CURRENCY CORRELATION — REPRODUCIBLE CHECK RECORD
An educational record, not a trade recommendation. Save a copy for each calculation.

1. Question and data
Question: ______________________________
Series A (base/quote): ______  Series B (base/quote): ______
Source, retrieval date and permission to use the data: __________________
Price basis (Bid / Ask / midpoint (Bid+Ask)/2 / last): ______
Timestamp meaning: ______  Timezone: UTC  Interval: exactly one hour
The supplied table must contain 4–128 consecutive hourly observations, with both positive prices on every row. Do not disguise weekends or missing hours as one-hour changes. Record exclusions or preparation separately; the lab does not sort, join or fill gaps.

2. Calculation settings
First and last included timestamps: __________________
Trailing return count N: ______  Required prices: N+1
A inverted? ______  B inverted? ______
Log return = ln(current price / previous price). Pearson measures linear association between paired returns. Three returns is only the tool minimum, not evidence of adequate sample size.
Return correlation: ______  Price-level correlation (different statistic): ______
If undefined, record zero/numerically flat variance instead of writing zero.
Inspect the scatter and table; note outliers: __________________
Preselected comparison window and result: __________________
Save the valid CSV with its assumptions. Editing inputs hides a previous result until calculation succeeds.

3. Separate currency map
For each of two positions, record pair, long/short, quantity in base units (not lots), reference price in quote units per base unit, and why that price was chosen.
Long Q base/quote at P: +Q base, −Q×P quote. Short reverses both signs.
Add signed contributions only within the same currency; gross is the sum of their absolute amounts. Never total different currency units.
Reference example: long 10,000 EUR/USD at 1.10 and short 10,000 GBP/USD at 1.25 gives EUR +10,000, GBP −10,000, USD +1,500 net; USD gross 23,500.
This is conceptual notional exposure, not cash balances, margin, P/L or an effective hedge.

4. Interpretation and limits
Zero correlation does not establish independence. Inverting the quotation of one series reverses its log-return signs and the sign of correlation, when defined; inverting both preserves the correlation sign. Correlation is neither a hedge ratio nor the probability of joint losses. No live prices or trading signals are supplied.
All bundled data are fictional. The default has 17 prices and 16 returns from 2024-01-08 00:00 to 16:00 UTC: r ≈ 0.026687804. The final eight returns give r ≈ −0.998691088. These are teaching results, not market estimates.
Conclusion supported by these observations only: __________________
Unanswered questions / conditions that would change the conclusion: __________________

Files: fictional-hourly.csv; reference-results.csv; exposure-reference.csv.
Displayed values are rounded. Keep the original CSV values when recording or checking a calculation; small differences in the final digits do not change the stated example.