Guide

Game gacha mechanics design explained

Gacha (from Japanese capsule-toy vending machines) is the randomized reward loop at the heart of many mobile RPGs, hero collectors, and live-service titles. Players spend premium or earned currency on a pull (single or multi-summon) that returns characters, weapons, or items from a weighted table — usually sorted into rarity tiers with flashy reveal animation. The mechanic is lucrative because it combines collection completion, social status, and variable-ratio reinforcement. It is also the fastest way to lose a community if odds feel rigged, pity counters reset without warning, or power-gated content demands a specific limited unit. This guide covers banner types, rarity and power curves, soft and hard pity, duplicate systems, economy integration, regulation, a Harbor Summon limited banner worked example, a model decision table, pitfalls, and a production checklist — alongside our game monetization overview, economy design guide, and loot table math deep dive.

What gacha is and why it works

At its core, gacha is a weighted random draw with presentation layered on top. The psychology is well studied: variable rewards (you might get something great this pull) drive repeated engagement more reliably than fixed payouts. Collection framing ("gotta catch 'em all") adds long-term goals beyond any single session. Limited-time banners introduce scarcity without requiring real inventory limits — the item exists in code forever, but the window to obtain it easily closes.

Gacha differs from generic loot boxes in scope and expectation. A shooter crate might drop a skin; a gacha banner is often the primary progression vector for roster-based games. Players expect transparent tier rates, pity guarantees, and duplicate handling that makes bad luck recoverable. When those expectations break, refund requests and store-review bombs follow — even if the math is technically disclosed.

The standard pull loop

Player selects a banner (permanent pool, rate-up, or limited event).
Spends currency — typically 1x or 10x multi-pull with a "guaranteed rare" bonus on ten.
Server rolls against the published table; client plays reveal sequence.
Item grants to inventory; pity counters update server-side.
Duplicate logic fires (shards, constellation, universal token).

Every step should be server-authoritative. Client-side RNG invites exploits and destroys auditability when players dispute outcomes.

Rarity tiers and the power curve

Most gacha games use three to five visible rarity bands — common (1–2 star), uncommon (3 star), rare (4 star), and ultra-rare (5 star) — with different base rates. A typical published table might show 3-star at ~80%, 4-star at ~15%, and 5-star at ~1–2% per pull on a standard banner. Rate-up and limited banners split the 5-star pool: half (or more) of 5-star hits go to the featured unit while the banner runs.

Design the power curve carefully. If 3-star units are unusable trash, every bad pull feels like a slap. Many successful titles make 3-stars viable in early content and 4-stars competent through mid-game — reserving 5-stars for endgame optimization and collection prestige. Selling exclusive power that free players can never approximate breaks PvP and poisons co-op matchmaking.

Featured vs off-banner 5-stars

On a rate-up banner, players chasing the featured character accept that most 5-stars will be "wrong" ones. The emotional contract is: pity and rate-up tilt odds enough that dedication (or spending) eventually wins. Hide the off-banner list in the UI and you breach that contract. Show it prominently, including standard-table units that dilute the pool.

Banner types and scheduling

Permanent (standard) pool — evergreen roster at base rates. No FOMO; pity often carries over indefinitely. Good onboarding sink for free currency and a safety valve when limited banners end.
Rate-up (non-limited) — one or two units get boosted 5-star share but remain in the pool after the banner ends. Lower anxiety than true limiteds; still drives spending spikes.
Limited-time character banner — featured unit unavailable (or severely diluted) after the event. Highest revenue per day; highest community stress. Pair with rerun schedules announced in advance.
Weapon / equipment banner — parallel track for gear gacha. Often harsher economy; consider separate pity currency so bad weapon luck does not brick character pulls.
Step-up / beginner banner — discounted multi-pulls with escalating guarantees (e.g. 20-pull box with fixed 5-star on step 3). One-time per account; strong first-purchase conversion when tuned to early friction.
Collaboration (collab) banner — licensed characters with stricter rerun uncertainty. Legal contracts may forbid reruns — disclose that in banner copy before players commit savings.

