A decision-logic filter for the July-November soybean spread

A spread is a price relationship

A commodity spread is the changing price relationship created by buying one futures contract and simultaneously selling another related contract.

An intercommodity spread pairs related but different markets. An intracommodity spread, also called an interdelivery spread, pairs two delivery months of the same commodity. July versus November soybeans is the common example.

Spread positions can carry higher commission overhead and two-sided risk. They are described as using generally lower margin, smaller day-to-day price change, and more seasonally regular behavior than an outright futures position.

July versus November as demand

Because soybeans are typically planted in March and harvested in November, the July contract is treated as old-crop inventory and the November contract as new-crop supply. The July-November spread measures immediate demand versus later demand.

When immediate soybean demand is high, the July contract is described as gaining more or losing less than November. When immediate demand is low, July is described as lagging November on rallies and weakening relative to November.

A weekly close can accept or skip the window

A seasonal window is a calendar span, here including the twenty-first through twenty-fifth weeks, in which a soybean spread is historically more likely to follow a recurring crop-cycle pattern.

A decision-logic filter is a weekly higher-close or lower-close rule that accepts, rejects, or delays a seasonal spread trade. Six decision-logic seasonal systems, labeled SS-1 through SS-6, apply a higher-week or lower-week close in a specified week to a sell signal. Holding periods are two to four weeks, and the archive tabulated win, loss, and average profit or loss counts.

SS-4 uses a lower weekly close in week 21 as a sell filter and holds four weeks. It is later compared with executing the same seasonal window every year.

Historical samples and a chi-square check

A chi-square statistic is a goodness-of-fit figure reported on the seasonal and filtered trade samples. It compares observed win and loss counts with an explicit quantitative baseline.

In the summarized seasonal-window sample, unfiltered seasonal trades numbered 28 with 20 profits, while decision-logic trades numbered 13 with 11 profits. The corresponding chi-square statistics were 4.32 and 4.92.

For the twenty-first through twenty-fifth week window, the unfiltered seasonal sample again had 28 trades and 20 profits, while the SS-4 decision filter produced 15 trades and 12 profits. The chi-square statistics were 4.32 and 4.27.