchmark, with direction
  • Volume behavior — high, low, or normal relative to historical norms
  • Multi-horizon forward outcome tracking — 4-week, 13-week, and 40-week horizons
  • Market breadth and sector leadership — regime context across exchanges

The framework is designed to improve the distribution of investment outcomes by supporting more systematic, evidence-based decision-making. It does not issue simplistic buy/sell commands or guarantee results.

ST-IM evidence

ST-IM Select: Mature Realized Outcome Evidence

The Stock Trends Inference Model (ST-IM) uses base-period means derived from the full historical observation record as evaluation thresholds. ST-IM Select identifies observations whose modeled return distributions exceed those thresholds across forward-return horizons.

The table below shows mature realized outcomes from historical ST-IM Select observations. These are actual forward returns from observations that have already completed their measurement window — not projections. The data is available live through the API at GET /v1/selections/stim-select/outcomes/summary.

Measurement window: these figures come from the endpoint's default trailing ten-year window — observations dated 2016-03-04 through 2026-02-27. They are not drawn from the full record back to 1980, and they cover a different span than the strategy and portfolio histories below. Pass explicit start_date and end_date parameters to evaluate a different window.
Horizon Count Avg return Median return Base mean Positive rate Outperform base rate
4 weeks 156,868 0.72% 0.30% 0.00% 51.5% 51.5%
13 weeks 156,889 2.68% 1.00% 2.19% 52.7% 46.7%
40 weeks 139,743 6.80% 1.60% 6.45% 52.3% 43.0%

How to read this table:

  • Avg return exceeds the base-period mean at all three horizons, indicating the ST-IM Select filter has historically identified observations with above-baseline average returns.
  • Median return is lower than the average, reflecting a distribution with positive skew and a wide range of outcomes — most observations cluster near or slightly above zero, with meaningful upside cases pulling the average higher.
  • Outperform base rate at 13w and 40w is below 50%, meaning most individual observations do not outperform the base-period mean — this is consistent with average performance being driven by a skewed distribution rather than uniform outperformance.
  • These results are intended as framework evaluation evidence, not as projections of future returns.
ST-IM Select outcome data represents historical observations that can be evaluated against base-period assumptions — not guaranteed outcomes, not price targets, and not direct buy/sell advice. Past historical outcome distributions do not guarantee future performance.
Understanding Our Assumptions: A Decade Later
Process analysis

Picks of the Week: Monte Carlo Simulation

Picks of the Week is a long-running weekly Stock Trends report based on stable, rule-based selection criteria. Monte Carlo simulation characterizes process-level outcome distributions: rather than measuring isolated average returns, it stress-tests repeated application of the same selection process across many simulated portfolio sequences.

Two simulation methods are used:

  • Realistic simulation — follows actual historical chronology and weekly available Picks of the Week populations, as the process would have been applied in practice
  • Theoretical simulation — samples from the full return distribution without chronological constraint

Simulation results (portfolio growth multiples from a normalized starting value of 1.0):

Method Max final Median final Min final Mean CAGR Sharpe ratio
Realistic 29.13 1.34 0.10 3.38% 0.38
Theoretical 28.83 1.48 0.20 4.39% 0.61

The wide range from minimum to maximum outcome reflects realistic uncertainty in repeated process application. These simulations characterize possible outcome distributions — they do not prove guaranteed alpha or eliminate the possibility of loss.

Methodology note: Results depend on simulation assumptions and historical data. Real-world implementation may involve additional considerations such as taxes, transaction costs, execution slippage, and investor behavior. Simulations evaluate repeated process application across historical data — they do not guarantee future results.
Interactive Monte Carlo simulation Read the article
Pillar 3 · Strategy and model-portfolio evidence

Inspectable Process-Level Evidence

1981+ portfolio history

Most performance claims ask you to trust a number. This pillar asks you to inspect a process. The Stock Trends API exposes official model portfolios together with the declared rules that governed them, the return history they produced, and the individual closed positions underneath that history — each as a separate, queryable endpoint.

Return histories for the longest-running model portfolios begin in 1981 and continue to the current week. Twenty official model portfolios are currently exposed, spanning NYSE, NASDAQ, TSX, index-constituent, ETF, and ST-IM Select strategies, with declared transaction-cost and stop-loss assumptions attached to each strategy definition.

One distinction matters when reading these histories: the portfolio record extends further back than the Stock Trends weekly publication record, which begins in 1993. The full span from 1981 should therefore be described as a long-running model-portfolio history, not as a forward-published one, and neither should be conflated with the ST-IM realized-outcome evidence above, which covers a different period entirely.

