Guide

Game ranked decay and inactivity systems explained

Harbor Brawl's ranked telemetry flagged a retention cliff that had nothing to do with balance patches. Players who returned after a two-month break logged into Diamond-tier lobbies, lost their first five matches at an 82% rate, and churned within 48 hours. The hidden MMR still reflected peak-season form; their reaction times and matchup knowledge did not. Conversely, active grinders at the same MMR band reported “free wins” against rusted opponents — queue quality collapsed at both ends. The ranked refactor introduced inactivity drift on hidden rating, a visible LP decay band for top tiers, and a five-match placement re-entry sleeve after 30 days offline. Return-player seven-day retention rose 19 points; match-quality variance in the Diamond band dropped 11%. This guide explains why decay exists, how hidden vs visible rating interact, curve design, season resets, the Harbor Brawl refactor, a technique decision table, pitfalls, and a production checklist.

Ranked decay is any rule that lowers a player's competitive standing when they stop playing or when a season ends. It is not punishment for taking a vacation — it keeps matchmaking honest about current skill, prevents ladder inflation at the top, and gives seasonal content a meaningful reset without wiping hundreds of hours of progress. Decay pairs with live-ops seasons, anti-smurf heuristics, and match format rules that define how many games count toward rating changes.

Hidden MMR vs visible rank

Most competitive games maintain two numbers:

Hidden MMR (matchmaking rating) — the estimate matchmaking uses to find opponents. It updates every match via Elo, Glicko, or TrueSkill math and is often invisible to players.
Visible rank — the tier badge, LP bar, or league points players see on profile screens. It may lag MMR (promotion series, demotion shields) or diverge entirely (separate ranked and casual pools).

Decay can target either layer or both. Drifting hidden MMR fixes queue quality without demoting someone's trophy case mid-season. Drifting visible LP creates social pressure to log in — common in Challenger or Grandmaster bands where top-tier queue depth is thin. Applying decay to both without coordination produces the worst UX: a player demoted to Platinum while still matching against Diamond MMR opponents.

Rule of thumb: matchmaking reads hidden rating; players read visible rank. Any decay that affects hidden MMR should eventually surface in visible rank through normal promotion/demotion, or players will feel the system is lying.

Types of ranked decay

Time-based MMR drift

After N days without a ranked game, hidden MMR moves toward a population mean (often the median of the active player base) at rate r points per day. Example: a 3200 MMR player inactive 45 days might drift to 2950 while a 1400 MMR player drifts up to 1550. Drift stops at a floor/ceiling or when the player completes placement matches.

Visible LP decay (ladder demotion)

Used in League of Legends Challenger, Overwatch top 500, and similar thin queues: if you do not play at least one ranked game every X days while above tier T, you lose Y LP per day until you drop below T or play again. This keeps the leaderboard representing active competitors, not historical peaks.

Soft season reset

At season rollover, compress everyone's hidden MMR toward the mean while preserving relative order. A player at the 95th percentile might start the new season at the 85th percentile — still high, but not stomping placements. Visible rank often resets to a lower tier with a short placement series to re-anchor.

Hard season reset

Wipe visible rank to a baseline; optionally reset or heavily drift hidden MMR. Common in battle royale seasonal ladders where there is no persistent promotion series. Cheaper to explain to players but angers veterans unless accompanied by exclusive cosmetics tied to the prior season peak.

Activity grace and decay immunity

Grace periods exempt new placements, account-linked smurf detection holds, or the first K days after reaching a new peak tier. Without grace, players who hit Diamond on Sunday night lose LP Monday because they cannot queue during work hours.

Design goals and tradeoffs

Decay is a product decision, not a math requirement. Teams add it when one of these goals outweighs player frustration:

Match quality — stale MMR creates blowouts; drift pulls inactive accounts out of tight bands.
Queue depth at the top — without LP decay, Challenger fills with ghosts who never queue, blocking active climbers.
Seasonal engagement — soft reset gives everyone a reason to play placements when new battle passes launch.
Anti-inflation — over long horizons, rating systems inflate as the player base improves; decay toward mean slows unbounded growth.
Smurf signal — accounts that spike MMR then sit idle may be sold or shared; drift reduces their matchmaking leverage over time.

Costs are real: returning players feel punished, casual competitors avoid ranked entirely, and support tickets spike when decay rules are opaque. The Harbor Brawl team A/B tested “rust badge” UI that showed “skill estimate recalibrating” instead of silent MMR drift — support volume fell 34% with identical math.

