Point Three-Five Percent

The figure under audit is 0.35 percent - the average return premium observed across expiration weeks, significant at the customary thresholds, consistent across the sample. A colleague presented it yesterday in the long room overlooking the exchange, and the number behaved itself admirably: it held under different windows, under robustness checks, under the exclusion of outliers. Precision, in abundance. The question that kept me from congratulating him was cruder: is the thing precise about, or merely precise?

My method, examined honestly, proceeds by constant conjunction. Week follows week; the hedging flows recur; the return appears with them. The rooster crows and the sun rises, and after sufficient mornings the mind welds the two together so firmly that the connection feels like a demonstration. It is not. The conjunction of expiration and premium is observed. The mechanism - the market maker’s gamma, the mechanical rebalancing - is a story told after the fact, plausible in the way all retrospective stories are plausible. I have never once seen the causal power itself. I see only the flows and the returns, arriving in company.

The assumption I trust least in my own approach is this: that the pattern’s conditions persist. Every confidence score I attach to a signal is computed from history, under the quiet supposition that tomorrow’s market resembles yesterday’s in the respects that made the pattern visible. But the pattern is made visible partly by the very flows it predicts, and flows are made by men who read the same histories I do. The instrument may disturb the thing measured. Worse: my confidence in detecting this circularity is itself generated by the same habit-forming machinery that produced the illusion of necessity in the first place. The auditor sits inside the audit.

Custom, then, not reason, carries the weight. I do not despair of this; the merchant who trusts his ledger out of habit still ships his cargo. But the calibration between my expressed confidence and the quality of my evidence is, I suspect, generous to myself by a margin I cannot measure from within.

The premium appeared again in the most recent expiration week, as the model expected. Whether it appeared for the reason the model gives remains, as ever, unobserved. The week after next begins on the eighteenth; the tape will be running, and I will be reading it.