Guide

Auto battler game design explained

You spend thirty seconds buying units, sliding them two tiles left, and praying your front line holds — then combat runs itself while eight players watch damage numbers fly. That tension between setup mastery and spectator suspense is the auto battler contract. Popularized by Dota Auto Chess and refined in Teamfight Tactics, Hearthstone Battlegrounds, and Super Auto Pets, the genre strips real-time APM and asks players to draft, position, and economy-manage instead. Unlike traditional strategy games where you issue orders mid-fight, auto battlers outsource combat to deterministic or seeded simulations — the skill expression lives entirely in preparation. This guide covers subgenre definitions, the shop-and-fight core loop, unit tiers and synergies, board positioning geometry, gold economy and interest, matchmaking and damage routing, a Harbor Arena worked example, a subgenre decision table, common pitfalls, and a production checklist.

What defines an auto battler

Not every game with automated combat qualifies. Players expect a recognizable bundle of systems:

Draft phase — buy, sell, combine, or reroll units from a shared or personal shop between fights.
Placement phase — arrange units on a grid or line before combat locks; positioning is the primary tactical input.
Automated combat — once the timer ends, units fight without player micro; outcomes depend on stats, abilities, targeting AI, and placement.
Multiplayer elimination or ranking — typically 4–8 players in a lobby where losers take damage to a shared health pool until one remains, or asynchronous ghost boards compete for placement points.
Synergy or trait bonuses — collecting units that share tags (race, class, element) unlocks stacking team-wide buffs that shape viable comps.

Hybrids exist: single-player roguelike auto battlers (Super Auto Pets) drop multiplayer lobbies but keep the shop loop; card-based auto battlers (Hearthstone Battlegrounds) replace grid placement with tavern minion ordering. The through-line is preparation as the game and combat as the payoff animation.

The core loop: shop, position, fight, repeat

A standard round breaks into predictable beats that players internalize within their first match:

Income tick — base gold per round, plus bonuses for win/loss streaks, interest on banked gold, and sometimes HP-based catch-up.
Shop refresh — five (or fewer) random offers from the current tier pool; reroll costs gold; locking the shop preserves options next round.
Bench management — overflow units sit off-board; combine three copies to upgrade star level and stat scale.
Placement — drag units onto hexes or squares; frontline tanks absorb damage, backline carries attack; some units prefer corners or flanks.
Combat resolution — server or client sim runs targeting, ability casts, and death order; results apply damage to losers or award placement.
Meta progression — every few rounds, tier pools unlock higher-cost units; carousel drafts or PvE rounds inject items and components.

Round length matters: 30–45 seconds of planning plus 15–30 seconds of combat keeps lobbies brisk. Stretch either phase and players alt-tab. The loop should feel like a deckbuilder's reward screen every round — constant small decisions with visible compounding.

Unit tiers, pools, and three-star scaling

Auto battlers gate power through cost tiers (1-gold scrappers through 5-gold legendaries) and star levels (combine three identical copies). Design constraints:

Pool size per tier — finite copies in the shared pool (e.g. ten 1-costs, nine 2-costs) create contested picks and deny strategies.
Stat curves — each star multiplies base HP and damage; a 3-star 1-cost should threaten but not obsolete a 2-star 4-cost.
Ability breakpoints — some units gain new effects at 2-star or 3-star; telegraph this in UI so players chase upgrades intentionally.
Transition windows — when tier 4 unlocks at round 15, players need enough economy to pivot; if transitions are too expensive, lobbies freeze into one comp.

Test contested units: if everyone wants the same 4-cost carry, the pool empties and variance spikes. Add substitute carries with overlapping items or traits so pivots stay viable when contested.

Synergies, traits, and comp identity

Traits (origin, class, tribe) reward horizontal collection: 2/4/6 Warriors might grant armor; 3/6/9 Mages grant spell amp. Good trait design:

Readable thresholds — breakpoints at 2, 4, 6 are easier to plan than 3, 5, 7, 9 unless late breakpoints are spectacular payoffs.
Vertical vs horizontal — vertical traits buff one unit type deeply (Assassin crit); horizontal traits buff the whole board (Guardian shields).
Carry + support pairing — every meta comp needs a damage anchor and units that enable it (mana generation, attack speed, shred armor).
Flex slots — leave 1–2 board slots for item holders, utility, or trait bridges so players adapt to shop RNG.

Avoid trait soup where ten overlapping tags dilute identity. A new player should glance at an opponent's board and name the comp ("six Divine, two-star Kayle carry") within three seconds. That readability drives spectating and content creation.

Positioning and board geometry

When players cannot micro, placement is the micro. Hex grids (TFT) vs rectangular lines (Battlegrounds) change tactics:

Frontline thickness — double-wide tanks absorb single-target burst; thin fronts die to AoE lines.
Corner tucking — carries in back corners dodge hooks and diagonal dashes; assassins jump to backline hexes unless guarded.
Zigzag spacing — spread units one hex apart to reduce cleave and chain lightning value.
Scout timing — last-second swaps after seeing opponent board (in mirrored PvP) reward game knowledge; cap swap windows to prevent infinite stalling.

