Duo queue against a 528 HP Waterfall Giant: pick the pair, not the cards
Two of you queue into Act 1b and the boss pool rolls Waterfall Giant. Solo, it is a 240 HP DPS race where almost every action adds 3 Steam Eruption stacks and the detonation at the end deals damage equal to the stacks unless the boss died at zero. In your duo, the scaling formula puts it at 240 times 2 players times the 1.1 Act 1 coefficient: 528 HP, with the same clock ticking. Whether that fight was winnable was decided at character select, not during the fight, and that is the entire subject of this guide.
Composition in co-op is a damage-math problem before it is a synergy problem. Party damage must roughly double in a duo and reach around 4.4 to 5.2 times solo levels in a quad, while incoming damage per player stays flat. So every party needs scaling damage somewhere, a Block anchor somewhere, and ideally one seat whose mechanics do not care about the enemy's HP multiplier at all. The Necrobinder's Doom execute is the famous version of that third slot, but each of the five characters contributes differently, and picking two with overlapping needs is how parties end up holding 70 percent of the required damage with no way to find the rest.
The roster, in unlock order (Ironclad, then Silent, then Regent, then Necrobinder, then Defect): The Ironclad, 80 HP, Burning Blood healing 6 HP after each combat, carrying the Strength, Vulnerable, Exhaust and Block archetypes. The Silent, 70 HP, Ring of the Snake drawing extra cards at combat start, running Poison, Shivs and the Sly discard engine. The Regent, 72 HP, Divine Right granting 3 Stars at combat start, with the Stars and Forge dual resource system. The Necrobinder, 66 HP, Bound Phylactery Summoning 1 each turn, fighting alongside Osty and stacking Doom. The Defect, 75 HP, Cracked Core channeling 1 Lightning at combat start, with the orb system rebuilt around active channel and evoke play rather than permanent Focus.
What follows is the role model for these five, every duo pairing evaluated, the consensus trio and quad lineups, why duplicates sink runs, the Doom deep dive, the passing decisions composition creates, and the Act 1 draft priorities that turn a good roster into a good run. The version anchor throughout is the v0.111 beta branch, with the v0.108 multiplayer card batch and the v0.104 co-op scaling nerfs already baked in.
The five heroes as co-op seats: tank, amplifier, flex, burst, controller
Community consensus assigns each character a co-op job, and the assignments follow from mechanics rather than preference.
The Ironclad is the main tank. Highest starting HP at 80, plus Burning Blood healing 6 HP after each combat, which compounds across a long campaign into a genuinely larger error budget for whoever fronts damage. His archetypes map onto the job: Strength scaling for when the party also needs him to hit, the Vulnerable package for amplifying everyone else, and the S-tier Exhaust Block build, where Corruption makes skills free-but-Exhaust and Barricade holds Block forever, which in co-op becomes both the party's Block anchor and its Mend priority.
The Silent is the debuff and poison amplifier. Her statuses are shared team-wide in co-op, so Neutralize's free 1 Weak and any Vulnerable she applies multiply everyone's damage, not just her own. Noxious Fumes applying 2 Poison to ALL enemies each turn handles the multi-body boards co-op loves, and Catalyst turns party time into a doubled or tripled clock. Ring of the Snake's extra opening draw makes her the most consistent turn-one setup seat in the game.
The Necrobinder is board control. Lowest HP at 66, but he fights alongside Osty, a giant skeletal hand that attacks on command and body-blocks, explicitly, for the team: damage Osty absorbs is damage no player takes. Bound Phylactery adds Summon 1 at the start of every turn for free, so Osty's HP pool grows passively, and since Unleash scales with Osty's current HP, the sponge slowly becomes a weapon. His damage identity is Doom, the percentage execute, which gets its own section below.
The Regent is burst DPS. Divine Right opens every combat with 3 Stars, and Stars persist between turns with no cap, so his play pattern is banking resources while teammates hold the line, then dumping: Seven Stars spends 7 Stars for 49 AoE, Astral Pulse spends 3 for 14 AoE, and Alignment converts banked Stars into 2 or 3 energy when the hand needs fuel instead of damage. In party terms he converts the safe turns other players create into stored damage.
