You Are Choosing a Branch, Not a Node
Floor 1 of every act drops you at the bottom of the map with two or three lanes crawling toward the boss icon, and the most expensive routing mistake in Slay the Spire 2 gets made in the first ten seconds: judging a lane by its first node. A ? node at the bottom means nothing on its own. What matters is the sequence behind it, because the moment you step on a node, your future options collapse to whichever branches connect upward from it. Routing is the art of paying attention to what you are giving up three floors before you actually give it up.
Learn the seven node types first: Combat, Elite, Shop, Rest, Event, Chest and Boss. Each converts a different resource, and we will price them individually in the next section. Before that, internalize the three questions that evaluate any branch. How many rest sites does it contain, and where? How many elites does it expose, and are they optional or forced? Where are the junctions where you can still change your mind? A branch that looks greedy but keeps two exits open is often safer than a modest branch that commits you at floor 2.
Version matters here. This guide targets v0.111 beta, and two recent changes reshaped early maps: v0.101 took elites off floor 6, turning the opening six floors into a near-free greed window, and from v0.105 onward the boss pools churned (Aeonglass replaced Doormaker in Act 3, and v0.109 nerfed it again). Memorized routes from launch-year content rot fast. What does not rot is branch discipline: reading the whole lane, counting rests, and refusing to let one shiny node make a five-floor decision for you.
One structural habit separates decent routers from good ones: rerun the evaluation after every node. The map is not a plan you set at floor 1 and execute; it is a plan you re-underwrite fourteen times. A lane that was correct at floor 3, when you were healthy and needed gold, becomes wrong at floor 7 after the elite took 30 HP out of you, and correct again at floor 11 after a rest and a shop. Every time the state changes, the value of every remaining node on your branch changes with it, and the junction you preserved last floor is what lets you act on that.
A final framing for the whole guide: nodes do not have fixed values, they have prices and payoffs quoted against your current run. The sections ahead price each node type, give you a two-gate test for elites, a decision tree for campfires, and worked lines through each act. Treat all of it as underwriting, not as a script. The map hands you information for free; the only question is whether you read it before or after you commit.
What Each Node Is Worth, Ranked With Reasons
Ranking nodes is unpopular among high-level players because context bends every answer, but beginners need a default order before context exists. Here is ours, quoted against a healthy deck in mid-act, with the conditions that move each node up or down.
Elite sits on top, conditionally. The payoff is the largest in the game: a guaranteed relic drop plus gold and a card reward, and elites are the only reliable renewable source of relics outside shops and events. Relics compound; Black Star, an Ancient-tier find, makes elites drop an additional relic, and the earlier you take it the harder the whole run snowballs. The condition is survival, which is the entire subject of the next section. An elite you can beat is the best node on the map; an elite that beats you is a run-ending node, which is why greedy elite deaths are among the most common losses in Early Access.
Rest ranks second, and its value swings with position. Two rests in the first eight floors are nearly worthless to a healthy deck; the rest directly before an elite or before the boss is often the single most valuable node on the map. When you evaluate a branch, count rests the way you count exits: as insurance whose worth depends entirely on when it pays out.
Shop ranks third on a gold-adjusted basis. The product you actually want is card removal, because every Strike or Defend that leaves the deck raises the average quality of every future draw, with situational cards, relics and potions behind it. A shop you reach with 150-plus gold around floors 8 to 12 is a strong node; the same shop reached broke at floor 4 is furniture. Membership Card, when stocked, halves all prices and rewards planning a whole branch around the shop that sells it.
Event, the ? node, ranks fourth on average and has the widest variance of any type. The current build holds 66 events across act-specific pools, running from pure boons (Self-Help Book empowers a card with no downside) to scripted traps (refusing at Wongo's downgrades one of your cards). Events also trade HP for power more often than any other node, so their value rises with your health and falls with your desperation.
Combat ranks fifth, and being fifth is a feature: it is the floor value of the map. Gold plus a card reward plus a small, predictable HP cost, with full information about what you are fighting. Combats are the default you take when nothing better is reachable or when your state says heal-and-stabilize.
Chest ranks last, which still means free. A relic with zero combat risk and zero agency; you never path into a chest, but you never dodge one either.
Now the three deck states that reorder the list. Healthy and strong: Elite over Rest over Event over Shop over Combat. Average: Rest and Elite swap, shop climbs with gold. Bleeding: Combat over Rest over Shop over Event, and elites leave the list entirely until a campfire fixes you.
