When Sports Statistics Disagree: Check Definitions, Periods and Updates
A step-by-step check of the event, metric definition, period, timestamp and source version when two summaries show different numbers.
Different numbers do not always mean an error
Two counters can report different totals while each applies its own definition correctly. One source may record a corner when it is awarded; another when the ball is put into play. One key-pass system may count the final pass before any shot, while another may add assists or attach a modelled chance-quality value. Before comparing the numbers, establish what was counted.
This matters for team statistics that look identical in a broadcast graphic and an analytical table. Sports event dictionaries define attempts, successful actions, fouls, serves and other fields separately; similar terms can have different counting rules. The match statistics guide helps identify the intended metric, while the sports prediction checklist reminds you to verify data freshness and match context.
Five points to reconcile
Event: Is it the same match, player, team and competition stage? Check identifiers and current participant identity. Definition: Does an action count only when completed, on target, officially awarded or tagged in another way? Period: Full match, regular time, a half, extra time or an interval after an update? Time and version: When was the value observed, published and corrected? Source: A statistics feed, organiser’s report, broadcast graphic or your own video review?
Then check for rounding, duplicate events, time-zone differences or a changed participant identifier. For percentages, verify both numerator and denominator: pass accuracy may exclude some types of pass, while save percentage depends on what counts as a shot and a save. Compare raw events when available, not only the final total.
Example: a corner or a key pass
One match report may call a corner “won” when the referee awards the set piece; another field records the corner once it is taken. If the final whistle follows the award, the totals can differ without an arithmetic error. Confirm which metric the bet or model uses. Likewise, “key pass” may mean the pass immediately before a shot, while xA assigns a value to the pass using the chance that the subsequent shot becomes a goal. Those are different measurements, not two names for one event.
If the discrepancy changes the conclusion, keep both values and label their definitions, snapshot dates and reason for choosing one. Do not average them: the mean of incompatible scales has no clear interpretation. If market terms specify a settlement source, its criteria govern the bet; for price comparison, see the guide to identical markets.
Practical consequence
For trend analysis, use a continuous series based on one version of a definition. If the source or format changes, mark the boundary and do not join the series without checking equivalence. If primary events are unavailable, state a range of possible values or mark the conclusion as unstable.
For a bet, a discrepancy is a reason to check the market, settlement source and underlying data—not a reason to choose whichever number supports a preferred outcome. If the conflict cannot be resolved, avoid a strong conclusion and preserve the uncertainty in the record.