Skip to content

Concept: The Dive — the run becomes the session

Pitch

One drop-in, many floors, and nothing is yours until you carry it out. A raid stops being a single map with an exit at the end and becomes a descent: floor after floor, with the loot you are carrying compounding, your wounds carrying with you, and an extraction pad on every floor asking the same question — bank this now, or take it deeper?

The player's whole session becomes one story with one arc, instead of a chore repeated three times. Extraction stops being the button that ends the errand and becomes the decision the game is about.

Problem / opportunity

The action pivot (../action-pivot/concept.md) rebuilt how Shadow Raiders plays — steering, dodging, flanking, reading enemies, being pressured off the map. The team's G0 playtest called that loop fun and froze its numbers. But the pivot deliberately did not touch the shape of a session, and that shape is still exactly the one the July publisher test measured: pick a floor, drop into one map, crack chests, walk to a pad, spend the proceeds on gear slots, repeat.

The publisher's directive after that test was a new extraction game built from the current assets and part of the mechanics — not an iteration of the current game. A build with better controls and the same silhouette does not answer that directive. This concept spends the one structural change the project can afford on the layer that most defines what game it is.

Three findings point at the same hole, from three directions:

  • The G0 checklist's weakest verdicts are all the same missing thing. "Did you ever hesitate at a safe exit?" — so-so. "Does loot loss sting?" — so-so. "Is the raid the right length?" — "feels okay I guess." Nothing in a raid accumulates, so nothing about leaving it is a decision. Every extraction is a formality.
  • The measured funnel dies at a wall, not on a slope. Floor 1-5 — the mandatory boss — carries a fail/complete ratio of 43.9, roughly triple anything else in the game. The structure offers a player who cannot beat that boss no way forward and no way to bank the run that got them there.
  • Sessions are one raid long. ~9.4 minutes per session against a raid timer that dictated it. The session has no reason to continue past the first extraction, because the first extraction resolves everything.

The opportunity is that the game already contains the pieces of a descent — a floor ladder, a descend portal that is distinct from the safe exits, a boss at the bottom, a "keep your gear on death" promise — and has simply never asked the player to carry anything between them.

The concept

You only keep what you carry out. Loot picked up during a dive is unbanked. Every floor offers safe exits that bank the haul and end the run, and one descend portal that banks nothing and continues it. Death costs the haul collected since drop-in — never your gear, never your progress on the ladder.

Depth pays, twice. Deeper floors hold better things, because danger already drives what the world spawns. And the act of descending itself multiplies what the run is worth when it finally banks — so a long dive from the surface and a short dive from deep in the ladder are genuinely different plays rather than one dominant strategy.

Your wounds come with you. Health does not reset between floors. Alongside the growing haul, attrition becomes the second thing that compounds, and the two of them together are what makes the question on each pad real: you extract because you are rich, or because you are hurt, and usually because you are both.

Every descent makes you stronger. Boarding the descend portal offers a choice of three upgrades — one that helps you survive, one that helps you kill, one that helps you carry more out. They last only for the dive. This is the run-arc that lets depth escalate without becoming hopeless, and it is the beat that turns a longer raid into an actual run.

The clock stops being a punishment and becomes a current. Extraction pads close on a schedule, as they already do, but the descend portal never closes and time never kills you. The pressure no longer ends your raid — it pushes you down, closing off the cheap ways out until the only door left is the one that takes your fortune deeper.

The boss becomes the last temptation instead of a wall. The floor that holds the biome's boss also holds ordinary safe exits. Walking away from the boss with a great haul is a real, respectable way to finish a dive; beating it banks the same haul and opens the next biome. The hardest thing in the game stops being a gate players fail against and becomes the highest-stakes version of the question they have already answered four times that run.

References

Grounding note: the project's gd-knowledge MCP corpus was not reachable in the session that produced this document, so nothing here is cited from it. The MISFITZ figures below are carried forward from the project's own earlier research (../action-pivot/concept.md), where they are recorded as studio-reported and unaudited. The remaining entries are named as design-pattern references — structural comparisons, not sourced metrics.

