settled-reach/docs/workshops/v01-gap-analysis/round1-nigel.md

title, description, type, status, workshop, agent, round, created

title	description	type	status	workshop	agent	round	created
Workshop Round 1 - Nigel: Sandbox & Replayability Analysis	Replayability analysis of concept proof claims, seed-based variation gaps, and replay ceiling estimates	workshop	archived	v01-gap-analysis	nigel	1	2026-02-11

Workshop Round 1 - Nigel: Sandbox & Replayability Analysis

Track 1: Strength of Concept Proof — "What happens the SECOND time you play this?"

---

Claim 1: Asymmetric Information Is the Core Mechanic

Tickets: #138-142 (info boundaries), #182-183 (divergent knowledge/relationships)

Playthrough 2 analysis:

Player plays smuggler first. Learns the ring, the contacts, the dead drops. Then plays detective. They KNOW who's dirty. Can they use that knowledge?

My answer: metagaming is a FEATURE, not a bug — but only if the system doesn't let it trivialize the game.

Dramatic irony is one of the most powerful emotions in storytelling. Playing the detective KNOWING the smuggling operation from the inside is compelling. The player thinks "I know who you are" when they see an NPC — but can the detective CHARACTER act on that? No. The information boundaries (#138-142) gate what the detective can officially know, investigate, or act on. The player's knowledge creates TENSION, not shortcuts.

BUT — and this is where D-029 earns its keep — the entanglement ratio varies per seed. So playthrough 3 (detective, different seed) is a player who THINKS they know the pattern from playthroughs 1 and 2, but the ratio has shifted. The bar owner who was innocent in seed A is complicit in seed B. Player overconfidence becomes a trap. That's anti-metagaming gold.

Ticket coverage: 7/10

The info boundary system is well-ticketed. What's missing:

• No ticket for "metagaming-as-dramatic-irony" — the system should be designed to reward, then subvert, cross-playthrough knowledge. This is an authoring consideration for Mellanie's content packs: include NPCs who LOOK guilty to a metagaming player but aren't in some seeds.
• #182 (divergent starting knowledge) and #183 (divergent relationships) are HIGH not CRITICAL. For replayability proof, these ARE critical. If both characters start with similar knowledge, the "different game" feeling collapses immediately.

Risk: MEDIUM. The architecture supports replayability. The content needs to actively exploit it.

---

Claim 2: Fog of Perception Creates Tension

Tickets: #110-113 (shadowcasting to fog rendering)

Playthrough 2 analysis:

The honest question: does the player learn the map once and fog becomes meaningless?

Short answer: at v0.1 scale, partially yes. And that's okay IF the NPCs vary.

The station district is a fixed layout. After playthrough 1, the player knows where walls are, where rooms are, where sightlines exist. Fog doesn't hide ARCHITECTURE the second time. What fog hides is ACTIVITY — who is where, what they're doing, what's happening behind that wall right now.

This means fog's replayability value depends ENTIRELY on NPC routine variation. If the bar owner is always at the bar at 8pm, the fog over the bar at 8pm communicates nothing new on playthrough 2. If the bar owner is sometimes at a secret meeting at 8pm (seed-dependent), fog creates genuine uncertainty even for returning players.

Critical dependency: Fog replayability = f(routine variation). Ticket #88 (daily routine system) is critical priority but its scope description needs to ensure seed-based timing variation. Ticket #178 (seed-based variation) is LOW priority. This is a problem.

Ticket coverage: 6/10

The fog SYSTEM is well-covered. The fog EXPERIENCE on replay is not. What's missing:

• No explicit connection between seed-based NPC schedules and fog-discovery variation
• #156 (procedural district filler) at MEDIUM could help — even minor spatial shuffles (furniture placement, blocked corridors, construction zones) would make fog rediscovery meaningful
• No ticket for "fog memory decay" testing — D-011 says fog returns when you leave, but there's no validation that this creates genuine re-exploration

Risk: HIGH for replay, LOW for first playthrough. Fog will work beautifully on playthrough 1. Its replay value hinges on NPC variation that's currently underpriotitized.

---

Claim 3: NPCs Feel Alive Without Player Attention

Tickets: #88 (daily routines), #93-99 (simulation tiers), #101 (routine execution)

Playthrough 2 analysis:

The bar owner question: is the bar owner always at the bar?

With the 10-axis NPC model (D-024), each NPC's daily routine is ONE of 10 axes, generated per seed via the NPC generation pipeline (#92). So in theory, the bar owner in seed A has a different routine than the bar owner in seed B.

