Nested calendar clocks in long-bond futures

A rate market without a crop cycle

The case used long-bond futures as the market proxy for a 1979-1994 sample unless a table specified another window.

Unlike agricultural markets, interest-rate futures were treated as having no planting-harvest cycle. The search was for calendar biases tied to economic-report timing, federal refunding, and other rate-setting events.

A quasiSeasonalBias, in this workflow, is a calendar tendency inferred from report, refunding, tax, or holiday timing rather than from planting and harvest.

The weekday clock

Among weekdays, Friday had the highest share of both weekly highs and weekly lows, while Monday had the highest share of weekly low closes. Those counts identify the weeklyExtremeDay.

Each week was coded from Friday-to-Monday settlement direction, with Monday always labeled X and later sessions labeled X or O according to whether the close matched or opposed that direction. That label is the directionCode.

After Monday's coded close, the next session more often reversed that direction than continued it, with XO at 60.67% and XX at 39.33%.

The one-session reversal procedure entered market-on-close on Monday opposite the Friday-to-Monday direction and exited market-on-close on Tuesday, with no stops, slippage, or commissions applied.

Month-end and month-to-month clocks

A monthEndWindow ran from five sessions before the last trading day through five sessions into the next month. A bias was counted only when the indicated direction occurred at least 50% of the time with a signed close-to-close result of at least 1-00 point and no costs applied.

On a back-adjusted continuous contract, October, November, and December posted the highest shares of positive month-to-month closes at 62.5%, 75%, and 62.5%, while February, March, and April posted the highest shares of negative closes at 58.8%, 64.7%, and 58.8%.

Across 17 years, no calendar month monopolized the yearly high or yearly low, and April and July recorded no yearly lows.

Holiday and tax-date brackets

Holiday and mid-April windows were chosen by scanning long and short combinations from five sessions before to five after a fixed calendar date. That scan is how a holidayBracket was selected.

One listed combination was a June-contract short two sessions after April 15, covered four sessions after that date.

A chart condition as a later hypothesis

Editorial note: the approved facts specify calendar coding and window scans, not a crop-cycle chart analog. A repeatable chart condition can be stated as a seasonal hypothesis only after the weekday directionCode and the enclosing monthEndWindow or holidayBracket are named. That last step is editorial method order, not an archive finding.