Live-ops calendars stagger banner types so whales always have a chase target while free players can skip a month without falling permanently behind. Overlap two limited 5-stars simultaneously only if your audience expects it — otherwise revenue cannibalizes and both banners underperform.

Pity systems: soft, hard, and sparking

Pity (bad-luck protection) turns open-ended gambling into a bounded grind players can plan around. Three patterns dominate modern design:

Hard pity (guaranteed rare at N)

If no 5-star by pull 90, pull 90 is 100% 5-star. Hard pity is the backstop that prevents infinite dry streaks. Publish N clearly; never reset it silently when a banner ends unless you also migrate progress to a shared counter.

Soft pity (rising odds)

Starting around pull 70–75, each miss increases the next pull's 5-star rate (e.g. +5% per pull) until hard pity triggers. Soft pity makes the average cost lower than N × single-pull price while preserving variance for early luck. Document the ramp in player-facing FAQ, not only compliance PDFs.

Sparking / selector currency

Every pull on a banner awards progress toward a spark — at 200 pulls (example), pick any rate-up unit from that banner even if RNG never hit. Sparks reward whales and steady free players alike; they also cap maximum spend for a guaranteed outcome. Some games make sparks banner-specific (use or lose); others use universal tokens exchangeable in a shop. Banner-specific sparks increase urgency; universal tokens reduce burnout.

Pity carry-over rules

Players remember carry-over policies forever. Common patterns:

Hard pity counter carries between all character banners.
Rate-up guarantee (50/50) resets per banner but pity count persists.
Spark progress resets when banner ends — disclose countdown in UI.

The infamous "50/50" system: on winning a 5-star, there is a 50% chance it is the featured unit; if not, the next 5-star is guaranteed featured. It doubles expected cost for target-chasers but feels fairer than pure RNG alone.

Duplicates, ascension, and dead pulls

Pulling a unit you already own should not be a complete waste. Standard mitigations:

Constellation / ascension — duplicates unlock bonus stats or abilities up to C6 (or similar cap). Stop at C0–C1 being transformative and C6 being whale-only power or cosmetic flair.
Universal shards — duplicate converts to currency buyable toward any unit in the same rarity tier.
Max-duplicate refund — at max ascension, convert to premium currency or banner tokens.

Weapon duplicates hurt more than character duplicates because pools are wider. Consider a "fate point" system (extra 4-star or 5-star weapon becomes tokens toward a selector) to prevent rage quits when a player hits five off-banner swords in a row.

Economy integration and free-to-play paths

Gacha does not exist in isolation — it sits inside the broader economy. Map every currency source and sink before tuning rates:

Premium gems — bought with real money; also drip from battle pass, achievements, and events in ethical titles.
Free summon tickets — banner-specific or generic; primary F2P path. A steady F2P player should expect one limited 5-star per several months via pity alone — document that cadence internally.
Daily/weekly stipends — login rewards that accumulate toward ten-pulls keep retention without instant gratification.
Stamina vs summons — if players must choose between farming stages and pulling, they feel punished for engaging with content.

Pair gacha with a battle pass that grants summon currency on the free track — it lifts conversion without making the pass feel mandatory for basic roster viability.

Regulation, disclosure, and ethics

Apple and Google require odds disclosure in many territories. Belgium and the Netherlands have treated some loot-box models as gambling; China mandates published rates and pull logs. Even where law is gray, transparency is cheaper than chargebacks and brand damage.

Show per-banner rates in-game before purchase, not buried in a website PDF.
Log every pull server-side with timestamp, banner ID, and outcome hash.
Avoid selling pulls to minors with credit-card friction bypassed by third-party gift cards without parental controls.
Never manipulate reveal animation to imply a higher rarity than rolled.
Publish rerun policies for limited units — uncertainty is not the same as surprise.

