FOXVALE • PLAYER RETENTION

Who comes back—and when?

A report based on actual session starts, complete follow-up windows and a minimum amount of PLAN-recorded non-AFK time.

October 2–26, 2026 · As of October 26 at 12:01 PM Pacific · SMP only

Back during the first week89.6%69 of 77 eligible accounts
15 accounts are still too new to measure

Back during the second week79.7%55 of 69 eligible accounts
A return on any day in days 7–14

Latest arrivals: day-one return3 of 837.5% among mature recent arrivals
7 more arrivals need longer follow-up

The old day-seven zero should be “not yet measurable” for recent joiners. All 15 accounts first seen in the latest seven days are still too young for a complete day-seven window. That does not mean they all quit.

What the data suggests

Most eligible accounts came back at least once during week one. This is broader than returning on one exact day and does not mean daily play.
Recent arrivals deserve a closer look, but there is no proven decline yet. Day-one returns are 3/8 for the latest arrivals versus 5/8 in the preceding seven-day arrival group. Both groups are very small; their uncertainty ranges overlap.
Do not read every new session as a human returning. Short disconnects around restarts can create extra sessions. This report merges gaps of 30 minutes or less and uses a non-AFK threshold.

How much data?

93 accounts · 4,350 sessions
92 accounts have usable first-visit history. One account is excluded because its initial visit was not observed.

48.5% of connected time is marked AFK by PLAN. Non-AFK is still only an activity proxy. Historical Unplugged retained intervals are not present in these exports, so fully human-only retention cannot be calculated.

Return rates with fair denominators

First-join group

Measure	Returned / eligible	Rate	Still too new	95% sample interval

Primary rule: a separate visit with at least 10 PLAN non-AFK minutes inside the measurement window. The initial visit is excluded. Intervals are approximate Wilson sample intervals; they do not capture AFK/telemetry bias.

Player explorer

Choose a player

Same October 26 snapshot. Names and summary statistics in this section are visible to anyone with this link.

Qualifying return milestones

A return requires a separate visit and 10 PLAN non-AFK minutes inside the stated window. “No qualifying return” does not mean the player quit.

Top and bottom 10

Rank by

All 93 accounts, over the observed October 2–26 history. Newer arrivals have had less time to accumulate activity. These are activity rankings, not player quality or proof of churn. Ties use alphabetical order.

Top 10

Player	Value	First tracked

Bottom 10

Player	Value	First tracked

PLAN non-AFK time is an activity proxy, not verified human playtime. Visits merge gaps of 30 minutes or less. Visit-start days count Pacific dates with a visit start, not every day a continuous session spans. Historical retained Unplugged time cannot be fully separated.

Compare first-join cohorts

Grouped by the Pacific week of first tracked join. Every cell shows its own eligible denominator. The first and current weeks are partial; first tracked join may include existing community members.

First tracked join	Day 1	Day 7	Day 14	First week	Second week
10-02 → 10-0428 first joins	71.4%20/28 eligible	64.3%18/28 eligible	57.1%16/28 eligible	96.4%27/28 eligible	89.3%25/28 eligible
10-05 → 10-1141 first joins	58.5%24/41 eligible	43.9%18/41 eligible	30.0%12/40 eligible	90.2%37/41 eligible	73.2%30/41 eligible
10-12 → 10-188 first joins	62.5%5/8 eligible	37.5%3/8 eligible	Not yet measurable	62.5%5/8 eligible	Not yet measurable
10-19 → 10-2513 first joins	37.5%3/8 eligible	Not yet measurable	Not yet measurable	Not yet measurable	Not yet measurable
10-26 → 10-262 first joins	Not yet measurable	Not yet measurable	Not yet measurable	Not yet measurable	Not yet measurable

Pale amber = fewer than 20 eligible accounts. Grey = no complete follow-up window yet. Do not rank tiny cohorts as reliable winners or losers.

First-visit engagement

PLAN non-AFK minutes during the first merged visit, capped at 24 hours. This is not proven hands-on playtime.

Under 2 minutes2

2–10 minutes9

10–30 minutes12

30+ minutes69

11 of 92 had under 10 recorded non-AFK minutes on that first visit. Returning later is still possible; this is not a churn label.

What to do next

Check in with recent newcomers. Ask about onboarding, finding a place to build and getting help. The current sample is too small to blame a specific change.
Watch the October 19–25 group as it matures. Its day-seven and week-one measures are not ready yet.
Keep definitions fixed. Compare the same windows and minimum-time rule next time; keep counts next to percentages.
Track retained intervals prospectively if human-only retention is needed. Existing cumulative Unplugged counters cannot reconstruct historical activity windows.

Why this differs from the dashboard curve

The dashboard curve measures last-seen minus registration time. It can count someone as retained across days they never visited, and recent joiners have not yet had time to reach later thresholds. This report measures separate visits and excludes accounts whose complete window has not elapsed.

Measure	Raw logins, ≥10m non-AFK	Merged visits, any duration	Merged visits, ≥10m non-AFK
Day 1	65.9%	64.7%	61.2%
Day 7	57.1%	51.9%	50.6%
Day 14	50.0%	42.6%	41.2%
Back during the first week	89.6%	90.9%	89.6%
Back during the second week	79.7%	84.1%	79.7%

These are different measures, not alternative estimates of the same “true human” retention. Merging reduces short-reconnect artifacts but can miss someone returning to their keyboard while remaining continuously connected.

Definitions, source coverage and limitations

A return is a separate visit beginning inside the elapsed-time window with at least 10 PLAN non-AFK minutes within that window. Adjacent sessions separated by <=30 minutes are merged. Initial visit excluded. AFK is only a session total, so boundary overlaps subtract the entire session AFK amount as a conservative lower bound; activity after the window never qualifies an earlier return. A person who stays connected and returns to their keyboard may not create a new visit: this is separate-visit retention, not all active-day retention.
First merged visit, capped at its first 24 hours; non-AFK minutes use the same conservative overlap accounting.
Only players whose complete measurement window has elapsed are in its denominator. Exact day N uses [N,N+1) days and requires N+1 days of follow-up. First week uses [1,7); second week [7,14). First 72h uses [0,3), excluding the initial visit.
PLAN AFK subtraction is an activity proxy. No event-level Unplugged history is available here, so these metrics cannot prove human activity or fully separate retained sessions.
Weekly cohorts use Pacific Mondays. First/last cohorts are partial. Cohort rates use only mature members. The launch cohort may include pre-existing community members; first tracked join is not proof of a new community acquisition. Wilson intervals do not account for telemetry/AFK bias.
Session export contains one SMP UUID, 4350 nonduplicate/nonoverlapping sessions and 93 accounts. Registration snapshot is 30 minutes older, but UUID sets match exactly; first-join dates are stable. One account with no observed first session/history mismatch is excluded from all new-player rates. No server changes were made.

Checks passed: all session accounts match the registration export; one SMP identifier; no duplicate session keys, no overlapping adjacent sessions, no negative durations and no AFK total greater than its session duration. Seven synthetic calculation tests verify reconnect merging, elapsed-window boundaries, AFK thresholds and exclusion of the first visit.

30-day retention is unavailable: the observed history is only about 24 days. No causal link to lag, claims changes or a specific mod is established by this report.