Overlapping game-history exports: remove duplicate records by identity
Two history exports can contain the same round when their date windows overlap. Appending both files and summing every amount would count that round twice. Removing every repeated amount is not a safe fix either: two different rounds can legitimately have the same stake and return.
For the topic overview, see Learn the details before the spin.
The unit to identify is the recorded event, using its reference and context—not merely the number in the amount column.
Keep the source files unchanged
Work in copies and record each export’s account scope, currency, time zone and date range. The UK account-history standard is a useful reference for the distinction between an account balance and its historical records. It does not define the fields in every local export.
In the invented example below, each row is one completed round-level record. If an actual system exports separate stake, award and refund events under one round ID, those are not automatically duplicate rows.
Find overlap without deleting a real round
| Export | Round reference | Stake | Return |
|---|---|---|---|
| A | R101 | 10 | 4 |
| A | R102 | 10 | 0 |
| B | R102 | 10 | 0 |
| B | R103 | 10 | 0 |
R102 appears in both files with the same completed data. R103 has identical amounts but a different identity. Keeping R101, R102 and R103 once gives stakes of 30 and returns of 4. Adding all four rows gives 40 and 4, overstating the loss by 10. Deleting all equal 10/0 rows would wrongly remove a real round.
A shared reference can also reveal a revision
Now imagine the second copy of R102 has status pending rather than completed, or a different return. That is a conflict to resolve, not a row to discard silently. Check the source timestamps and documented export meaning. A later file might reflect a revision, but “later” alone does not explain which field changed or why.
Keep an audit note showing both records and the reason for the chosen version. If the conflict cannot be resolved, exclude it from a claimed final total and report the unresolved item separately.
Choose a key that fits the actual data
A reference may only be unique within one provider, account or game. Preserve those qualifiers rather than assuming R102 is globally unique. For event-level data, include the transaction type and event reference when the system supplies them.
After deduplication, compare row counts and total amounts with the originals. The goal is a transparent reconciliation, not a better-looking session result. Even a perfectly cleaned history describes recorded activity; it does not reveal the game’s complete probability model or predict future outcomes.
The Jili777 homepage brings together the site’s other reading topics.
Source-based editorial research and original examples. Sources checked 9 October 2026. No real-money play or account transaction was performed for this article.