Spurious Correlation Alert

Overview

This pillar identifies strong statistical correlations that lack a logical, causal connection. It acts as a critical safeguard, preventing traders from making bets based on coincidental patterns or misleading data.

What It Does

The Spurious Correlation Alert scans for high correlations between market prices and unrelated external datasets. It then applies a series of checks, including searching for common causal variables and using a knowledge base to assess the logical plausibility of the relationship. The pillar flags any connections that appear statistically strong but are likely just random noise.

Why It Matters

Human brains are wired to see patterns, but many are meaningless in predicting future events. This pillar provides a disciplined, automated check against this cognitive bias, protecting capital from being risked on nonsensical relationships that fool other traders.

How It Works

First, the system calculates the Pearson correlation coefficient between a market's price history and thousands of external time-series datasets. For any correlation above a set threshold, it then searches for a third, confounding variable that could be driving both trends independently. Finally, it assesses the causal logic of the link, flagging it as spurious if no logical connection or common cause is found.

Methodology

The pillar identifies pairs with a Pearson correlation coefficient (r) where the absolute value is greater than 0.7 over a 90-day window. It then performs a Granger causality test to check for predictive information. A causal plausibility score is generated using a large language model to evaluate the semantic relationship between the two phenomena. A final 'Spuriousness Score' from 0 to 100 is assigned based on a weighted average of these checks.

Edge & Advantage

This provides a defensive edge by systematically filtering out popular but baseless narratives, allowing you to avoid trades that are built on statistical illusions.

Key Indicators

• Spuriousness Score

  high

  A 0-100 score indicating the likelihood that a detected correlation is non-causal and purely coincidental.

• Confounding Variable Flag

  high

  Indicates if a third, common-cause variable has been identified that better explains the correlation.

• Causal Plausibility

  medium

  A rating of how logically sensible the proposed causal link between the two variables is.

Data Sources

• Tyler Vigen's Spurious Correlations

  A database of known and humorous spurious correlations used for baseline comparisons.

• Google Trends

  Provides public interest data that can often act as a confounding variable for many market trends.

• Wikidata

  A structured knowledge graph used to check for logical or established relationships between entities and concepts.

Example Questions This Pillar Answers

• → Will the S&P 500 close higher this week if ice cream sales increase?
• → Will a candidate's odds of winning an election improve because their name correlates with a rising stock?
• → Will Bitcoin's price go up following a rise in Google searches for an unrelated meme?

Tags