Scroll investigation · Dataset essay
The calendar is not a catalyst.
Monthly averages can reveal recurring structure. They can also invite a forecast the data never made. Eight markets, twelve months, and one rule: frequency, magnitude, and sample size must agree before a pattern deserves attention.
Green is positive or above 50%. Red is negative or below 50%. Amber is never used for direction.
Averages make a persuasive picture.
The first view colors each cell by its mean calendar-month return. Repeated bands are easy to see, but a mean can be pulled by a few exceptional years. The picture is a summary of realized history, not a trading instruction.
Ask how often, not only how much.
The second view replaces magnitude with hit rate. A positive average backed by only half the observations is a different claim from a modest average that appeared repeatedly. Neither is sufficient on its own.
One month does not mean one regime.
In this build, September ranges from -0.38% for EFA to -1.95% for TLT across the selected markets. Calendar labels do not erase differences in asset structure, listing history, or economic exposure.
The sample is part of the result.
The final view shows observations. Newer listings cannot carry the same evidentiary weight as seasoned instruments. Even fifteen observations remain a small sample, so the responsible conclusion is descriptive: seasonality can frame a question, never settle it.