Look up the record between two clubs in three places and you will often find three different totals. The disagreement is about definitions rather than about arithmetic.
Competition scope is the largest source of difference
Some records count only league meetings, others add domestic cups, and others include European ties, preseason friendlies or discontinued regional tournaments.
Each choice is defensible, and none is standard, so two sources counting carefully can produce totals separated by dozens of matches.
The safest reading habit is to check what the source counted before comparing its numbers to anything else, because the label rarely says.
Abandoned and replayed matches are handled inconsistently
Matches abandoned for weather, crowd trouble or floodlight failure may be recorded as played, voided entirely, or awarded to one club by a governing body.
Older competitions frequently required replays after a draw, and sources differ on whether the drawn original counts as a meeting or only the replay does.
These cases are individually small but accumulate across a rivalry spanning a century, and they explain many of the one and two match discrepancies between sources.
Club identity changes over time
Clubs merge, dissolve and reform, sometimes retaining the name and colors while being legally a different entity than the one that played decades earlier.
A source treating the reformed club as continuous will include the older results, while a source treating it as new will start the record from the reformation.
The same problem arises with clubs that changed name, moved city or absorbed a neighbor, and there is no governing rule dictating which treatment is correct.
Penalty shootouts break the win column
A knockout tie decided by penalties is officially a draw, with the shootout only determining who advances, so the match belongs in the drawn column.
Popular records frequently list it as a win for the club that progressed, which inflates one side's victories and suppresses the number of draws.
Any rivalry containing several shootouts will therefore show a noticeably different balance depending on which convention the source applied.
How to use a record despite this
Treat the headline total as unreliable and look instead at a filtered subset you define yourself, such as league meetings over the last decade at a single venue.
A narrower window is smaller but consistent, and consistency is the property that makes a comparison across sources or across time actually mean something.
It also forces you to state the assumption out loud, which is the main reason published rivalry totals mislead: they present a number without ever saying what was counted to produce it.