Killer, Decisions & Replay Planning

The confirmed replay structure

Steam says every playthrough chooses a supernatural serial killer from a pool of candidates. The selected killer can change the campaign's reasoning, kill style, and targets. The same listing also says decisions are checked and autosaved, so you should treat important choices as persistent until the finished game proves that a branch can be undone.

This is not a promise that every outcome is permanently missable. It is a practical rule for a game whose central mystery changes from run to run: write down what happened instead of reloading on instinct.

Keep a spoiler-safe suspect ledger

Use one row per resident with only observations in the first pass:

Field	Record
Identity	Name, home, work, and any confirmed relationship.
Routine	Where and when you saw the resident, including season.
Evidence	Direct clue, second-hand statement, or only a suspicious coincidence.
Decision	What you chose and the immediate result.
Confidence	Confirmed fact, observed event, interpretation, or unknown.

Keep interpretation in its own column. A resident being absent from a location is an observation; calling that absence proof of guilt is a theory. This distinction will make a second run much easier to compare.

Log decisions before the result

Before selecting a major dialogue option, note the date, location, people present, and the question you are trying to answer. After the scene, add the result. Because the game advertises checked and autosaved decisions, this record is more useful than a series of reloads that hides the branch logic.

• Mark whether the choice changed a relationship, a clue, a location, or only the dialogue.
• Use a separate tag for romance decisions; do not assume romance and investigation outcomes are independent.
• Capture the season and time of day. The 24-hour clock and changing seasons can alter what you can observe next.
• When a case closes, list the unanswered questions before starting another route.

Romance without losing the mystery

The official store page lists 12 potential romanceable people and warns that a few may be the killer. That makes relationship progress a source of character context, but also a spoiler boundary. Keep a romance notebook separate from the public suspect summary, and label any guide screenshot that reveals a route or outcome.

Pre-release material does not confirm which romance options belong to which killer scenario, how affection is earned, or whether a relationship can be repaired after a decision. The [romance guide](/romance-guide/) tracks confirmed scope and will need launch-build route testing.

Compare runs instead of chasing one perfect run

1. Finish a first run while recording observations, not spoilers.
2. Copy the farm, season, and resident baseline into a new sheet.
3. Compare only changed clues, targets, routines, and decision results.
4. Test one hypothesis at a time so a new killer scenario is not confused with a changed dialogue choice.

The useful replay question is not "what is the universal best route?" It is "which evidence and decisions changed this case?" The official design intentionally makes the answer variable.

Known unknowns

Killer identities, candidate names, exact target schedules, murder scenes, route locks, and the number of distinct outcomes are not published in the official sources checked for this page. Do not use a fan theory as a confirmed spoiler until the release build or a first-party source supports it.

For the time pressure behind this system, read [Grave Seasons Tips](/tips/). For the town's official scope, see [Residents](/residents/) and [Characters](/characters/).