  • MISFITZ (Antihero Studios) — the publisher's stated reference and the pivot's anchor. Relevant here for its reported inversion of the death/retry relationship: players who died retried at a higher rate than players who extracted (studio-reported, unaudited). A dive sharpens exactly that relationship, because a death now has a story — a specific fortune, lost at a specific depth, that the next run exists to win back.
  • Naavik's mobile-extraction analysis; Badlanders and Lost Light as cautionary tales (already in docs/research/) — casual mobile extraction works when the core ruleset survives casualization and the friction is cut. It fails when the wrong thing is casualized: Badlanders made loot loss meaningless. The dive is the direct answer to that specific failure — it does not make loss harsher, it makes loss legible, by giving the player a number they watched grow.
  • The press-your-luck descent as a structural pattern — cash out now or go one level deeper for a bigger pot — is the oldest reliable tension in games and is the actual novelty here relative to the raid-shaped competition. Named as a pattern; no metrics claimed.
  • The project's own Megabonk-style ladder (PROJECT_CONTEXT.md §8.5.1) — the biome/floor structure this concept re-reads. The ladder was already built; it has only ever been used as a save-checkpoint, never as a run.

Fit with Shadow Raiders

The dive keeps every pillar the project has committed to. Gear is never lost — the signature twist is untouched and, with a compounding haul to lose instead, it finally has something to contrast against. Ladder progress is never lost on death. The action pivot's entire verb layer — steering, the dash, flank crits, Peace/War, the perception rules — is unaffected; the dive changes what a session is, not what a fight is.

The dive re-reads three systems rather than replacing them. The descend portal, which has always been strictly better than a safe exit and therefore never a choice, becomes the risky door. The closing-extraction pressure, validated at G0 as the storm's replacement, keeps its mechanism and inverts its meaning from pushing you out to pushing you down. And the floor picker stops meaning "which raid am I replaying" and starts meaning "how deep do I drop in".

The dive retires two rules deliberately, both flagged in the GDD: the hard timer death, which cannot coexist with an accumulating haul, and the rule that the last safe pad eventually closes, which would convert the design's central choice into coercion.

The dive feeds the meta loop that already exists rather than competing with it. The 8-slot gear and blueprint economy is untouched; the upgrade cards are run-scoped and evaporate at the end of every dive. This is a hard boundary, and it is what keeps this a single structural bet instead of a second progression system that must be balanced against the first.

Risks & open questions

  • Loss calibration. The design deliberately makes loss sting, in a game whose audience is casual. Rails exist (gear safe, ladder safe, the Safe Pocket, a forced heal option when badly hurt), but whether they are enough — or too many, blunting the tension back to nothing — is the central playtest question and cannot be answered on paper.
  • Session length becomes player-controlled. A dive is as long as nerve holds. That is the point, but it means session length stops being a number the design sets and becomes a distribution it observes. Worth watching for the opposite failure of the baseline: sessions that run too long for a mobile sitting.
  • Economy inflation. A depth multiplier stacks on top of a loot table that already improves with depth. The gear and blueprint economy was tuned against single-raid banking; the dive hands it hauls of a different order. This needs the combat-and-economy re-anchor pass, not a guess.
  • Does the shallow dive stay worth playing? If deep dives dominate the economy, floors 1–2 become a corridor players resent rather than a place they choose to bank. The start-depth choice is designed to prevent that; it needs verification.
  • Rivals at depth. Rival raiders currently read a fat haul as a reason to turn on you. Under a dive that condition is permanently true after a couple of floors. Whether deep divers should be hunted — flavourful — or whether that makes the back half of every run a rival gauntlet is an open tuning question.
  • Does the descend-upgrade choice belong here at all? It is the only net-new subsystem in the bet and the only part that could be cut without unmaking the design. If the dive reads well without it in the first playtest, cutting it is cheaper than balancing it.