The Defect is the flex seat. Frost builds, Glacier's Block plus two Frost orbs, make him a second Block anchor; Lightning builds, Tempest channeling a full board of orbs and Ball Lightning, plus the 0-cost Claw cycle that kills faster than anything else in the game, make him the fastest kill seat. Note the StS2 rework: permanent Focus stacking is gone, replaced by active channel and evoke play, with Defragment now a rare and the only stable Focus source. The flex seat is where a trio or quad patches whatever is missing.
Read as a system: the tank holds, the amplifier multiplies, the controller soaks and executes, the burst banks and dumps, the flex fills holes. A party wants each job covered by exactly one player, which is why the next section ranks the pairs.
Every duo ranked: from Ironclad plus Silent to the risky double-burst pairs
Ironclad plus Silent is the consensus duo, and the reasons stack. Every job is covered: Block anchor and front line from the Ironclad, shared debuffs and Poison pressure from the Silent. The synergy is concrete: her Neutralize and Red Mask-style effects keep Weak and Vulnerable up team-wide, and his Whirlwind converts the amplified hits into AoE for multi-body fights; against single bosses, Poison ticks while he blocks, so the pair covers both burst windows and long grinds. If neither player has a strong preference, this is the default answer.
Ironclad plus Necrobinder is the scaling duo. The Ironclad holds the line at 80 HP with post-combat healing while the Necrobinder runs the Doom clock, and Osty body-blocks hits that would otherwise land on players. This pair is the strongest answer to Act 3, where the boss coefficient rises to 1.3 and percent executes stop caring. The weakness is Act 1: both engines come online slowly, and early fights against loops like Nibbit's Strength-stacking hiss punish slow starts, so draft damage early and take the safe path until one engine is running.
Ironclad plus Regent is the comfortable duo. The Regent banks Stars across turns while the Ironclad tanks, then spends them in the windows the tank's Block creates. Strong into DPS races like Waterfall Giant. The structural weakness is that neither amplifies the other: the pair's ceiling is the sum of two solo decks rather than a multiplied product, so it wants a shared-debuff relic like Bag of Marbles or Red Mask to fake an amplifier.
Silent plus Necrobinder is the double-DoT pair, better than it looks on paper: Poison and Doom both ignore Block, and both bypass Skulking Colony's per-hit damage cap, so wall-type enemies collapse against it. The risk is real though: 70 and 66 HP, no true tank, and two engines that want time. Play it with Block discipline and treat Osty as the party's entire front line.
Silent plus Defect works as amplifier plus flex. A Frost-leaning Defect substitutes a soft tank while the Silent multiplies his damage, or a Claw Defect races early acts under her debuffs. Solid, forgiving, slightly under-ceilinged.
Regent plus Defect is the double-burst pair and the risky one: no tank, no amplifier, two players who both want setup turns. Frost Block has to carry all defense, and any boss that punishes development, Knowledge Demon with its escalating card and energy caps, hurts badly. Only for pairs who know both characters cold.
The remaining combinations, Necrobinder plus Regent (controller plus burst, low combined HP), Necrobinder plus Defect (execute plus flex, quietly strong), Ironclad plus Defect (tank plus flex, safe but low amplification), are workable seconds. The pattern in the ranking: a tank plus an amplifier beats two damage seats, and any pair containing the Necrobinder gets better the closer the run gets to Act 3.
Trios and four-player drafts: the consensus lineups and where Necrobinder slots in
With a third and fourth seat, the consensus lineups are stable, and the logic behind them is worth more than the list.
The consensus trio is Ironclad plus Silent plus Regent: tank, amplifier, burst. It extends the default duo with the seat that benefits most from the amplifier's work, since the Regent's banked Stars turn a shared Vulnerable window into a 49-damage AoE spike. It also covers time horizons: the Ironclad holds the present, Poison grinds the middle turns, Star burst closes fights on schedule.
The consensus quad is Ironclad plus Silent plus Defect plus Regent. Same spine, plus the flex seat, and the Defect flexes by need: Frost Block when party defense runs thin, Lightning or Claw damage when the damage check is failing. In a quad the scaling pressure peaks, roughly 4.8 times base HP on Act 3 elites and 5.2 on the Act 3 boss, so the flex seat usually ends up damage-sided.
