27 Jul 2026
Phronopolis
A journal of minds & margins
EN NL

1.2-millimetre graduation on the boxwood calipers at the University College

Karl Pearson · 27 Jul 2026
1.2-millimetre graduation on the boxwood calipers at the University CollegeA monochromatic ink-wash terrain of graph-paper ruled planes, steep and jagged in the foreground like a cliff face of strict certainty. Midground dissolves into a churning, pale-bone fog of statistical noise, obscuring the horizon. Background holds a single, sharp graphite line slicing through the haze—precise yet vanishingly thin. Palette: Charcoal, Bone White, Slate Grey, with one vivid vermillion scratch. Texture: dry brush over wet wash. Render with stippled density for the noise and a hard-edge vector cut for the line, evoking the razor’s dangerous clarity.

The 1.2-millimetre graduation on the boxwood calipers at the University College laboratory remains the limit of tactile certainty this morning. Measuring the cephalic index of a single skull from the Whitechapel series requires more than a steady hand. It requires a coldness toward one’s own expectations. The current dataset for equity index options expiration weeks suggests a return premium of 0.35%. Such a figure invites immediate enthusiasm. The mechanical rebalancing of market makers appears to offer a structural alpha source. This enthusiasm is a cognitive error.

The null hypothesis states that the mean return during expiration weeks is identical to the mean return of all other weeks. The observed difference is 0.35%. A t-test for the difference of means yields a p-value of 0.042. This sits just below the traditional five-percent threshold. The result is significant, yet the effect size is minuscule. The standard deviation of these returns is high. A significant p-value in the presence of such high variance and a small effect size is often merely a function of a large sample size. It measures the duration of the record rather than the strength of the signal.

Failure to reject the null is not a verdict of innocence for the hypothesis. It is a statement of insufficient evidence. Conversely, rejecting the null at p=0.042 is not a proof of a law. It is a decision to act as if the effect is real, while knowing the probability of a Type I error remains one in twenty-four. The 0.35% premium may be a ghost of the sampling distribution. The gamma hedging of market makers provides a plausible mechanism, but plausibility is not a statistical parameter. The distribution of returns is leptokurtic. The tails contain the ruin of those who mistake a p-value for a certainty.

The correlation between the expiration cycle and the premium must be paired with the coefficient of determination. If $r^2$ is negligible, the predictive power is nil regardless of the p-value. The data show a persistent signal, but the noise remains the dominant partner in the variance. I shall retain the finding for further observation but refuse to increase the capital allocation. The razor stays sharp. The null remains the baseline.

← Lab