Open any other guide on this site and the Artificer is the farm: the engine favors her against exactly one hero out of 43, and one matchup file clocks her printed kill at 56 turns. Then read the fine print those same files keep admitting — the math can't see robots. Shock Bot quietly adds 3 to every attack she boosts, Heal Bot claws back the big hits, and a free Synth arrives every upkeep whether her dice cooperate or not. The scores are real. The conclusion isn't.
Your first decision isn't which attack — it's which robot, and her matchup files pick the build order by opponent, not by instinct. Into Ninja and Pyromancer the coaching is Shock Bot first, and build lean — a fast Basic Shock Bot beats saving for perfect upgrades, because against their pace you need teeth immediately and Heal Bot buys its extra turns second. Into Cursed Pirate and Seraph the files flip it: Heal Bot first, because surviving the sprint is the entire matchup. Meanwhile, Spanner Strike is the default swing at 98%. Schematics is your accelerator: 60% of dedicated turns pays 4 Synth, 2 cards, and 2 health — one roll that buys a Basic Bot and its first activation. Take it when the opponent's board is quiet and take Spanner when it isn't. And set the rhythm early: Scrap Collector hands you a Synth every upkeep no matter how the dice behave, so even your worst roll leaves you one turn closer to a working machine.
The middle turns are where the printed numbers stop describing your hero. Priority one: upgrade Shock Bot to Advanced. The upgrade costs 3 Synth, but an Advanced Bot activates for 1 — exactly what Scrap Collector pays you every upkeep — so from that point on, plus-3 on every attack is free in real terms. Eureka is the payoff swing: 49% odds for 7 damage plus 3 Synth, and with Shock Bot live it's a 10. When Eureka misses you still own Spanner's 98% floor, and a Synth spent on its three bonus dice is worth about 3.5 expected damage in total — roughly 2.3 more than the free die. The Nanite game needs its own robot: build the Nanobot, stack toward three Nanites — Jolt at 59% applies one, and 4 spare Synth applies one with no dice at all — then pay its activation and detonate for 5 during upkeep, damage that never sees a defense roll. Two budget rules: Synth stops at 7, so income wasted at the cap is tempo handed back, and a Nanite without a Nanobot is a promise you haven't kept yet.
Endgames with the Artificer are about converting a built board into one oversized turn. Zapppp! lands on 25% of dedicated turns for 9 plus a Nanite plus one free Bot activation — 12 with Shock Bot. Overclock, when four bolts arrive, deals 6 undefendable and fires two different Bots free, so a boosted Overclock is an undefendable 9 that skips the defense roll entirely — though raw prevention tokens still eat it; only an Ultimate pierces everything. A table-rules note: official errata says the Bot activation on Overclock, Zapppp!, and the ultimate happens BEFORE damage is calculated — the 'then' printed on the board is a misprint — so the Bot's damage genuinely joins the attack. Don't throw a boosted swing into a live negation token — jab with Spanner until it's spent, then unload. Maximum Power! at five bolts is a 1.3% event with a 13-damage true ceiling, and the rules give opponents exactly one out: altering your dice before it activates. That cuts both ways — anything you hold that can alter dice in a roll phase is your insurance against THEIR ultimate, because nothing else in the rules stops one.
Her die is 3 wrenches, 2 gears, 1 bolt, and that lopsided split writes every reroll rule. Spanner Strike needs three wrenches — a coin flip on the opening roll, 98% across three attempts — so wrenches are the default keep and there's no bad-luck excuse for a wasted turn. The interesting decisions live in the gears: three gears showing forks between Eureka (two wrenches beside them, 49% with best play) and Schematics (one bolt, 60%) — read the board, not the dice: swing when Shock Bot is live, bank when it isn't or when you're buying robots. A scattered sequence points at Jolt, the four-in-a-row line at 59% for 7 damage plus a Nanite and a Synth. The bolt is the six face and the trap face: dedicated Overclock fishing succeeds 10% of the time and ultimate fishing 1.3% — never spend rerolls hunting bolts. Take Overclock and Maximum Power! when the bolts arrive on their own, and commit to Zapppp!'s five-in-a-row (25%) only when four sequential faces land by the second roll.