Ethical gacha sells collection and convenience, not mandatory power for ranked modes. If your game is PvE-only, modest power gating is tolerable; in PvP, keep gacha rewards cosmetic or sidegrade at most.

Worked example: Harbor Summon limited banner

Harbor Summon is a mobile tactical RPG shipping its first limited 5-star healer, Selene Tidecaller. Live-ops wants strong day-one revenue without community backlash. Design choices:

Duration: 21 days — long enough for monthly-paycheck cycles, short enough for urgency.
Rates: base 1.6% 5-star; soft pity from pull 74 (+6% per miss); hard pity at 90. Featured rate-up: 50% of 5-stars are Selene; lost 50/50 guarantees next 5-star featured.
Spark: 180 banner pulls earns a selector for Selene or the other rate-up 4-star support — unused sparks convert to 10 universal stellar shards at banner end (announced upfront).
Ten-pull: guaranteed at least one 4-star; cannot be a weapon on character banner.
Free path: event grants 15 tickets + ~40 pulls worth of gems for active F2P — enough to touch soft pity once if they save two weeks prior.
Duplicate: Selene duplicates grant constellation up to C4; further dupes become 25 stellar shards each.

Pre-launch spreadsheet: median pulls to featured 5-star ≈ 78; 90th percentile ≈ 145. Marketing copy uses "guaranteed featured within 180 pulls via spark" — not "cheap pulls" — so whales budget sparks and dolphins know the ceiling. Post-banner survey tracks satisfaction; rerun Selene six months later on a rate-up (non-limited) banner to catch newcomers.

Decision table: gacha vs alternatives

Goal	Gacha fits	Consider instead
Roster collection fantasy	Character banners with pity	Direct unit shop if roster is small (<20 units)
Recurring revenue spikes	Limited banners + sparks	Battle pass if cosmetics-only and stable MAU
Whale monetization	Constellation dupes, weapon banners	Cosmetic shop without power variance
Regulatory simplicity	Published rates, no real-money trade	Premium game + paid DLC, no random paid rewards
Competitive PvP	Cosmetic gacha only	Never sell exclusive meta units in ranked
Small live-ops team	Permanent pool + monthly rate-up	Skip limited collabs requiring legal rerun negotiation

Common pitfalls

Opaque pity migration — resetting counters on banner rollover without compensation is a top community complaint.
Power creep every banner — each new 5-star obsoleting the last tells collectors their investment expires.
Weapon-banner trap — character on banner requires signature weapon also gacha-only; stack costs explode.
Fake scarcity timers — "24 hours left" banners that return weekly erode trust faster than honest rerun schedules.
Client-side pity display desync — UI shows 80/90 but server has 72; support tickets explode.
Under-tuned 4-star pool — useless 4-stars feel worse than 3-stars because they consumed pity-adjacent rarity slots.

Production checklist

Simulate 100k pulls per banner in Monte Carlo before publishing rates.
Server-authoritative RNG with auditable logs and idempotent grant handlers.
In-game odds screen matches legal disclosure in all supported regions.
Hard pity, soft pity, 50/50, and spark rules documented in player FAQ.
Pity carry-over behavior tested across banner open, close, and maintenance.
Duplicate handling defined for every rarity before first live banner.
Free-to-play pulls-per-month modeled and playtested on zero-spend accounts.
Reveal VFX cannot imply higher rarity than server result.
Rerun calendar communicated for limited units within six months.
Track ARPPU, average pulls to 5-star, spark redemption rate, and banner NPS.

Key takeaways

Gacha is economy plus psychology — tune tables with spreadsheets, not gut feel.
Pity is the contract — hard pity, soft pity, and sparks define maximum cost.
Duplicates must convert — dead pulls are retention poison.
Transparency beats tricks — disclosed odds and honest timers outperform dark patterns long-term.
Schedule for humans — paycheck cycles, rerun clarity, and skippable months keep players paying willingly.