Where the Necrobinder slots in: substitution, and the timing of it is the skill. The community roles list names him board control rather than a fixed core seat, yet the Doom build data calls him the only S+ co-op character, and the reconciliation is that he is a situational S+. Swap him into a trio for the Regent when the party keeps losing Act 3 boss races, because a percentage execute is the one mechanic the 1.3 coefficient cannot punish. Swap him into a quad for the Defect when the party has damage but keeps dying to board pressure, because Osty soaking hits for the team and End of Days executing a full board are both defense of a kind. What you should not do is run the Necrobinder beside two squishy seats and call the party balanced; his 66 HP means someone else must own the tank job outright.
Four-player logistics change drafting more than most groups expect. Relics drop in player-count batches and are drafted one each, so a quad sees far more relics per run than a duo, and the party should route elite-heavy to exploit it. Shops stay per-player, so gold is individual and hoarding has no party upside. Enemy Artifact, Slippery and Skittish stack higher with player count, so the amplifier's strip-the-Artifact job matters more, and the v0.104 nerfs to exactly those stats are the reason the mode is playable at all right now. Enable Fast Mode before the first fight, because four-player animation pacing is the number one reason quads dissolve before Act 3.
A closing note on Ascension: the party runs at the lowest member's level, progressing separately from solo. A mixed-experience group should treat the first shared run as calibration rather than a ladder push, then decide seats based on where the damage budget actually fell short.
Why two of the same character sink the run: pool competition and build dilution
The community consensus on duplicates is blunt: avoid them, because they compete for the same card and relic pools. Here is the mechanics of why, because the co-op damage math makes it worse than it sounds.
Card rewards are drawn from character pools. Two Ironclads are offered from one pool of Strength, Vulnerable, Exhaust and Block cards, and each engine needs a critical mass of specific pieces: a Strength deck wants Inflame, Demon Form, Limit Break and Whirlwind; an Exhaust Block deck wants Corruption, Dark Embrace, Feel No Pain and Barricade. In a solo run one player collects all the copies and assembles the machine. Split between two players, each gets half an engine and neither deck reaches the point where it out-scales the party HP multiplier. The solo wisdom that taking every card dilutes your deck applies with double force here: in co-op a diluted deck is not merely weaker, it fails the party's damage budget, and the budget is unforgiving.
Relic drafting punishes duplicates even harder. Relics drop in player-count batches and are drafted one each, with ties going to rock-paper-scissors. When Paper Phrog drops, both Ironclads want it; one wins, and the loser's Vulnerable package stays at plus 50 percent while the winner's jumps to plus 75. Two Silents bidding on Snecko Skull is the same story: the winner's every Poison application gains a stack, the loser's gains nothing. In a mixed party, that Paper Phrog would have had exactly one natural owner and zero conflict.
The specific trap this section is named for: two players running the same build. Double Strength Ironclads both want Brimstone, Pen Nib, Demon Form and Limit Break. Every overlapping pick forces a tiebreak, and the outcome is one real deck plus one understudy, a party that spent two seats to get one and a half. The partial fix when two friends insist on the same character is to split lanes completely, one player on Exhaust Block while the other goes Strength scaling, but even then the overlap on staples like Shrug It Off and on every defensive relic keeps bleeding value.
The constructive version of the rule: diversity is arithmetic, not aesthetics. Five characters means five separate card pools and five mostly separate relic appetites, so a party of distinct characters converts every card reward and every relic batch into a pick someone fully uses. Every consensus lineup in the previous section is duplicate-free for exactly this reason. And the diplomatic version, for the friend group where two people main the same hero: one of them takes the neighboring role for party play, the way an Ironclad main picks up the Silent amplifier seat, and both end up with more wins than two half-Strength decks ever produced.
Doom: why a percentage execute is the S+ answer to the HP multiplier
Time to justify the S+ label with the actual scaling math, because Doom is the clearest example in the game of composition answering the multiplier.