When to Fight an Elite: The Two-Gate Test
There is no fixed formula for elite count anymore, and our FAQ says so plainly: fight them once you have answers and a rest site before them. What we can give you is a two-gate test to run at the junction, thirty seconds, honest answers only.
Gate one is the deck gate: can this deck produce the fight's requirements? Translate the local elite pool into requirements. Front-loaded damage: can you output roughly 20-plus damage by turn three? Act 1 elites range from about 61 HP (Phrog Parasite) to 140 (Terror Eel), and every extra turn an elite lives is a turn it hits you. A Block plan for the elite's worst telegraphed turn, meaning actual Block cards in the deck, not aspirational ones. And one answer keyword if the local pool demands it: AoE for Phrog Parasite's death-spawned Wrigglers, multi-hit or damage over time for Skulking Colony's per-hit cap, exhaust tools for the Burns Mecha Knight will shove into your hand.
Gate two is the health gate: can your HP budget afford the attempt? Workable defaults for a first clear: enter Act 1 elites above 65% HP, Act 2 elites (Infested Prism sits at 200) above 70%, and treat Act 3's Mecha Knight, a 300 HP machine widely called the deadliest elite in the game, as opt-in only. The gate has two sub-checks. A rest site before the elite, not after; healing you have not collected yet is not money you have. And a potion backup in the belt: a Block Potion or Weak Potion converts directly into a survived turn, which is why hoarding them for the boss is a routing error as much as a combat error.
Skip conditions deserve equal billing. Skip the elite when the deck fails gate one, when the only path to it funnels through a dead-end branch with no exit, or when the act's boss pool punishes the elite's damage type (going into Knowledge Demon at 40% HP because Infested Prism ate your bar is a trade no one should accept). Also skip, honestly, when you do not know the elite's pattern yet; a first meeting with Terror Eel or Soul Nexus is a tuition payment, and it is fine to pay it on a run you can afford to lose.
The strategic reason the test matters more in StS2 than it did in StS1: elites are your relic engine, and this game moved relic acquisition away from between-act boss drafts into Ancients (act-start NPCs offering Blessings) and elite drops. A run that dodges every elite arrives in The Hive underpowered and gets graded by Infested Prism anyway. One elite per act is the survival minimum on a normal run, two is the snowball configuration, three usually requires either Black Star or a deck that came together early. When in doubt, take the elite that is one node after a rest, and leave yourself the junction exit if the fight goes badly.
Read the Elite Pool at Act Start, Then Route Against It
Every act publishes its elite pool, and the pool is a matching puzzle: your deck is either equipped for specific members or it is not. Routing without reading the pool is shopping without a list. Here is what each pool asks for, act by act, in v0.111.
Act 1a, Overgrowth: Bygone Effigy (127 HP) rewards setup decks, because its sleep phase is a free window to bank Powers and resources; the wake-up turn is the one your Block has been waiting for. Byrdonis (91) alternates Dive and Peck, steady pressure that rewards Weak on the Peck turns and a fast clock. Phrog Parasite (about 61) seeds Infection statuses into your discard pile between Lash turns, then bursts into 4 Wrigglers on death, stunned for their first turn only: this elite is a questionnaire with one field, do you have AoE?
Act 1b, Underdocks: Skulking Colony (79) puts a ceiling on each individual hit, so decks built around one enormous swing starve here; the Forge-heavy Regent swinging his Sovereign Blade wants multi-hit or damage over time riding along. Terror Eel (140) is a staying-power check, ten turns of blocking. Phantasmal Gardener is four bodies of roughly 26 to 31 HP with Skittish 6, which converts your opening hit of each turn into Block for them: multi-hit decks feed the garden, single big hits kill one body at a time, and you prioritize whichever is nearest its Enlarge turn.
Act 2, The Hive: Infested Prism (200) is a raw output check with no trick; either the engine is online or the branch should not include this node. Entomancer (145) keeps the board growing with bees all fight, so bring sweeping damage or a rush plan aimed straight at the caster. Decimillipede is three segments around 40 HP each, and any segment that dies while its siblings live comes back at 25 HP, which converts AoE from asset into trap: focus-fire decks want this elite, sweep decks fear it.
