Skip to content
Johan Lee
All work

Desktop game · Personal project

Animal Tactics

An auto-battler for desktop, the genre Teamfight Tactics made popular: draft animal spirits from a shop the whole lobby shares, combine copies into stronger forms, build trait synergies on a hex board, and outlast seven AI opponents. It's in development, and a web release may follow.

Animal Tactics logo
Role
Design and development
Timeline
Version 0.9 · In development
Stack
Godot 4.7GDScriptDesktop
Animal Tactics in combat: two teams of animal spirits fighting on a hex board with summoned Saplings, traits on the left, the players and a pinned unit on the right, and the shop along the bottom
  • 38

    Animal spirits, each with its own ability

  • 7

    AI opponents, at three difficulties

  • ~33,000

    Checks in the headless test suite

The game

Each round starts with a planning phase: collect income, buy from a shop shared by all eight players, reroll, level up, and arrange your board. Then the fight plays out on its own, against another player or a round of monsters, and the loser takes damage. Three copies of an animal combine into its Awakened 2-star form, and three of those into a Sovereign 3-star. The last player standing wins.

There are 38 animal spirits across five habitats, eight roles and two traits that belong to a single animal, and each animal fights its own way: the Tortoise hides in its shell, the Hippo sends enemies running, the Orca leaps into the biggest group of enemies. Most traits change how a fight plays rather than adding stats: Savannah animals get stronger standing together, Woodland grows a healing Sapling, and Arctic attacks build Frost that freezes. Around them sit 36 items built from eight components, augments you pick three times a game, a shared draft every stage, and a Season that changes one rule for the whole game.

Seven bots fill the other seats, at Easy, Normal or Hard. Each picks traits to focus on, levels on a schedule, keeps gold back for interest, and rolls for upgrades when its health gets low; Hard bots also build toward trait breakpoints.

How I built it

I built Animal Tactics with Claude Code and OpenAI's Codex, working from a design document that keeps every rule and number in one place. The first commit was already a complete game, the first nine releases, 0.1 to 0.6.1, landed within a day of starting it, and 0.9 followed three days later.

The animal art was generated with ChatGPT's image generation: one sheet per animal with its three forms, which the game cuts apart and animates in code. The item icons are SVGs drawn by a small script, and the 25 sound effects are synthesized from scratch, with no recorded samples.

Fights you set up but don't control

Combat is a fixed-step simulation at 30 ticks a second, seeded so the same fight always plays out the same way. Units walk the hex grid around each other, build mana and cast, and the two sides take turns acting first each tick so neither gets a head start.

The game state and the simulation are plain objects with no scene attached. During combat every fight in the lobby runs side by side, the board draws whichever one you're watching, and the tests play whole eight-player games with no window at all. Rules, combat, every ability, item and trait, and eight full games add up to about 33,000 checks, and separate UI tests drive the real scene with simulated clicks and drags.

A deterministic simulation is also what makes an online lobby practical later, since a server could run the same code.

Balancing 38 animals

All the tuning lives in data files: shop odds by level, how many copies of each unit the shared pool holds, the economy, and every unit's stats. Pricier animals start stronger and grow more with each star, from 1x, 1.8x and 3.24x for 1-costs to 1x, 1.9x and 4.6x for 5-costs.

To check the numbers, a harness fights role-matched teams against each other, and every animal wins between 44% and 57% of its fights against others of the same cost. When version 0.9 gave nine traits real mechanics, each was measured against its old version by running the same fights with the trait on and off, then across 96 full bot games, where the nine traits' average placements landed between 4.33 and 4.92.

Built for a desktop screen

The game opens on a title screen: pick the bots' difficulty and a planning timer (45 seconds, 30, or untimed), read a 12-step picture guide, browse every animal in the spirit gallery, or look back at your match history with stats for each animal and trait. First-game tips explain each new idea the first time it comes up, and the end screen shows all eight players' final teams.

In a game, the layout works at any window size, with panels pinned to the edges and the board centered. Tooltips appear the moment you hover an item, trait or shop card, the item bench lists the recipes you can build right now, and most actions have a key: D to reroll, F to buy XP, E to sell, W to move a unit between bench and board. Clicking a unit pins its live stats, and a combat recap shows damage dealt, damage taken, and healing and shields for every unit.

Where it stands

Animal Tactics is at version 0.9 and runs on desktop. Next on the list: more trait mechanics and Seasons, keybinds, saving a game in progress, and possibly online play. A web release may follow. The game was called Unit Tactics until version 0.8.1, and the journal below covers every release so far.

Screenshots

Inside the game

  • The Animal Tactics title screen, with new game options for bot difficulty, the planning timer and the Season, and a row of 3-star spirits along the bottom
    The title screen, with options for a new game.
  • Planning a round: animal spirits on the stone arena board, trait badges on the left, a 3-star Orca pinned in the unit panel, and the shop along the bottom
    Planning a round, with a unit pinned on the right.
  • The final standings: every player's placement, final team with stars and items, and active traits, with Play again and Main menu buttons
    The final standings show every player's team.
  • The Team planner: a searchable roster, a formation of eight units on the hex grid, and the items planned for each
    The team planner, with positions and items.

Devlog

Animal Tactics in the journal

All entries
  1. Animal Tactics 0.9: traits that change fights, and Seasons

    Nine traits now change how a fight plays instead of adding stats, every game rolls one of eight Seasons, and the title screen keeps a history of your games. Plus music and sound.

  2. Animal Tactics 0.8: a title screen and a new name

    The game now starts on a title screen with bot difficulty and timer options, teaches itself with first-game tips, ends with full standings, and has a new name.

  3. Unit Tactics 0.7: belts, gloves and ten new animals

    Two new components bring the item list to 36, two new classes and two one-animal traits join the roster, and ten new animals take it to 38.

  4. Unit Tactics 0.6: 28 animals, 28 abilities

    Every animal now fights its own way, stats scale with cost and were tuned with a fight harness, and a polish pass reworks the board, shop, traits and unit panel.

  5. Unit Tactics 0.5: augments, special items and a shared draft

    Three times a game you pick an augment that lasts the whole game. Spirit Totems make emblems, new tools rework your items, and every stage has a shared draft.

  6. Unit Tactics 0.4: Wild Spirits

    The placeholder tokens become 28 painted animal spirits with three forms each, plus a pause menu, a team planner, fairer loot, shop odds and a round timeline.

  7. Unit Tactics 0.3: any window, instant tooltips, live scouting

    The layout works at any window size, tooltips appear the moment you hover, and scouting now shows every fight live, side by side.

  8. Unit Tactics 0.2: items, scouting and a damage chart

    Monster rounds now drop item components that combine into 21 completed items, you can scout any player's board, and a live chart shows who's doing the damage.

  9. Unit Tactics 0.1: a playable auto-battler

    The first commit is already a whole game. Draft units from a shared shop, build traits on a hex board, and outlast seven AI opponents in fights that play themselves.