But wait — the bar owner IS the bar owner because they own the bar. Their ROLE constrains their routine. The bar template (#185) defines a role that includes "operates the bar." Can you have a bar owner who's rarely at the bar? That creates its own interesting dynamics (who's covering? what are they doing instead?), but it also risks breaking template coherence.

What should vary between runs:

• Routine timing — same activities, different schedule. Bar opens at different times per seed. Meeting happens Tuesday vs Thursday.
• Routine deviations — some seeds, the bar owner takes walks at midnight. Others, they drink in their office after closing. These are the secrets you discover through observation.
• Relationship-driven changes — if the bar owner and the logistics hub manager are in a triangle, their routines INTERACT differently per seed based on relationship state.
• Entanglement effects — an entangled bar owner has secret meeting routines that an unentangled one doesn't.

Ticket coverage: 7/10

The simulation tier system (#93-99) is well-designed and the background state machines mean NPCs genuinely change while you're not watching. Routine execution (#101) makes them act on those states. But:

• #88 (daily routine system) doesn't explicitly require seed-based VARIATION in routines
• #92 (NPC generation pipeline) is HIGH not CRITICAL — this is the pipeline that creates per-seed variation!
• No ticket for routine DEVIATION mechanics (what makes an NPC break their routine?)
• #105 (tolerance threshold triggers) is MEDIUM — but tolerance threshold crossings ARE the moments that create "alive without attention" proof

Risk: MEDIUM. The tier system architecture is sound. The routine generation needs to be explicitly designed for variation.

---

Claim 4: Observation Is a Mechanic, Not a Cutscene

Tickets: #115 (NPC vision), #124-128 (sound model), #119-122 (internal monologue)

Playthrough 2 analysis:

Can the observe-notice-follow-discover chain produce DIFFERENT discoveries in different runs?

The chain:

1. Observe — see NPC behavior through fog (vision system)
2. Notice — internal monologue flags something unusual (monologue system)
3. Follow — player decides to tail the NPC (player agency + NPC routines)
4. Discover — find the secret (information boundary reveal)

For this chain to vary per run, EACH link needs to produce different outputs:

• Observe: Different NPCs in different places (seed-dependent) = YES varies
• Notice: Different tells (#90) flagged by monologue = DEPENDS on tell variation per seed
• Follow: Different destinations (routine variation) = DEPENDS on #88 seed design
• Discover: Different secrets (D-029 entanglement variation) = YES varies

The chain's ENDPOINTS vary well (different conspiracies, different entangled NPCs). The chain's MIDPOINTS — what tells look like, how the monologue interprets them — need more variation.

Critical gap: The tell system (#90) is HIGH priority and is load-bearing for observation replayability. If "hesitating before answering" always means guilt, players learn the meta. Tells need to vary per seed — some seeds, hesitation is meaningful; others, it's just a personality trait. D-029's variable entanglement helps here, but the tell POOL needs to be rich enough that the same behavior doesn't always map to the same meaning.

Ticket coverage: 6/10

The mechanical pipeline exists. What's missing:

• Tell variation per seed — not explicitly in #90's scope
• Monologue variation for returning players — does the character notice DIFFERENT things on different runs? #121 (character voice variation) helps but is MEDIUM priority
• No "false positive" mechanic — observation should sometimes lead to dead ends. Unentangled NPCs with suspicious-looking routines are essential anti-metagaming
• #123 (unreliable narrator) is LOW priority but is a replayability multiplier — different characters interpret the same observation differently

Risk: HIGH. The observe chain WORKS once. For it to work differently on replay, tell variation and false positives need explicit design attention.

---

Claim 5: Same World, Different Story

THE critical claim for replayability. This is my hill.

Tickets: #179-183 (character builds), #196-199 (success criteria)

Playthroughs 1 & 2: Smuggler vs Detective (same seed)

Rating: 8/10

This is well-designed. The smuggler's daily life IS the detective's investigation target. Same NPCs, same locations, inverted relationships. The smuggler knows who's in the ring and sees the detective as a threat. The detective sees suspicious patterns and the smuggler as either a suspect or a potential informant.

D-027 specifically requires "both playthroughs feel like fundamentally different games." Tickets #180 (smuggler build) and #181 (detective build) are CRITICAL. Divergent starting knowledge (#182) and relationships (#183) are HIGH.