Act 3, Glory, runs a five-elite pool. The Knight Gang (Magi Knight 82, Spectral Knight 93, Flail Knight 101) is one package with a fixed kill order: Magi first (Aura Dampen downgrades your cards while it lives, and it winds up toward a 35-damage Magic Bomb), Spectral second (Aura Hex makes your hand Ethereal), Flail last while blocking its Rams. Mecha Knight (300) wants exhaust for 4 Burns, discipline to skip attacking on Windup turns, and 40-plus Block saved for the 35-damage Heavy Cleave. Soul Nexus (234) opens with Soul Burn for 29 every time, so it is literally a test of whether 29 Block on the opening turn is within your deck's reach.
The routing rule that falls out of this: at every act start, name the elite you are hunting and the one you are dodging, then prefer branches that serve the first and starve the second. Branches differ by one or two nodes more often than players notice, and against a pool this readable, that single node is the whole decision.
The Campfire Decision Tree: Rest, Upgrade, or Rarely Something Else
Campfires are where routing converts into power, and the default beginner error is treating rest as the only option. Health above one is spending money, and most of the time the campfire should be spending it on an upgrade instead. Here is the tree, top to bottom.
Question one: price the next two nodes against your health bar. Name them off your branch, list the worst telegraphed turn in each (Owl Magistrate's Verdict is 33 plus 4 Vulnerable; Mecha Knight's Heavy Cleave is 35; Vantom's turn-3 swing is 25-plus). If a realistic sequence of those turns can take you to zero, rest. Healing is boring and frequently correct.
Question two: do I hold a keystone upgrade that changes fights? Some upgrades are not improvements but transformations. Body Slam at 1 cost is a card; Body Slam at 0 cost is the finisher of a Block deck. True Grit's random Exhaust becomes your choice, the starter motor of every Exhaust deck. Glacier climbs to 9 Block while channeling two Frost. If a card at this power level is sitting unupgraded in your deck, the campfire is an upgrade, full stop, and the rest can wait one more fight.
Question three: am I near the boss with health to spare? Above roughly 70% HP with the boss inside three floors, upgrade the card you will actually play in that fight, usually your best Block source or your finisher. Pantograph owners can push that threshold higher, since it refills 25 HP whenever a boss fight opens, effectively converting the final campfire into a free upgrade for anyone above half health.
Question four, the default: upgrade the card you play most. Not the rarest card, the most played one. An upgrade on a card you cast six times a fight outperforms an upgrade on a flashy card you cast once, and frequency is something you can actually count from your last few combats.
Two situational branches complete the tree. In co-op, Mend heals another player for 30% of their Max HP, and triage logic applies: funnel the heal toward whichever player is tanking, and use Mend to lift a revived ally off 1 HP, since a fallen player returns at exactly that. And Eternal Feather changes the math of rest itself, paying 3 HP per 5 deck cards at every campfire: a 20-card machine heals 12 per rest, which quietly turns napping into a real option for decks that have grown past thin.
One anti-pattern to name: rest-spamming out of fear. Fear-driven resting upgrades nothing, shortens no fights, and just schedules the same decision for the next campfire with a slightly higher health bar. If question one said no and question two found no keystone, the answer was question four, and the answer was upgrade.
Routing the First Half: Floors 1 Through 8
The first half of an act (floors 1 to 8, roughly the phase before elite HP starts gating you) has one mission: convert safe nodes into the resources that make the second half's decisions possible. Gold, two or three good cards, and a mapped elite plan. Nothing else matters yet.
Start with combats, plural. Early normal fights are the cheapest card-and-gold engine on the map, and with v0.101 clearing elites off floor 6, the opening stretch is safer than it has ever been: something like Combat, Event, Combat, Combat, Combat across the first five floors costs almost nothing. The cards you want out of these rewards are damage now: Anger and Pommel Strike on Ironclad, Backstab on Silent, Astral Pulse or Solar Strike on Regent, Scourge on Necrobinder, Ball Lightning or Cold Snap on Defect. The first-half test is whether you can delete a 60 HP normal in three turns; if you cannot, keep taking combats and keep picking damage.
The first elite window opens once two conditions hold: one or two answer cards in the deck (the two-gate test from earlier) and a rest site on the branch before the elite node. Not after it. Pathing an elite with the rest behind you converts your insurance into a story you tell later. Ideal first-elite geometry is Rest at floor 6 or 7, Elite at 7 or 8, with a junction immediately after the elite so a bad fight (you lose 35 HP to Terror Eel) still leaves you a soft branch of combats and an event instead of a forced second elite.
