How to Interpret Special Event Percentages

A denominator-first guide to the event-hunting page, sample size, and claims the dataset can support.

By Artimuz · Updated 2026-08-25 · 7 minute read

Every percentage needs a denominator

The event-hunting page counts the records in a group that carry an event and divides by the size of the group. If 2 of 5 records carry it, the page shows 40%. The "2 of 5" tells you more than the 40% does, because it also tells you how small the group is.

Change the Nightlord, the Shifting Earth or the pool and the denominator changes with it. Two percentages taken from different pools are not comparable, however close they look. Limveld records are counted nowhere on that page, and the DLC pool holds only the Balancers and Greg, so every denominator there is smaller than the dataset totals quoted elsewhere on this site.

The Old Gaol card is a useful example, because it puts two numbers next to each other that are not the same kind of number.

The Old Gaol card on the event-hunting page. Two figures sit at the top right, Base at 20.0 percent and DLC at 13.8 percent. Below them the base panel reads 40 percent, noted as one of ten pairings at that share, and shows Adel plus Crater at 2 of 5 with links to seeds 065 and 066. — 20.0 percent is 32 records out of the 160 base records on a Shifting Earth. The 40 percent beside it rests on five records. The page prints the 2 of 5 so you can see how little is holding it up, and that ten other pairings match it exactly.

Seed 065 open in the result view, a fully identified Crater map with the Old Gaol encounter circled in red around Godskin Duo. Every other location carries a label naming what it holds, and the Nightlord medallion sits at the bottom left. — This is one of the two seeds behind that 40 percent, opened by clicking 065 on the card above. The record still carries no special event, because the Old Gaol is an encounter on the map rather than an event card.

This is a share of records, not the odds the game rolls

All the page does is count. What it cannot tell you is how the game picks among those records, because nobody outside FromSoftware knows that.

So read every number on it as a share of known records, not as per-run odds. The distinction bites hardest when two options are close: 22% against 20% in the dataset is no reason to expect one more often than the other over your next ten runs.

Small groups swing wildly, so read the fraction

One more matching record turns 1 of 5 into 2 of 6, which is 20% becoming 33.3%. That same single record moves 20 of 100 by less than a percentage point. A big percentage sitting on a tiny denominator is not a strong result. It is a small sample.

This is why the page prints the fraction next to the pairing it shows, and says how many pairings share a top figure when more than one does. In the base pool each Nightlord and Shifting Earth pairing holds exactly five records, so a top share of 20 percent means a single sighting, and anywhere from three to ten pairings can be sitting on that same figure. Treat the one shown as somewhere to start, not as a rule.

The pairing shown is settled by Nightlord order, earliest first, so a tie in the base pool resolves to Gladius before Adel before Gnoster, and a tie in the DLC pool resolves to the Balancers before Greg.

Picking a combination to hunt

Decide first whether the DLC is available to you, because that settles which of the two panels you are reading. After that, weigh the share against how many records are behind it. A slightly lower share with a larger sample is usually the better bet, and the larger samples on that page are the Great Hollow pairings for the Balancers and Greg, which hold twenty records each where everything else holds five.

The two panels do not cover the same Nightlords. The base panel is the eight base Nightlords; the DLC panel is the Balancers and Greg alone. The eight do appear in DLC-range records, but the DLC panel leaves them out, so it compares DLC Nightlords against each other rather than against the base roster.

If you want to know the real frequency, the only way is to keep a log of your own runs: the pairing you ran, the seed you got, whether the event turned up, the date and the game version. Do not pool runs from either side of a patch without marking where the boundary was.