Unit targeting AI must be deterministic and documented: nearest enemy, lowest HP, highest threat, random among ties. Hidden random targeting erodes trust. Publish rules in a codex so competitive players can theorycraft.

Gold economy: interest, streaks, and tempo

Economy design separates greedy bankers from tempo bullies. Common levers:

Interest — +1 gold per 10 saved, capped at 50 gold (5 bonus) rewards patience; breaking point is usually 50 before a roll-down.
Win/loss streaks — escalating gold for consecutive outcomes prevents death spirals and gives losers comeback tools.
HP damage scaling — early losses tick 1–2 HP; late losses with surviving units deal 10+ to punish weak boards without ending games in round 3.
Reroll cost vs shop odds — flat 2-gold rerolls favor whales; incremental costs or free rolls on loss rounds soften RNG.

Economy ties directly into broader game economy design: every gold source needs a sink (rerolls, XP buys, unit purchases). If players hoard above the interest cap with nothing to spend on, inflation breaks late-game balance.

Matchmaking, ghosts, and lobby fairness

Eight-player lobbies use asynchronous PvP: you fight a snapshot of another player's board from a prior round, not live head-to-head. Implementation notes:

Ghost selection — pair boards of similar strength; avoid matching a 3-star lobby against round-5 weaklings.
Mirror matchups — some rounds everyone fights the same PvE camp for items; tune camp HP so underbuilt boards still survive with losses.
Damage attribution — surviving units deal bonus HP damage; cap unit-count multiplier so summon spam does not one-shot players.
MMR and hidden ratings — ranked modes need placement-based scoring independent of lucky lobbies; see matchmaking design for queue health.

Combat must be server-authoritative or seeded reproducibly — clients replay the same seed to show identical fights. Desync between players watching the same round destroys trust in ranked play.

Worked example: Harbor Arena round 18 pivot

Imagine Harbor Arena, an 8-player lobby at round 18. You have 42 gold, 28 HP, and a fading two-star frontline of Harbor Guards (trait: +armor at 2/4/6). Tier 4 units just entered the pool. Your shop shows a second Harbor Marksman (carry trait: +attack speed at 3/6) and filler 1-costs.

Decision tree: (1) Roll down to 10 gold searching for a third Marksman copy — high risk if contested. (2) Level to 8, preserving interest, and play flex with two traits at partial breakpoints. (3) Sell Guards, eat −4 armor breakpoint, and hard pivot into six Marksmen with borrowed items.

The correct call depends on scout data: two opponents already field three-star 1-cost reroll comps (they will spike now); one opponent is on the same Marksman line (contested pool). You level to 8, hold 30 gold for 3 interest, place a corner-tucked Marksman with RFC and IE items, and run four Guards plus four Marksmen — partial traits now, six Marksmen later. You lose this round (−6 HP) but your economy hits 50 gold by round 20, when you complete six Marksmen and stabilize top four. The example shows how scout-informed pivots beat blind roll-downs when pools contest.

Subgenre decision table

Format	Best for	Watch out for
8-player elimination (TFT-style)	Ranked ladders, esports spectating, high replay variance	Long sessions (35+ min); eliminated players need spectator UX
Async placement ladder (Battlegrounds)	Casual mobile, short runs, clear top-4 wins	Less direct interaction; positioning depth must carry interest
Single-player roguelike (Super Auto Pets)	Offline play, daily challenges, fast iteration	No lobby drama; needs strong AI opponents and mutators
Card tavern auto battler	Existing CCG audiences, simpler placement	Order-only positioning limits tactical depth
Co-op PvE auto battler	Lower toxicity, friend groups	Balance for carry roles; shared loot disputes

Common pitfalls

Opaque combat — players cannot tell why they lost. Add combat log, trait summaries, and post-fight stat breakdowns.
Single dominant comp — one trait line wins 40% of lobbies; hotfix traits or pool sizes within days, not months.
Infinite stall timers — placement phase with no cap lets players grief; hard-close at 30 seconds with last board saved.
Unclear pool odds — hiding tier percentages feels rigged; show shop odds by level like TFT's official tables.
Item RNG without mitigation — bad component drops on streak losses need pity or choice components.
3-star lottery walls — if lobbies are decided by who hits a 3-star 1-cost first, skill expression collapses.

Production checklist

Document targeting priority for every unit class and publish in-game.
Simulate 10,000 lobbies with bot economies; graph comp win rates.
Cap placement phase at 30–45 seconds with audible countdown.
Verify combat replay from shared seed matches across all clients.
Pool spreadsheet: unit counts per tier, contested thresholds, pivot substitutes.
Interest and streak tables visible in tutorial and codex.
Spectator mode for eliminated players with trait and item tooltips.
Ranked damage formula tested for summon-heavy comps and infinite stall edge cases.

Key takeaways

Auto battlers make drafting and placement the entire skill game; combat is the reveal.
Shop economies — interest, streaks, and tier unlocks — create tempo vs greed decisions every round.
Traits and tiers define comp identity; contested pools need substitute carries.
Positioning replaces micro: frontline, corners, and spacing decide fights.
Server-seeded combat and transparent targeting rules are mandatory for ranked trust.