Month-of-Year Seasonality in Trading EURUSD
Jan 18, 2026
What Happens When You Do the Analysis Correctly
Keywords: month of year seasonality, seasonality in trading, calendar effects trading, trading seasonality analysis, statistical power trading
Reading time: ~7–9 minutes
The Myth of Month-of-Year Effects
Month-of-Year (MOY) seasonality is one of the most persistent ideas in trading folklore.
Claims like “January is bullish” or “September is weak” are repeated so often that they are rarely questioned.
But repetition is not evidence.
This article documents a complete, statistically disciplined Month-of-Year analysis — and reaches a result that surprises many traders:
After proper testing, no calendar month shows a statistically significant return edge.
This is not a negative outcome.
It is a correct outcome.
Stage 1: Data Preparation and Calendar Alignment
The analysis begins with strict data hygiene:
Daily returns, not prices
Alignment by calendar month
Pooling the same month across all years
Treating each year as an independent observation
No smoothing, no pre-filtering
At this stage, the objective is simple:
Ensure January is always January.
No assumptions are made about profitability.
Stage 2: Descriptive Aggregation by Month
For each calendar month, we compute:
Average return
Standard deviation
Win rate
Number of years observed
This step is intentionally descriptive.
It answers:
What does the raw data appear to show?
Average montly returns boxplot shows that some months may look better or worse than others. This is exactly where most analyses stop — and exactly where mistakes begin.
Stage 3: Statistical Significance Testing
To test whether apparent patterns are real, each month’s average return is evaluated against the null hypothesis:
H₀: Average monthly return = 0
We compute:
t-statistics - a measure of how significant a result is, calculated by comparing the size of an observed effect to the amount of "noise" or random variation in the data.
p-value - is the probability that the results you observed occurred purely by random chance, assuming there is no real effect or difference.
A practical interpretation:
|t| < 1.5 → noise
1.5 ≤ |t| < 2 → weak / unstable
|t| ≥ 2 → statistically meaningful
|p-value| < 0.05 → the result is likely "real" and not just a fluke. Researchers often call this "statistically significant."
|p-value| > 0.05 → the result could easily be a coincidence or "noise," meaning there is no strong evidence of a real effect.
t-Stat by months barchart shows that there is no month crosses the ±2 threshold. This means: There is no statistically detectable Month-of-Year return edge.
Stage 4: Statistical Power: Why “No Signal” Often Means “Not Enough Evidence”
A failed significance test does not always mean no edge.
It often means low statistical power.
Statistical power depends on:
Effect size
Volatility
Sample size (years)
Significance threshold
In financial markets, effects are small — power matters more than intuition.
Key insight:
Most months have adequate sample size
Yet |t| remains low
This indicates absence of edge, not lack of data
This is an important distinction:
Low power + no signal → inconclusive
Adequate power + no signal → no edge
We observe the latter.
Stage 5: Quality Filtering Confirms the Result
To ensure nothing meaningful is missed, we examine quality dimensions:
Win rate vs t-statistic
Directional consistency
Economic magnitude
Some months win more often
Others have slightly higher returns
None combine frequency, magnitude, and statistical strength
This confirms the earlier conclusion.
Summary
This result does not mean:
Markets are random
Seasonality is useless
Research failed
It means something much more precise:
Month-of-Year alone is too coarse to capture persistent market structure.
Calendar effects that survive tend to exist at finer granularity:
Day-of-Month
Day-of-Year
Event-driven windows
Regime-dependent conditions
Month-level aggregation averages away these effects.
Key Insights
No calendar month shows a statistically significant return edge
Apparent patterns disappear under proper testing
Statistical power is sufficient to trust the result
Month-of-Year effects are weaker than commonly believed
Finer-grained seasonality is more informative
Conclusion
This study shows that Month-of-Year seasonality doesn’t survive proper testing.
That result isn’t the end — it’s the starting point.
When seasonality fails at the month level, it usually means the signal lives at finer resolution.
The next articles in this series zoom in:
Day-of-Month (DOM): flow-driven calendar effects
Day-of-Year (DOY): short seasonal windows that months average away
Same framework. Same statistical discipline. Higher resolution.
Month-level analysis removes false expectations.
Day-level analysis shows where structure actually appears.
Next: Day-of-Year seasonality
Then: Day-of-Month seasonality