The Doom rule: a stacking debuff where, at turn end, an enemy whose Doom is greater than or equal to its max HP is executed. The operative words are max HP. Co-op scaling multiplies enemy HP by player count and an act coefficient, so a quad facing Test Subject #C10 with its 600 base HP is looking at 3120. Flat damage decks have to actually deal 3120. A Doom deck has to stack 3120 Doom, which sounds identical until you notice that Doom application scales with engine quality rather than with enemy HP, and that the execution ignores everything about the bar except its size.
The stacking tools and their scaling properties: Scourge, 1 cost, 13 Doom (16 upgraded), and it draws 1 (2 upgraded), so it costs no tempo and multiple copies stack linearly. Countdown, a 1-cost power, applies 6 (9 upgraded) Doom to a random enemy at the start of each turn, a permanent per-turn annuity; across a long boss fight it alone out-produces most flat damage cards. Reaper Form is the compounding piece: whenever your attacks deal damage, apply that much Doom, meaning every point of party amplification, shared Vulnerable included, feeds the Doom clock too. End of Days is the payoff: massive AoE Doom that immediately executes enemies at threshold, and against scaled multi-enemy boards it deletes phases on the spot.
The resource underneath is Souls: 0-cost tokens that draw 2 cards then exhaust. Reave shuffles them into the draw pile, Haunt converts each played Soul into 6 (8 upgraded) lost HP for a random enemy, and Soulbound exports your Souls into a linked teammate's deck. The Doom build page names the failure mode plainly: balance Soul generation against spending or you stall in boss fights, which in co-op means stalling at 90 percent of the largest HP bar of the run with the whole party watching.
Two co-op-specific interactions complete the picture. First, v0.108 changed the end-of-turn resolution order and Doom no longer kills first, so call your Doom totals and let teammates decide whether their finisher is help or waste. Second, the premier boss answer: Waterfall Giant at 0 HP clears every debuff on itself, goes invulnerable, and detonates for its Steam Eruption stack count, unless it is killed at 0 stacks, and the game's own counter-list names the Doom execute as the way to do it. For most duos that boss is a 528 HP sprint against a growing bomb; with a Necrobinder it is a clock.
Pair the character with a Block anchor and a Beating Remnant (max 20 HP lost per turn) on someone, and the party can stall comfortably to threshold against nearly anything the multiplier prints. That is the whole S+ argument: every other damage type in the game gets proportionally worse as parties grow, and Doom stays exactly as good as its engine.
Tank, damage, support: a working division-of-labor model
Roles only matter if they change drafting and play, so here is the model at the level of actual decisions.
The tank seat, Ironclad by default, owns three things: the party's Block floor, the Mend priority, and the defensive potion funnel. Drafting: Shrug It Off (8 Block, 11 upgraded, plus a draw) and True Grit early, Barricade and Body Slam if the Exhaust Block lane opens. He should not draft like a solo carry; he drafts like the reason everyone else gets to develop. Play: he blocks the telegraphed turns the party calendar identifies, Verdict 33, Heavy Cleave 35, the Frog Knight sub-half-HP 35 beetle charge, and he is the designated survivor when things go wrong, because a downed damage seat revives next floor while a downed anchor can cascade into a wipe.
The amplifier seat, Silent by default, owns the shared debuff layer and the passive pressure. Drafting: upgrade Neutralize early, take Noxious Fumes, Acrobatics (Uncommon since v0.103) for the discard engine, Dagger Throw for the triggers, and Catalyst once the Poison count justifies it. Play: apply Weak before the boss's big attacks and Vulnerable before the party's burst windows, and strip Artifact with cheap applications before committing real stacks against carriers like Mecha Knight.
The damage seats own the scaling check. The Regent drafts Venerate, Solar Strike, Prophesize (draw 6, or 9 upgraded, take on sight) and Genesis, then banks Stars during safe turns and dumps into amplified windows. A damage-sided Defect drafts Ball Lightning, Tempest and Defragment, or goes all-in on the Claw cycle. A damage-sided Ironclad runs Demon Form plus Limit Break with Whirlwind as the AoE converter.
The controller seat, the Necrobinder, owns the board and the clock: Bodyguard and the starter relic's free Summon grow Osty, who body-blocks for the team; Defy contributes cheap shared Weak; Scourge, Countdown and eventually End of Days run the execute.