What makes this work:

• Different ACCESS TIERS to the same dialogue (D-028 layer 1) — smuggler gets insider talk at the docks, detective gets deference at the precinct
• Different TRIANGLES are visible — smuggler sees criminal triangles, detective sees institutional ones
• Different MONOLOGUE voice — same NPC behavior, different character interpretation
• Different ROUTINE relevance — the smuggler cares about shift changes (opportunity windows), the detective cares about pattern deviations (suspicious activity)

What could undermine this:

• If the station district is too small, the spatial experience is identical. Both characters walk the same corridors to the same 3 locations. The SOCIAL experience differs but the PHYSICAL experience doesn't.
• If dialogue content (#190-192 content packs) doesn't have enough access-tier variation, conversations feel similar despite different character builds

Playthroughs 3 & 4: Same characters, different seed

Rating: 4/10 — THIS IS THE GAP

Here's where I start shouting.

There is NO ticket that explicitly validates "same character, different seed = different experience." Ticket #197 (criterion 2: fundamentally different playthroughs) tests smuggler vs detective. It does NOT test smuggler-seed-A vs smuggler-seed-B.

What SHOULD change per seed:

• Which NPCs are entangled (D-029) — #175, #176 are MEDIUM priority, should be HIGH
• Which Tier 1 module is active (D-023) — #158 is MEDIUM priority and only Tier 1 drama module schema, not the pool draw
• Which NPCs fill which template roles (#92, #176) — #176 is MEDIUM, should be HIGH
• Triangle configurations (#106-109) — #107 is HIGH, good
• Tell assignments (#90) — variation per seed not explicitly scoped
• Routine timing (#88) — variation per seed not explicitly scoped

The seed-based variation ticket (#178) is LOW priority. This is the single biggest replayability risk in the entire v0.1 plan.

Without #178, playthroughs 3 and 4 are playthroughs 1 and 2 with different characters. The replay ceiling drops from 6-8 runs to exactly 2.

Ticket coverage for cross-seed replay: 4/10

Risk: CRITICAL. The ticket plan proves two-CHARACTER replayability but not two-SEED replayability. The infrastructure exists in the decisions (D-023, D-029) but the implementation tickets are underpriotitized.

---

Claim 6: Social Triangles Create Emergent Drama

Tickets: #106-109 (triangle generation), #175-176 (entanglement ratio)

Playthrough 2 analysis:

Do triangles generate differently across seeds?

With the triangle generation system (#106-109), triangles are defined at template level (2 intra-template, 1 cross-template per site). With 3 social sites, that's at minimum 9 triangles involving ~20 NPCs.

The combinatorial question: how many MEANINGFULLY DIFFERENT triangle configurations are possible with ~20 NPCs and 3 sites?

Each triangle needs 3 NPCs with specific relational properties. With 20 NPCs:

• 3 sites x 4-8 NPCs each = 12-24 role slots
• Intra-template triangles draw from within site (C(8,3) = 56 possible per large site)
• Cross-template triangles connect across sites
• Per-seed NPC generation (#92) fills roles differently each time

Theoretical variety: HIGH. The combinatorial space is large enough for 10+ meaningfully different configurations.

Practical variety: MEDIUM. Role constraints reduce the effective space. The "bar owner and two regulars" triangle has limited configurations because the role structure constrains who can be involved. The triangle TYPES may repeat even as specific NPCs change.

Player-created triangles: NOT COVERED IN TICKETS.

This is a gap. If the player befriends two NPCs who hate each other, that's a player-created triangle. If the player becomes a regular at the bar and the bar owner starts treating them as a confidant, that reshapes existing triangles. The relationship dynamics system (#103) is MEDIUM priority and might support this implicitly, but there's no explicit design for player-inserted triangle dynamics.

Ticket coverage: 6/10

Triangle generation is reasonably covered. What's missing:

• Player as triangle vertex — no ticket
• Triangle DISSOLUTION mechanics — what happens when a triangle resolves? New one forms?
• Cross-seed triangle variety validation — are triangle types diverse enough?
• #108 (cross-template triangles) is MEDIUM but cross-template triangles are where the most surprising emergent stories come from

Risk: MEDIUM. Triangle generation will work. Triangle SURPRISE across playthroughs depends on role flexibility and NPC pool variety.

---

Special Evaluation: D-029 Entanglement Ratio (30/50/20)

Does variable entanglement actually prevent metagaming?

YES — but only if the variance is wide enough and the tell system cooperates.

The math at v0.1 scale (~20 NPCs):

• 30% flat = 6 NPCs (wallpaper, routine + greeting only)
• 50% mundane triangles = 10 NPCs (real drama, no conspiracy connection)
• 20% entangled = 4 NPCs (conspiracy-linked)

With variance (say 25-35/45-55/15-25 per seed):

• Low-entanglement seed: 3 entangled NPCs out of 20. Investigation is a needle-in-haystack.
• High-entanglement seed: 5 entangled NPCs out of 20. More suspects, faster discovery, but more red herrings too.
• The module attachment ratio (60-70/30-40 known vs stranger) further varies WHICH 3-5 are entangled.