Curve design: grace, rate, and floor

A workable inactivity curve has four parameters:

Grace days (G) — no drift for the first G days offline. Harbor Brawl uses G = 14 for all tiers and G = 7 for Grandmaster+ visible LP decay.
Drift rate (r) — MMR points per day toward mean after grace. Linear drift is easy to explain; exponential approaches mean faster for extreme outliers. Cap r so a one-year break does not dump a pro into beginner queues.
Floor and ceiling — drift stops when hidden MMR reaches mean ± band or when visible rank hits a demotion tier. Floors protect identity (“I was never below Gold”) at the cost of match quality.
Re-entry placements — after D days offline, require P placement matches with elevated K-factor before rejoining the main pool. Placements act as a high-bandwidth skill probe that overrides slow drift.

Simulate curves on historical churn cohorts before shipping. Plot “expected MMR error” (absolute difference between drifted MMR and post-return win-rate-implied MMR) against days inactive. Minimize error subject to a maximum demotion players will tolerate in user research.

Harbor Brawl ranked refactor

Before the refactor, Harbor Brawl stored one hidden Glicko-2 rating per mode with no inactivity handler. Season two soft reset compressed ratings 30% toward mean but did not address mid-season breaks. The refactor shipped three layers:

Hidden drift — after 14 idle days, MMR moves 8 points per day toward the mode median (2400) until within 150 points or the player queues ranked.
Grandmaster LP decay — top 500 players lose 25 LP per idle day after 7 days without a ranked win; demotion at 0 LP to Master with demotion shield disabled for decay-only losses.
Placement re-entry — accounts idle 30+ days play five matches with 2× rating deviation update, then rejoin normal matchmaking.

UI copy explicitly states last-ranked date and projected drift if the player waits another week. Esports qualifiers use a separate non-decaying “circuit” rating so pro brackets are unaffected.

Technique decision table

Your situation	Prefer	Avoid
High churn on returning players in ranked	Hidden MMR drift + placement re-entry; transparent rust UI	Visible-only demotion with unchanged matchmaking rating
Thin top-tier queue, inactive leaderboard names	LP decay above tier T with short grace	Hard reset every month without seasonal rewards
New season battle pass launch	Soft MMR reset + visible tier compression + 10 placement games	Full hard wipe with no recap of prior peak
Casual-friendly ranked with low stakes	No decay; wider matchmaking bands after idle	Challenger-style LP loss for part-time players
Fighting game 1v1 ladder	Glicko RD inflation on idle (uncertainty rises) + provisional return set	Flat Elo drift toward mean without RD reset
Team shooters with role-specific skill	Per-role MMR drift; team queue uses aggregate	Single number drift ignoring role swap on return

Common pitfalls

Decay without communication — silent MMR drift feels like broken matchmaking; show dates, rates, and recovery paths.
Mismatched hidden and visible decay — demoting LP while MMR stays high produces impossible opponents.
Decay during server outages — pause idle clocks when ranked is disabled region-wide or players will riot over forces beyond their control.
Too-aggressive top-tier LP loss — seven-day vacations should not eject lifetime competitors; tune grace for human schedules.
Ignoring party queue — decaying one member's MMR while they only queue in five-stacks skews team balance.
Same rules for esports and pub ranked — circuit ratings need freeze windows around tournaments.
No simulation on return cohorts — shipping drift constants without backtesting produces double punishment with existing demotion shields.
Decay on smurf-flagged accounts only — appears punitive and is legally sensitive; use drift universally with standard grace.

Production checklist

Document hidden vs visible rating and which decay applies to each.
Set grace days G, drift rate r, mean target, and floor/ceiling.
Define idle — ranked only vs any login; clarify per-mode independence.
Implement placement re-entry after threshold D days with elevated K-factor or RD.
Expose last-ranked timestamp and projected drift in client UI.
Pause decay during ranked outages and documented maintenance windows.
Separate esports/circuit ratings from pub ranked decay if applicable.
Backtest drift curves on historical return-player win rates.
A/B test copy and grace length before tightening LP decay at the top.
Log decay events for support tooling and match-quality dashboards.

Key takeaways

Ranked decay keeps matchmaking honest when skill changes during time away from the game.
Drift hidden MMR for queue quality; use visible LP decay sparingly for thin top-tier ladders.
Soft season resets compress ratings toward mean without erasing seasonal achievement.
Grace periods, placement re-entry, and clear UI prevent decay from feeling like arbitrary punishment.
Harbor Brawl cut return-player churn 19 points with 14-day grace drift, Grandmaster LP decay, and five-match re-entry placements.