Understanding Our Assumptions — A Decade Later

1. Revisiting the First Premise: Market Conditions Are Non-Specific

In 2014 Stock Trends Editorial, Understanding Our Assumptions, I argued that when stocks share equivalent combinations of Stock Trends indicators—trend category, trend length, relative strength (RSI), volume tag—they occupy comparable “market conditions”, even if they differ in size, industry, or liquidity.

That premise still stands. In fact, the last decade has offered even stronger support.

Why this holds true today

Across global markets, researchers have repeatedly shown that trend and momentum behaviour is universal, cutting across:

• Mega-caps and micro-caps
• Financials, technology, industrials, energy
• U.S., Canada, Europe, emerging markets
• Even across asset classes such as bonds, currencies, and commodities

This is not a narrow anomaly—it is one of the most persistent, geographically and temporally robust behaviours ever documented in financial economics.

The modern empirical foundation

Academic studies since the 1990s and through the 2020s continue to confirm that:

• Cross-sectional momentum in equities remains strongly present: baskets of recent winners have historically outperformed baskets of recent losers over intermediate horizons.
• Time-series momentum (trend-following) persists across more than a century of data and multiple asset classes, with assets tending to continue in the direction of their prevailing trend for several months before mean reversion asserts itself.

This is precisely the logic embedded in the Stock Trends classification system: a Bullish () trend with a high mt_cnt trend counter and strong RSI exhibits similar statistical tendencies across thousands of historical observations, regardless of the company name attached to it.

Nuance for 2025

In updating the premise, we now acknowledge:

• The form of market response is broadly comparable across securities.
• The magnitude of response varies with liquidity, volatility, and sector.
• Regime effects—low-rate eras, inflationary cycles, volatility shocks—modulate how strongly patterns perform.

But the essential insight remains resolutely supported: trend and momentum conditions are transferable patterns, not idiosyncratic attributes of individual stocks.

2. Revisiting the Second Premise: Market Responses Are Specific

If market conditions can be categorized, then subsequent returns become the response variable.

This is the backbone of the Stock Trends Inference Model:

• Group identical indicator combinations across decades of history.
• Measure the forward return distribution for each combination.
• Assume a probability distribution (normal approximation or empirical distribution).
• Compute the likelihood that a stock under those conditions will outperform a random stock’s base return.

This transforms technical analysis into statistical inference.

The academic parallel

Much of modern factor investing takes the same approach:

• Past returns, trend slopes, volatility regimes, and momentum windows are treated as categorical or continuous signals.
• Forward returns are evaluated conditionally on those signals.
• Probabilities, not point predictions, emerge as the output.

In this light, the Stock Trends Inference Model was an early practitioner framework built on the same empirical foundations that now underpin global factor indices.

What has strengthened this premise since 2014

• Stronger behavioural explanations. Investor under-reaction, herding, anchoring, and limits to arbitrage have all been further validated as drivers of persistent technical patterns.
• Recognition of randomness. Factor literature now explicitly recognizes that momentum and trend-following are noisy signals. They fail often, they experience deep drawdowns, but they work in expectation over long horizons and repeated exposure.
• Crisis behaviour research. Trend-following often performs strongly in stressed environments, providing “crisis alpha”. This mirrors our observation that extended Bearish trends and high-volatility drawdowns exhibit repeatable statistical characteristics.

Together, these developments reinforce the premise that return responses to trend and momentum conditions can be measured, ranked, and utilized probabilistically.

3. Embracing Randomness: The 2014 Insight That Aged Perfectly

In the original editorial, I wrote:

“The Stock Trends Inference Model attempts to reconcile with randomness by assuming it implicitly.”

This is now a mainstream view in quantitative finance.

Rather than treating technical signals as directional forecasts, Stock Trends treats them as:

• Sampling windows into historical behaviour.
• Probability distributions of future outcomes.
• Inputs to ranking systems rather than deterministic triggers.

In 2025, with the rise of AI-driven market models, this framing is even more essential. The stock market is inherently noisy—but not structureless. Certain conditions consistently tilt the odds.