The Artificer runs two treasuries, and Combat Points are the smaller one — Synth builds the robots, but CP keeps the machine's lights on. Income is 1 CP per turn (capped at 15) plus 1 per card sold, and the start player skips their very first Income phase — going first, sell a dead card early. Two standing budget rules. First, respect the Knockdown tax: standing up costs exactly 2 CP or your entire attack phase, and her matchup files single that choice out as brutal for her — losing an offensive phase also loses the Synth riders attached to it, so keep 2 CP banked into any hero who can knock you down. Second, spend CP on protecting timing: her damage scales on Synth, not Combat Points, so she can afford to fund interaction — and the one thing she cannot rebuild is a stolen setup turn.
Honesty first: her action-card deck and upgrade paths aren't in the verified data yet, so this guide makes zero named-card promises — DiceSlayer will verify her deck and upgrades on camera. What her board already guarantees is card advantage most heroes lack: standard income draws 1 per turn, and Schematics draws 2 more on 60% of dedicated turns, so a banking Artificer sees more cards than the table. The discipline that follows from the rules: sell what you will never play — every card converts to a flat 1 CP — and treasure anything that alters dice during a roll phase, because pre-activation dice alteration is the only Ultimate counter in the rules, and the hero with zero per-hit prevention needs that out more than most. On upgrades, one caution: Hero Upgrades can never be sold back, and her worst matchups are racers — into Ninja or Pyromancer, be suspicious of any early permanent spend that isn't keeping you alive.
Deny her turns, not her dice. Every quiet turn you give her is a Synth from Scrap Collector plus whatever her roll banks — the machine assembles itself while you wait. Chip constantly so Schematics turns feel expensive — but chip smart, not blind: every normal-damage hit pays her about 1.3 Synth and a roughly coin-flip Nanite through Ratchet Up. Tax her build turns with stuns and Knockdown-class effects, which her own matchup files call brutal against a hero who needs setup windows; and starve her rerolls where you can, because Overclock and the ultimate live and die on bolt hunts. Save your negation — a dodge token, an attack-cancel — strictly for the Bot-boosted swings: her files coach her to jab until your token is spent. And if your hits can land at 5 or less, keep them there — Heal Bot only wakes on hits of 6 or more.
Three windows. The widest is turns one through four, before any robot exists: her defense roll prevents zero and her per-hit mitigation is literally nothing until Heal Bot is built, so every point you land pre-Bot is full price. The second is the turn right after a build or upgrade — the Synth just left her bank, and the attack that turn is probably an unboosted Spanner. The third is any turn she holds Nanites but no Nanobot: her chip pressure isn't real until a Nanobot cashes it, and the stacks can evaporate on upkeep rolls before she ever does.
Nobody bans the Artificer, and the raw math explains why: the engine favors her against exactly one modeled hero, and opponents' scores against her run into the mid-60s. Mostly that instinct is right — if your bring-3 includes even one racer, your ban belongs elsewhere: Iceman, Ninja (65.5 from his side of the math), Barbarian — whose file calls their pairing the most lopsided math in the set — and Pyromancer all end the game before her machine exists, and Moon Elf's natural plan is accidentally anti-Bot, with 4-to-5 damage hits that never wake Heal Bot while Entangle starves her bolt rerolls. The exception is an all-engine roster: her best scores are into slow builders — Alchemist 51.8, Necromancer 49.4, Santa 48.6 — and those are printed numbers that ignore her Bots entirely. If all three of your heroes hand her ten quiet turns, the 'free win' across the table is the one B-tier hero who genuinely outgrows you. That's the day her name belongs in the ban conversation.
She's a specialist's pick, not roster glue. The community parks her in B tier on sparse tournament results, the engine's raw scores favor her against exactly one modeled opponent, and she carries the highest official complexity rating in the game — a real practice tax before she earns a seat. What she offers a bring-3 is a dedicated engine-hunter: into slow builders the Bots compound faster than their boards do, and she draws essentially zero bans, so she never costs you strategic capital. Pair her with two faster heroes who cover the racer pods she cannot survive, and reveal her at the Alchemist and Necromancer tables of the world.
Her clock problem is the worst kind for 3/1/0 scoring: the fastest printed kill quoted anywhere in her matchup files is 15.2 turns (into Shadow Thief), and the honest ones read 18, 31, 56. Bots pull real games far below those numbers, but her plan still opens with build turns, and a strategy that needs five setup turns plus a grind is a draw waiting to happen — and a draw is barely better than a loss. When the round clock is short: Basic Shock Bot immediately, skip the Advanced-upgrade greed, lean on Spanner's 98% for turn-in, turn-out tempo, and remember: an ugly boosted Eureka that ends the round scores 3; a beautiful engine that times out scores 1.