Slot the first shop between floors 8 and 12 on a branch you were taking anyway. The agenda is removal first: gold spent deleting a starter Strike or Defend buys improvement that never expires, and it beats buying any card. Budget gold backward from the shop, meaning if the branch gives you a shop at floor 9, your floors 1 through 8 combats are what fund the removal, so do not squander gold on events that sell nothing you need. Two common relic accelerators justify moving the greed up a floor or two: Festive Popper (9 damage to all enemies at combat start) and Akabeko (8 Vigor on the opening turn) plus Lantern (extra energy on turn one) turn early fights from leaky into free, and free fights mean the health budget for an earlier elite.
Events in the first half are usually worth their node because the pools price HP trades cheaply while you are healthy: Abyssal Baths soaks you for 3 damage to gain +2 Max HP repeatedly, and the Door of Dark at Doors of Light and Dark deletes exactly the card you name.
The half ends with a simple audit before you cross floor 9: count your answers, count your gold, count your HP. The second half of the act will spend all three, and the branch you pick at the midpoint is the contract that says how.
Routing the Back Half: Dead Ends, Escape Hatches and the Pre-Boss Ritual
The back half of an act (floors 9 onward, phase three territory) is where the map starts closing doors. Branches merge, options thin out, and the two structural hazards you are routing against are dead ends and commitment traps.
A dead end is a lane that funnels you into a node you cannot afford with no exit after it: the forced second elite at 30% HP, the mandatory event when every option costs health, the final stretch with no rest between you and the boss. Dead ends are visible in advance, which is the good news. Count the rest sites on each candidate branch the way a lender counts collateral: on a typical act you want two usable rests after the midpoint, with the last one positioned directly before the boss. A branch with zero rests and two elites is a dead end at any health below 80%, no matter how good the nodes look individually.
An escape hatch is a junction you preserve precisely so you can downgrade greed after new information arrives. The classic sequence: you path toward elite number two at 70% HP, the fight goes long, you limp out at 35%, and the junction you kept lets you pivot into combats and a shop instead of the second elite the greedy lane would have forced. The site's own pathing philosophy calls this branch discipline, and in Early Access, where act variants and patch churn keep breaking memorized formulas, keeping one pivot open is worth more than any single node upgrade. Practically: when two branches look equal, take the one that converges later, because converging later means you keep the option longer.
The pre-boss ritual is the last three floors before the boss, and it should be boring by design. Final rest, or final upgrade if Pantograph or a fat health bar covers you. Potions stocked: two-plus slots filled, and if a shop sits on the final stretch, Alchemical Coffer (fills every potion slot instantly) is a legitimate closer purchase. Answers loaded: name the boss and the two turns in its loop that can kill you, and confirm you have the Block or the potion for each. If the branch gives you a Chest near the end, take it; a free relic with zero risk is the best node on any tired map.
Bleeding-state routing deserves its own paragraph, because it is where runs are actually saved. Below 30% HP with the boss four floors away, the list reorders hard: Combat over Rest over Shop over Event, no elites at any price, and the shop becomes a potion store first (Block Potion, Weak Potion) and a removal service only with leftover gold. Lizard Tail holders get one stretch of extra courage, since the once-per-run revive at 50% Max HP is functionally a second health bar; everyone else should remember that a boring Act cleared at 25% HP still pays out, and a heroic elite death pays nothing.
One last back-half note for co-op routers: the map is shared, but campfire logic is not. With Mend in the option list, decide rests and Mends as a party before anyone locks in, because a heal spent on the wrong player is a rest you do not get back.
How Many Question Marks to Take, and When
Event nodes are the map's variable-rate instruments: high average value, high variance, and completely dependent on which act's pool you are drawing from and what state you are in. Sixty-six events exist in the current build, distributed across act-specific pools, with the alternate Act 1 variants (Overgrowth, Underdocks) carrying exclusive entries. Routing them well is mostly about knowing which category of event the local pool serves you.
Boon events are pure upside and justify detouring a node or two. Self-Help Book (any act) empowers an Attack, Skill or Power in your deck for free, always take it. Amalgamator (The Hive) removes 2 starter cards and hands you an upgraded Ultimate replacement in exchange, which is close to two free removals plus a card; almost every deck should accept. The Future of Potions? converts a spare potion into a card reward; with refilling potion drops, that is free value more often than not.