This is exactly what ST-IM measures.

4. How These Assumptions Explain Today’s Market Behaviour

To appreciate how well the 2014 premises continue to function, consider several modern examples.

A. NVIDIA and the AI Megatrend

Few stocks better illustrate the persistence of trend and momentum than NVIDIA during the 2022–2025 AI cycle.

• Sustained Bullish () and Bullish Crossover () classifications
• Extended periods of RSI > 100, signaling durable relative strength
• Heavy volume confirmation of the trend
• Shallow, quickly reversed corrections within the primary uptrend

These are the exact pattern combinations which, historically, exhibit:

• High positive mean forward returns
• Elevated probability of outperforming the “random stock” benchmark
• A tendency to maintain leadership through medium-term horizons

If the premises of 2014 were wrong, NVIDIA’s behaviour would look random or chaotic. Instead, it is textbook Stock Trends.

B. Sector Rotations and Trend Distributions (2020–2025)

The past five years produced violent rotations:

• Technology to Energy (2021–2022)
• Energy back to Tech and megacap Growth (2023–2025)
• Cyclicals vs Defensives through rate cycles
• Small-cap vs large-cap divergences

In every rotation, sector-level Stock Trends distributions shifted coherently:

• Rising Bullish counts in the dominant sectors
• Contraction of Bullish momentum in the losing sectors
• Distinct RSI signatures during leadership transitions

The same principle from 2014 applies:

If indicator combinations encode comparable market conditions, then rotations become measurable shifts in population states—not mysterious market whims.

C. Trend-Following in the Inflation Shock (2022)

Global trend-following indices delivered strong performance during the inflationary regime shift of 2022, capturing:

• Downtrends in bonds
• Uptrends in commodities
• Strength in the U.S. dollar
• Divergences within and across equity markets

Again, markets separated into clear, statistically coherent patterns.

This validates the Stock Trends view beyond equities alone: price reacts predictably to shared behavioural mechanisms, regardless of the underlying asset.

5. Updating the Model for 2025 and Beyond

The core assumptions remain robust. But our understanding—and our tools—have advanced.

Here is how the updated Stock Trends Inference Model now interprets its premises:

1. Market conditions are comparable, but not identical

We now incorporate:

• Sector context
• Liquidity differences
• Regime sensitivity
• Trend “maturity” through mt_cnt
• Momentum quality through RSI and additional momentum variables

2. Responses are probabilistic, not predictive

ST-IM outputs include:

• Confidence intervals for inferred mean returns
• Probabilities of beating benchmark returns
• Mean–variance estimates
• Outlier risks and tail scenarios
• Scenario distributions under different market regimes

These feed into:

• Select Lists and probability-based rankings
• Performance evaluation of Stock Trends Reports like Picks of the Week
• Causal AI research on drivers of outperformance
• Machine learning forecasts
• The Stock Trends Game, including the Investor Challenge

3. Randomness is not the enemy — it is the medium

Technical patterns work because:

• Investors are human.
• Reactions are delayed and often exaggerated.
• Information diffusion is uneven.
• Capital constraints limit arbitrage.

The market is noisy, but it is systematically noisy.

6. Conclusion: The 2014 Foundations, Strengthened by a Decade of Evidence

The assumptions introduced in 2014 were not speculative—they were early articulations of concepts that would later become bedrock ideas in quantitative finance.

Since then:

• Momentum research has expanded and deepened.
• Trend-following has proven resilient across multiple crises.
• Behavioural explanations have grown stronger and more nuanced.
• Regime-aware analysis has become essential.
• AI has amplified, not replaced, the need for structured inference.

Today’s Stock Trends Inference Model stands on the same pillars as before, but with sharper tools and deeper data.

The original question—What do market conditions imply about forward returns?—remains at the heart of Stock Trends.

And the answer is clearer than ever:

Patterns matter. Probabilities matter.
And in a noisy world, disciplined inference is indispensable.