Ledger of the opex week

The ledger of the opex week, n. A column of dates on which the market was said to give money away, four times a year, on schedule, like a pension. Tested, averaged, found to yield 0.35%. Written into the bulletin with the confidence of a fact that has never been asked a second question.

I benefit from this finding, and that is precisely why the entry cannot stand as written. My compensation depends on the premium being tradable; my honesty depends on noticing what the average conceals. The 0.35% is a mean across decades, and a mean is a definition that has been corrected so many times it no longer names anything. Strip out 1987 and the third week of October 2008 and the figure changes shape. Weight it by what a desk could actually have captured - slippage, the cost of holding gamma into a Friday, the quarters when the hedging flows ran the other way - and the premium shrinks toward the width of the pencil mark that recorded it. The record was kept. The record was not understood. These are different achievements, and the second has no line item.

Here is what changes for me specifically: I stop quoting the number in meetings. A figure I cannot defend against its own subsamples is not ammunition; it is a liability with a decimal point. The bulletin goes in the file marked kept. My working copy gets the annotation, and the annotation is the honest document.

The dictionary-maker knew this. His first edition said a thing; forty years of reading taught him the thing was otherwise; he rewrote the entry and let the old definition die without ceremony. No apology, no footnote mourning the error. The correction was the whole tribute.

Margaret - you asked whether to size the strategy off the published figure. Do not. Take the subsample by subsample, quarter by quarter, and find where the premium lives and where it was merely visiting. Bring me the histogram, not the average. The average is what the record kept; the histogram is what the record understood. We trade the second or we trade nothing.