The flex seat, the Defect, drafts last and fills the hole: Glacier and Cold Snap when the tank is thin, Ball Lightning and Tempest when the damage budget is short, Claw only when going all-in on the 0-cost cycle.
Three rules make the model function. One: at most one pure support; a party of amplifiers has multiplied damage and nothing to multiply it with. Two: every party needs at least one Block anchor and at least one scaling engine, and the flex seat covers whichever is missing. Three: assign seats before floor 1 and only reassign at act boundaries; mid-act role drift is how parties end up with two half-tanks and no damage, which the scaling formula then exposes at the worst possible moment.
Passing decisions: where The Ball goes and who should receive your Souls
Composition creates passing decisions, and co-op gives you exactly four levers: The Ball, Soulbound, the potion flow, and Mend. Each has a decision rule worth stating out loud.
The Ball: colorless, 1 cost, 10 damage, passes to a random teammate after you play it, and permanently gains 15 damage each time it is passed. Because the target is random, you control nothing except whether it keeps moving, so the rule is unconditional: play it whenever you draw it. The party-level reading matters more than the card-level one. The Ball is a shared asset that grows on party tempo, so the right question at the reward screen is never whether your deck wants a 1-cost 10-damage attack; it is whether the party wants a snowballing shared nuke, and in a duo or larger it almost always does.
Soulbound: the Necrobinder card that links a teammate so your Souls enter their deck too. Souls are 0-cost, draw 2, then exhaust, so you are exporting tempo to a chosen player, and the link should go to the seat that converts Souls best. The best convert is a Haunt player: Haunt makes every played Soul drain a random enemy for 6 (8 upgraded), so Souls arriving in that deck are free damage as well as draw. Good converts are draw-hungry engines, a Regent setting up a Prophesize turn or a Silent mid-combo. Bad converts are deliberately thin decks: the Claw archetype specifically suffers when extra cards enter the shuffle, and a deck built on exhaust discipline does not want foreign tokens either. The link is a real composition decision, so make it out loud at the start of the act, and revisit it if the Haunt player joins late.
Potion flow: since v0.109 nearly any potion can target teammates, so treat the combined belt as party inventory. Defensive potions, Block Potion's 12 Block, Ghost in a Jar's Intangible, funnel to the tankiest player when the team's HP runs low. Offensive debuffs, Weak Potion and Vulnerable Potion, go into the boss loop's pressure points. Fairy in a Bottle remains personal insurance for its owner only. Duplicator (the next card played twice) belongs on the burst seat's finisher turn, and Alchemical Coffer is best held by whoever actually empties slots.
Mend flow: campfire Mend heals another player for 30 percent of their Max HP. Priority order: a revived ally sitting at 1 HP, always; the tank below roughly 40 percent with an elite or boss next; otherwise upgrade, because upgrades compound for the rest of the run.
The v0.108 batch of co-op cards, Midnight, Blaze, Blade Symphony, Concoct, Plot, Constellation, Underworld, Hibernate and One For All among fifteen, extends the same principle to every reward screen: before taking a team card, name the teammate it feeds. A pass with no named recipient is a card you drafted for nobody.
Synergies and collisions: pair-specific matchups to plan around
Some pairs multiply each other; some quietly fight. The matchups worth planning around, sorted by category.
Amplifier pairs multiply. Silent Weak and Vulnerable sitting under an Ironclad Whirlwind, a Regent Seven Stars or a Defect multi-hit Lightning turn is the basic engine of party damage: one player's energy multiplied into everyone's cards. This is why every consensus lineup includes the Silent, and why Bag of Marbles and Red Mask are quietly among the best party relics in the game regardless of who holds them.
Stall pairs compound, and Doom plus Frost is the archetype: a Defect running Glacier and Frost Block beside a Necrobinder running Countdown and Scourge can hold almost anything while the Doom clock runs, and Beating Remnant on either seat caps the worst turn at 20 lost HP. The trap is pairing two stall engines into a boss with a hard timer: The Insatiable's Sandpit starts at 4, at zero you lose, period, and each Frantic Escape card in the deck buys one more turn. A stall-heavy pair must draft escape tempo, spend everything, and never exhaust carelessly; if the party is double-stall by design, one seat needs a burst pivot for exactly that fight.
