Probabilistic market intelligence built from structured market classifications

The Stock Trends Inference Model, or ST-IM, is designed to answer a practical question: when a stock is classified by the Stock Trends system in a particular market condition, what has historically tended to happen next?

ST-IM does not assume that markets are perfectly predictable. It begins from the opposite premise: markets are noisy, uncertain, and highly influenced by randomness. The model attempts to extract useful probabilistic information from that uncertainty by organizing very large populations of historical observations into meaningful Stock Trends classifications.

The core idea

The model treats each weekly Stock Trends observation as part of a larger statistical population. Every stock has a market condition: trend classification, trend age, relative strength, recent outperformance or underperformance, and volume behavior.

These classifications convert raw price and volume data into structured factor variables. Once those factor variables exist, the model can study what forward returns historically followed from similar conditions.

In simple terms:

• Markets are uncertain.
• Individual stock outcomes are noisy.
• Large populations of classified observations can reveal useful return distributions.
• Those distributions can support better screening, ranking, portfolio construction, and market interpretation.

Why the Stock Trends classification system matters

ST-IM is only possible because the Stock Trends system classifies market behavior systematically. The classification framework transforms weekly trading data into repeatable market states.

Instead of asking only whether a stock went up or down, the model can ask more precise questions:

• Was the stock Bullish, Bearish, Weak Bullish, Weak Bearish, or in a crossover condition?
• How long has that trend condition persisted?
• Was the stock outperforming or underperforming its benchmark?
• Was relative strength above or below the market baseline?
• Was volume unusually high, unusually low, or normal?

This classification structure gives the model meaningful sample spaces. Without classification, the historical population is much noisier and harder to interpret.

Randomness, normal distributions, and large populations

ST-IM is grounded in the recognition that market outcomes are partly random. A single stock may rise or fall for reasons that cannot be fully predicted. But across a large enough population of similarly classified observations, distributional patterns can become more informative.

This is where the central limit theorem and the normal distribution become useful reference points. As observations accumulate, the distribution of average outcomes can become more stable, making it possible to compare expected forward returns, probabilities, and confidence intervals across different Stock Trends conditions.

The objective is not to eliminate uncertainty. The objective is to measure it, compare it, and use it intelligently.

What ST-IM attempts to do

ST-IM estimates forward-return probabilities and return distributions conditioned on Stock Trends classifications.

It attempts to answer questions such as:

• Do stocks in this condition historically outperform a random stock benchmark?
• Is the expected return distribution meaningfully better or worse than the baseline?
• Which forward horizon shows the strongest probability profile?
• Does the current market structure support a bullish, bearish, neutral, or selective stance?
• Which stocks appear more attractive when ranked by probabilistic return expectations?

Forward horizons

ST-IM focuses on practical investment horizons that match the Stock Trends methodology:

These horizons allow agents and portfolio systems to distinguish short-term opportunity from intermediate-term and longer-term structural strength.

What the probabilities mean

ST-IM probabilities are not guarantees. They are estimates based on how similarly classified observations behaved historically.

A probability above 50% does not mean a stock must outperform. It means that, in the relevant historical sample, similar conditions produced a favorable outcome more often than the random benchmark.

Agents should interpret probabilities as decision-support signals, not as deterministic predictions.

STIM Select

STIM Select is a stricter selection framework built on top of ST-IM. It identifies instruments whose probability and confidence profiles meet stronger criteria across forward-return horizons.

The purpose of STIM Select is not to find every possible opportunity. It is to filter for cases where the historical return distribution appears more favorable than a broad random baseline.

In portfolio workflows, STIM Select can help agents:

• reduce weak statistical candidates,
• rank opportunities by probabilistic strength,
• screen for higher-confidence portfolio candidates,
• combine stock-level probabilities with regime and breadth context.

Strengths

• Interpretable: based on named Stock Trends classifications rather than opaque signals alone.
• Probabilistic: frames market outcomes as distributions, not certainties.
• Systematic: applies repeatable rules across large populations of observations.
• Portfolio-friendly: becomes more useful when applied across diversified groups and repeated decisions.
• Agent-ready: provides structured outputs that can be used in screening, ranking, allocation, and workflow orchestration.

Limitations

• ST-IM does not guarantee the performance of any individual stock.
• Historical relationships may weaken or change in future market regimes.
• Tail events, news shocks, liquidity events, and macro surprises can overwhelm statistical expectations.
• Small or narrow sample spaces may be less reliable than broad populations.
• Probabilities should be combined with portfolio discipline, risk controls, and market context.

These limitations are not weaknesses of the framework; they are part of honest probabilistic modeling. The model is most useful when its uncertainty is respected.

How agents should use ST-IM

AI agents should use ST-IM as a probabilistic reasoning layer inside a larger workflow.

• Use /v1/ai/tools to discover workflows.
• Use /v1/meta/stim and /v1/ai/context for conceptual grounding.
• Use /v1/stim/latest for instrument-level probability data.
• Use /v1/selections/latest and /v1/selections/published/latest for filtered candidates.
• Use portfolio endpoints to evaluate groups, not just individual securities.
• Use market regime, breadth, and leadership endpoints to add broader context.

Portfolio management applications

ST-IM is especially useful in portfolio management because probabilities can compound in value across diversified and repeated decisions.

A single favorable probability does not create certainty. But a systematic portfolio process can use many probability-weighted decisions to improve selection quality, avoid weaker candidates, and adapt to changing market structure.

• screening and candidate generation,
• ranking and filtering,
• portfolio construction,
• portfolio comparison,
• regime-aware exposure control,
• monitoring leadership and breadth changes.

Understanding the market more broadly

ST-IM is not only a stock-selection tool. It also helps interpret the market itself. By aggregating classified observations, the Stock Trends API can reveal broad market regime, sector breadth, leadership concentration, and the maturity of bullish or bearish conditions.

This makes the API useful for agents that need to reason about market environment, not merely rank individual instruments.

Related API endpoints

• GET /v1/ai/tools — primary agent discovery
• GET /v1/ai/context — indicator and dataset context
• GET /v1/meta/stim — ST-IM metadata
• GET /v1/stim/latest — latest ST-IM data
• GET /v1/stim/history — historical ST-IM data
• GET /v1/selections/latest — latest selection candidates
• GET /v1/selections/published/latest — published STIM Select candidates
• POST /v1/portfolio/construct — construct portfolio workflow
• GET /v1/market/regime/latest — current market regime
• GET /v1/breadth/sector/latest — sector breadth context
• GET /v1/leadership/summary/latest — leadership context

Further reading

For a broader explanation of the Stock Trends methodology and systematic investing framework, read the Stock Trends Handbook.