Resource events trade fungibles for deck quality, and they are the reason healthy decks take extra ? nodes early. Doors of Light and Dark offers two random upgrades (Light) or one targeted removal (Dark); the community pick is Dark, because targeting beats randomness. Wellspring's Bathe option removes a card at the price of a temporary curse, worth it whenever removal options are scarce. Trash Heap digs for a unique relic with HP, a good trade when the health bar is long. Sunken Treasury tempts with the large chest (lots of gold, permanent curse) and the answer is usually the small chest; gold-for-curse trades are almost always bad in this game.
Trap events are the variance tax. Wongo's looks like a shop and refuses you for free, except refusing downgrades one of your cards, so never pick the refuse option. Affliction events in general are the ones to steer branches around: Afflictions latch onto existing cards and cannot be removed at shops, which makes them the only event category worth actively avoiding rather than merely discounting. And the hidden chain rewards planning: picking up the Foul Potion at Potion Courier unlocks The Merchant??? later, where throwing the potion triggers a hidden fight paying 300 gold, knock-off relics and the exclusive Merchant's Rug (not available in co-op; a healthy deck with AoE should route toward it).
The density heuristic, then: take ? nodes in the first two-thirds of an act while above roughly 60% HP and your deck still wants removal or relics; take combats instead when bleeding, because combats pay in predictable small increments while events occasionally charge HP. And weigh event value against the branch, not the node: a ? that sits next to a rest is worth more than a ? that commits you to an elite lane, because event outcomes are exactly the kind of new information that makes an escape hatch valuable.
Worked Lines Through Each Act (and Why They Bend)
Three sample routes, one per act, to show the reasoning in motion. Do not memorize these lines; act variants and patches will invalidate them. Steal the bends.
Act 1, Overgrowth, healthy Ironclad at 80 HP. Combat, Combat, Event, Combat, Rest, Elite, Combat, Shop, Combat, Rest, Elite, Event, Rest, Boss. The bends: the first Rest at floor 5 is positioned before the first Elite (say Byrdonis at 91 HP), not because you need healing but because insurance precedes risk, and the Rest can still be an upgrade since Burning Blood covers combat drips. The Shop at floor 8 has your accumulated combat gold behind it, roughly enough for the Strike removal that should be your first purchase. The second Elite is conditional: take it above 60% HP with an AoE answer for whatever the pool rolls, dodge it otherwise via the junction after the shop. The floor 13 Rest is non-negotiable, because the boss pool (Vantom's Slippery loop, The Kin's acolytes, Ceremonial Beast's race) all punish entry below half health.
Act 2, The Hive, Silent with poison ticking. Combat, Event, Combat, Elite, Rest, Combat, Combat, Shop, Rest, Combat, Elite, Chest, Rest, Boss. The bends: this line front-loads two combats before the first Elite because The Hive grades builds, and if the first elite rolled is Infested Prism (200 HP), you want Noxious Fumes online and stacking before you knock. Entomancer (145) is the preferred target for an AoE-light poison deck since the Poison itself answers the bee summons; Decimillipede is the one to avoid, because its Reattach mechanic taxes exactly the attrition plan poison runs on. The mid-act Shop doubles as a safety valve: if the first elite went badly, the same branch can drop to combats and spend the gold on potions instead. Knowledge Demon on the boss floor is the nightmare matchup for slow decks, so the final Rest before the boss is a rest unless you are above 70%, and the potion belt should hold something that says either race (Fire Potion) or survive (Block Potion, Ghost in a Jar for one Intangible turn).
Act 3, Glory, any deck. Combat, Event, Rest, Elite, Combat, Shop, Combat, Combat, Rest, Elite, Rest, Combat, Rest, Boss. The bends: identify which elites your branch exposes before committing. The Knight Gang (Magi 82, Spectral 93, Flail 101) is the reasonable target for most decks because the kill order is known and the combined HP is manageable; Mecha Knight (300) is opt-in only with exhaust answers for its 4 Burns and 40-plus Block for Heavy Cleave; Soul Nexus (234) demands 29 Block on turn one, which you can count in advance from your own deck. Three rests in the back half is not cowardice: Test Subject #C10 spreads 600 HP across three phases and The Queen's Chains of Binding tax every turn, and both fights are lost in the last third, exactly where a tired health bar gives out.
All three lines share one signature: every greedy node sits after a rest and before a junction. That is the whole discipline in one sentence. Elites feed the relic engine, rests convert time into safety, and junctions keep you honest about which of the two you currently need. The map in v0.111 will keep changing under everyone's feet; that sentence will not.