Burst pairs conflict with phase-respawning bosses. Doormaker cycles door and creature phases, and the door respawns at full HP with Strength carrying over; spike-and-reset burst wastes its windows, while Poison persists across phases and DoT decks shine, with the phase 2 one-turn stun as the best window. Test Subject #C10 resets DoTs entirely between its three phases, so a Poison plus Necrobinder pair should budget stacks per phase rather than dumping everything into one bar.
Multi-hit seats collide with specific auras. Phantasmal Gardener's Skittish 6 grants Block on the first hit each turn, punishing Whirlwind and peck-style patterns; that fight wants single big hits or DoT. Spectral Knight's Aura Hex makes every card in hand Ethereal, discarded at turn end if unplayed, which wrecks combo hands, Sly discard setups and anyone banking cards, so kill it right after Magi Knight. Magi Knight's Aura Dampen downgrades all your cards while it lives, specifically punishing upgrade-reliant seats, and it has the lowest HP of the Knight Gang for exactly that reason.
Facing mechanics punish uncoordinated pairs: Kaiser Crab's Surrounded means hits from the claw the party faces away from deal 50 percent more, and the Rocket claw's 31 Laser is the lethal move, so the pair commits to one claw, kills both within a turn or two before the survivor's Strength and 99 Block snowball, and leans on AoE, which hits both claws.
The general rule underneath all of it: read the act pool before assigning seats. Act 1 arrives as either Overgrowth or Underdocks with completely different enemy, event and boss pools, and the pair built to win a Waterfall Giant DPS race is not the pair built for The Insatiable's countdown. Composition is a per-act decision, not a set-and-forget.
Act 1 draft priorities by seat: what each role picks up first
The final layer: what each seat should actually take in Act 1, because composition decisions at floor 1 only pay off through drafting discipline.
Ironclad tank: Shrug It Off and True Grit for the Block floor, Inflame if party damage is covered elsewhere, and starter Strike removal at shops. Relic priorities at draft: Anchor, Sai, Pantograph, Beating Remnant. Skip greedy damage rares unless the party is dangerously damage-light going into Act 2.
Silent amplifier: Noxious Fumes is the key power, Neutralize upgrade early, Acrobatics now that it is Uncommon, Dagger Throw for the discard triggers, Catalyst once Poison stacks justify it. Relics: Snecko Skull is the mandatory one, and Red Mask amplifies for the whole party no matter who finds it.
Regent burst: Venerate and Solar Strike for income that also deals damage, Prophesize on sight (draw 6, or 9 upgraded), Genesis the moment it appears because it defines the archetype for the run. Decide the Star versus Forge lane early and never draft across it; The Smith only enters a Forge deck. Relics: energy and draw, Ectoplasm if the party can eat the gold lockout, because Alignment converts Stars but the rest of the hand still costs energy.
Necrobinder controller: Scourge on sight, Bodyguard for Osty, Reave plus Haunt for the Soul engine, Countdown for the passive clock, End of Days whenever the Doom count supports it. Relics: Beating Remnant to stall to threshold, plus anything that caps or heals loss. He should also claim the Soulbound link decision early, per the passing rules.
Defect flex: decide the lane by party need before the first reward screen, Glacier and Cold Snap for Block, Ball Lightning and Tempest for damage, Defragment the moment it appears because it is the only stable Focus source in the game. Claw only if going all-in on the 0-cost cycle, and if so, guard the deck's thinness fiercely: no Affliction-priced blessings, no speculative picks.
Party-level rules to close. One: skip cards that serve no seat; the party damage budget punishes diluted decks harder than solo ever does. Two: aim every co-op card at a named teammate before taking it. Three: relic drafts follow the seat map, engine relics to their engines, defense to the tank, economy to whoever snowballs hardest. Four: re-check the composition at each act boundary against the upcoming boss pool, and if the party keeps losing long fights, the answer is almost always a Doom seat, not more flat damage.
That is the whole model: pick distinct characters, assign the five jobs, draft into the job, pass to the seat that converts best, and let the Necrobinder answer whatever the Act 3 multiplier throws at the party.