Four endpoint families combine into evidence you can audit rather than accept:

  • Strategy rules — the declared buy/sell conditions, stop-loss settings, and economic assumptions
  • + Portfolio history — the chronological return record those rules produced
  • + Closed-position history — the individual positions underneath that record
  • + Strategy provenance — the join proving which strategy governed which portfolio
  • = Inspectable process-level evidence

Portfolio metadata and history

  • GET /v1/stocktrends/portfolios — list official Stock Trends model portfolios
  • GET /v1/stocktrends/portfolios/{port_id} — metadata for one model portfolio
  • GET /v1/stocktrends/portfolios/{port_id}/returns — chronological return observations
  • GET /v1/stocktrends/portfolios/{port_id}/summary — public history summary, including annualized ROI using the Stock Trends average-investment method
  • GET /v1/stocktrends/portfolios/{port_id}/positions/history — closed-position records

Strategy rules and provenance

  • GET /v1/stocktrends/strategies — strategy definitions, condition counts, and economic assumptions
  • GET /v1/stocktrends/strategies/{strategy_id} — declared buy/sell conditions, stop-loss settings, and investment assumptions
  • GET /v1/stocktrends/portfolios/{port_id}/strategy — portfolio-to-strategy provenance

Comparison

  • POST /v1/portfolio/compare — evaluate two portfolios under identical decision-scoring and regime logic

The history endpoints supply the evidence. The strategy endpoints supply the methodology that explains what produced it. Together they let agents and researchers evaluate Stock Trends portfolios without treating them as a black box — and without relying on this page's description of them.

Scope and limitations: these are rule-based model portfolio records, not audited brokerage-account performance, not live client capital, and not a record of executed trades. Strategy conditions are metadata, not executable endpoints — these routes do not return current matching stocks, current buy or sell candidates, or current live holdings, which are excluded from public history endpoints by design. Declared transaction-cost and stop-loss assumptions may differ from real-world execution, which also involves taxes, slippage, liquidity constraints, and market risk. Long history does not eliminate uncertainty or regime change, and it does not guarantee future performance.

Agent workflow

Agent Evaluation Workflow

A practical workflow for agents and researchers evaluating the Stock Trends framework:

  1. Discover tools — GET /v1/ai/tools and /llms.txt provide machine-readable workflow guidance and endpoint metadata
  2. Understand assumptions — GET /v1/meta/stim exposes ST-IM base-period means and modeling assumptions
  3. Review ST-IM outcome evidence — GET /v1/selections/stim-select/outcomes/summary returns mature realized outcome data across 4-week, 13-week, and 40-week horizons
  4. Inspect strategy definitions — GET /v1/stocktrends/strategies and GET /v1/stocktrends/strategies/{strategy_id} expose the rules behind each portfolio strategy
  5. Evaluate a portfolio — GET /v1/stocktrends/portfolios/{port_id}/returns provides return history, GET /v1/stocktrends/portfolios/{port_id}/summary adds an annualized ROI summary, and GET /v1/stocktrends/portfolios/{port_id}/positions/history exposes the closed positions underneath it
  6. Confirm provenance — GET /v1/stocktrends/portfolios/{port_id}/strategy joins a portfolio to the strategy definition that governed it, so the rules and the record can be checked against each other
  7. Compare — POST /v1/portfolio/compare evaluates two portfolios under identical decision-scoring and regime logic
  8. Generate research output — combine outcome evidence, strategy context, and regime/breadth data into advisory-safe research summaries using framework evaluation language rather than investment advice, keeping the three evidence pillars distinct rather than merging them into one performance claim

Relevant resources:

  • /v1/openapi.json — full OpenAPI spec
  • /llms.txt — machine-readable agent discovery
  • GET /v1/ai/context — indicator and dataset grounding
  • GET /v1/ai/tools — primary agent workflow discovery
  • GET /v1/meta/stim — ST-IM assumptions and base-period data
  • GET /v1/selections/stim-select/outcomes/summary — mature realized outcome data
  • GET /v1/stocktrends/strategies — strategy framework definitions
  • GET /v1/stocktrends/portfolios — official Stock Trends model portfolios
  • GET /v1/stocktrends/portfolios/{port_id}/returns — portfolio return history
  • GET /v1/stocktrends/portfolios/{port_id}/summary — portfolio history summary
  • GET /v1/stocktrends/portfolios/{port_id}/positions/history — closed-position history
  • GET /v1/stocktrends/portfolios/{port_id}/strategy — portfolio-to-strategy provenance
  • POST /v1/portfolio/compare — portfolio comparison
Start with /ai/tools Read API docs Understand ST-IM