How to Interpret Special Event Percentages
A denominator-first guide to the event-hunting table, sample size, and claims the dataset can support.
By Artimuz · Updated 2026-07-28 · 7 minute read
Every percentage needs a denominator
The event-hunting table counts how many compatible seed records contain an event and divides that by the number of records in the same comparison group. If 3 of 15 records match, the displayed proportion is 20%. The “3 of 15” carries more information than 20% alone because it exposes the sample size.
Changing Nightlord, map, or the base-game/all-content scope changes the denominator. Percentages from different groups should not be compared as if they came from the same population.
Dataset share is not proven selection probability
The calculation assumes only that records can be counted. It does not establish that the game selects every compatible seed with equal frequency. Without a large sample of observed runs or documented selection rules, the result should be read as “share of known records,” not guaranteed per-run odds.
This distinction matters most when choosing between close values. A combination showing 22% in the dataset is not necessarily measurably better in live play than one showing 20%, especially when both groups are small.
Small groups produce unstable rankings
One additional matching record changes 1 of 5 from 20% to 2 of 6, or 33.3%. The same extra record changes 20 of 100 by less than one percentage point. Small denominators therefore create dramatic-looking differences that may not persist when the dataset changes.
Use the matching-record count as a confidence cue. A top combination with only a few compatible records is a lead worth testing, not a universal rule.
A practical hunting decision
First choose whether DLC Nightlords and maps are available to you. Then compare the highest dataset share, the number of matching seeds, and how quickly you can repeat that expedition. A slightly lower share may be the better practical choice if the setup is familiar or the sample is much larger.
Keep a simple run log if you want to test real-world frequency. Record the chosen combination, result seed, whether the event appeared, date, and game version. Do not mix runs from different patches without noting the boundary.