Building Bots for Pokémon Auto Chess
Every Pokémon Auto Chess lobby that isn't full gets padded out with bots — AI-controlled opponents built not by the developers, but by players, using an in-client Bot Builder tool. If you've ever faced a surprisingly well-constructed AI board and wondered who made it: someone drafted it stage by stage, submitted it, and a volunteer reviewer approved it into the pool every player can be matched against.
This page covers how to build one that actually gets approved — the tool itself, and the rules a submission has to follow to read as a realistic human game rather than an obviously scripted one.
What the Bot Builder actually is
It's built into the game client itself, not a separate website — open it from the main menu. You lay out a board for each stage your bot will reach, picking units, positions, and items exactly like you would mid-match, and the tool gives you a live read on what you're building: current synergies, an estimated difficulty/Elo for the board as a whole, and how many item components you've used so far relative to the stage you're on.
Save your work to a file as you go — there's no autosave, and a half-finished bot can't be submitted. When you're happy with it, Submit sends it to the #bot-creation Discord channel, where a Bot Admin reviews it and either approves it (it joins the real pool other players get matched against) or sends it back with notes.
The one rule everything else follows from
A bot has to be something a real player could have actually done. No gold to spend as a shortcut, no item or unit a real game couldn't produce at that stage, no behavior that breaks the game's own rules for players. Every specific guideline below is really just this rule applied to one part of the builder.
What gets a bot rejected
- No item components. Equip only crafted items. The game's recipes change between versions; a bot holding a component instead of a finished item breaks the moment a recipe does.
- Give it a real avatar and name. The default Ditto icon is rejected outright — pick a Pokémon from the team (ideally the carry, or one that covers its main synergies) so players get an instant read on what they're up against. If you ever need to update an already-approved bot, keep the same avatar and name; changing either submits it as a brand-new bot instead of overwriting the old one.
- Keep the Elo honest. The approval bands are Beginner (under 850), Easy (850–999), Medium (1000–1149), Hard (1150–1299), Extreme (1300–1449), and Master (1450+) — but in practice, bots much above 1100–1200 are rarely approved, and 1300+ essentially never is, since a team that strong isn't a realistic single game anymore, it's a best-case scenario stitched together.
- Only use what the board actually earned. Shiny items and Shiny Charm access require actually running a Baby-synergy stacking line for several stages; Artificial-synergy tools and the Normal-synergy scarf line are gated the same way — don't equip them unless the board's own synergies would have unlocked them in a real game. The same logic applies to anything tied to a special carousel or encounter: if it's not guaranteed every game, a bot can't rely on it either.
- Stay consistent stage to stage. A unit's shiny status, emotion icon, and held items shouldn't change arbitrarily between stages — in a real game they'd only change for a reason (an item was actually given, a unit was actually sold). If an item needs to move from one unit to another, that's only legal if the original holder was sold first, same as in a live match.
- Respect regional exclusivity. A regional form (Alolan, Galarian, Hisuian, Paldean) can't appear in the same region's encounter pool as its base form — check which region a stage belongs to before placing one.
Think like a player, not a spreadsheet
The builder doesn't track gold, so the realism check is entirely on you: would this board's level, item count, and evolution progress actually be affordable at this stage? The two things that trade off against each other are levels and rerolling — a bot that's hyper-leveled won't have many evolved units yet, and a bot that rerolled hard for fast 2-/3-star evolutions will be behind on levels. Use your own games as the reality check: a 2-star (3 copies) is realistic within a handful of turns at decent shop odds; a 3-star (9 copies) takes meaningfully longer.
| Playstyle | Stage 10 | Stage 13–15 | Stage 17–20 | Stage 22–25 | Stage 27+ |
|---|---|---|---|---|---|
| Standard pacing even leveling, modest rerolling | Level 5 | Level 6–7 | Level 7–8 | Level 8–9 | Level 9, likely econ-crashed |
| Reroll-heavy pushing for fast/high evolutions | Level 5 | Level 6 | Level 7 | Level 8, possibly crashed | Level 9 if it gets there at all |
Treat that as a sanity check, not a template to copy exactly — a board built around a reroll-heavy 3-star carry should look noticeably behind on levels compared to one that hyper-leveled past evolutions. An "econ crash" (spending down to near 0 gold to push levels or evolutions) is normal and expected in a real game; a bot that's both maximally leveled and maximally evolved with full items at every stage reads as impossible, not strong.
A few smaller things worth remembering: you don't have to evolve everything — a deliberate 2-star unit is a perfectly normal real-game outcome. Items don't need to be the theoretical best-in-slot either; a real game often ends up with the second-best option instead of the one you wanted.
Getting help and submitting
The in-client Submit button sends your bot straight to #bot-creation for review; questions, feedback on an in-progress build, and general bot-building discussion happen in #bot-builders. Both channels are on the official Discord server: