SD Clustering Signal

The headline Pattern Analysis reading — how strongly today's AI stock predictions lean one way, graded in standard deviations (1, 2 or 3 SD) against their own history.

The Standard Deviation (SD) Clustering Signal is a technical breadth indicator designed to measure how strongly a suite of artificial intelligence stock predictions leans in a single direction on any given trading day. By analyzing the collective output of convolutional neural network (CNN) price predictions across a fixed universe of stocks, the system calculates daily breadth statistics such as the ratio of bullish predictions (Up ratio), the ratio of bearish predictions (Down ratio), average prediction confidence, and the count of high-confidence predictions. Each of these daily statistics is then compared against its long-run historical baseline to determine how many standard deviations it sits from its historical average. A daily signal is generated when a primary direction statistic (either the Up ratio or the Down ratio) crosses a threshold of 1, 2, or 3 standard deviations above its historical mean, provided there are no conflicting signals from opposing statistics. For example, an "UP" signal is triggered if the Up ratio clears its standard deviation threshold and no bearish statistic does. If both bullish and bearish statistics cross their respective thresholds simultaneously, or if only secondary confidence metrics cross, the day is classified as neutral. The final headline signal represents the highest clean standard deviation level achieved, such as "2 SD UP" or "1 SD DOWN". For retail investors, this signal serves as a gauge of market-wide momentum and systemic sentiment shifts. Because the thresholds are built on standard deviations, most trading days naturally result in no signal, representing normal market noise. When a 2 SD or 3 SD signal occurs, it indicates an extreme statistical outlier where the AI models are in rare, highly concentrated agreement. Investors use these rare readings to identify potential market turning points, breakouts, or overextended trends, though they must remain cautious as 3 SD events are exceptionally rare and based on small historical sample sizes.

Suppose the historical Up ratio for a curated stock universe averages 42% with a standard deviation of 10%. To trigger a 2 SD UP signal, the daily Up ratio must exceed the historical mean by at least two standard deviations: 42% + (2 10%) = 62%. If on a given day, the AI models predict that 63% of the stocks will rise, the Up ratio of 63% successfully clears the 62% threshold. Assuming no bearish statistics (like the Down ratio) exceed their respective thresholds, the day is graded as a '2 SD UP' signal.