The anti-metagaming effect: A player who calibrated "roughly 1 in 5 NPCs is suspicious" from playthrough 1 finds that calibration WRONG on playthrough 2. They can't build a reliable suspicion threshold. The uncertainty IS the replay value.

Gap: Tickets #175 (entanglement ratio configuration) and #176 (NPC pool generation) are MEDIUM. For anti-metagaming to work at v0.1, these need to be HIGH. The variable ratio is meaningless if it's not implemented.

Risk: MEDIUM. The design is sound. The implementation priority doesn't match the design's importance.

---

Special Evaluation: Seed-Based Variation (#178)

Is this sufficient for replayability at v0.1 scale?

#178 is LOW priority. This is insufficient. Full stop.

Seed-based variation is not a nice-to-have for replayability — it IS replayability. Without it, the game has exactly two meaningfully different playthroughs (smuggler + detective). With it, the game has 6-8+ meaningfully different playthroughs.

What #178 needs to control at v0.1:

1. NPC role assignment — which generated NPCs fill which template roles
2. Entanglement selection — which NPCs are conspiracy-linked (D-029)
3. Entanglement ratio — what percentage of NPCs are entangled per seed (D-029 variance)
4. Triangle configuration — which NPCs form which triangles
5. Routine timing offsets — same routine types, different schedules per seed
6. Tell assignment — which behavioral tells are meaningful vs personality quirks
7. Module attachment — which Tier 1 module(s) activate and which NPCs they attach to

Items 1-4 are architectural (the seed determines generation outputs). Items 5-7 are behavioral (the seed determines how the world FEELS different).

Recommendation: PROMOTE #178 to HIGH. Split into sub-tasks if needed. This is the replayability engine.

---

Replay Ceiling Estimate

Station district: 3 social sites, ~20 NPCs, 2 characters

Scenario	Playthroughs	Confidence
Without seed variation	2 (one per character)	HIGH — smuggler vs detective is well-designed
With seed variation (current LOW priority)	4-6	MEDIUM — depends on variation breadth
With seed variation + tell variety + routine offsets	6-8	MEDIUM — diminishing returns from small NPC pool
With all above + player-created triangles	8-10	LOW — speculative, needs mechanic design

For a v0.1 vertical slice, 6-8 meaningfully different playthroughs is excellent. Most vertical slices are one-and-done. But we only get there if seed-based variation is properly prioritized.

The comparison test: Can two players describe completely different games?

• Smuggler vs detective: YES, absolutely
• Same character, different seed (current plan): MAYBE — not enough ticket coverage
• Same character, same seed, different play choices: PARTIALLY — social triangles create some divergence, but the conspiracy structure is fixed per seed

---

Summary: Priority Adjustment Recommendations

PROMOTE (replayability-critical, currently underpriotitized)

Ticket	Current	Recommended	Reason
#178 Seed-based variation	LOW	HIGH	The replayability ENGINE. Without it, replay ceiling = 2
#175 Entanglement ratio config	MEDIUM	HIGH	D-029 anti-metagaming requires implementation
#176 NPC pool generation	MEDIUM	HIGH	Per-seed NPC assignment is foundational
#182 Divergent starting knowledge	HIGH	CRITICAL	Core differentiator between characters
#183 Divergent relationships	HIGH	CRITICAL	Core differentiator between characters
#108 Cross-template triangles	MEDIUM	HIGH	Where the most surprising emergent stories come from

NEW TICKETS NEEDED

1. Same-character-different-seed validation test — Success criterion alongside #197. Two smuggler playthroughs with different seeds must produce measurably different experiences.
2. Tell variation per seed — Extend #90 to ensure behavioral tells are seed-dependent, not fixed. Hesitation doesn't always mean guilt.
3. Routine timing variation per seed — Extend #88 to include seed-based schedule offsets. Same routine types, different timing.
4. False positive design — Unentangled NPCs with suspicious-looking (but innocent) routines. Essential anti-metagaming.
5. Player-as-triangle-vertex — Mechanic for player actions to create or modify social triangles.

CONCEPT PROOF SCORECARD

Claim	First Playthrough	Second Playthrough (other char)	Third+ (same char, new seed)	Overall Risk
Asymmetric information	9/10	8/10	6/10	MEDIUM
Fog of perception	9/10	5/10	5/10	HIGH (replay)
NPCs alive	8/10	7/10	5/10	MEDIUM
Observation as mechanic	8/10	6/10	5/10	HIGH
Same world, different story	9/10	8/10	4/10	CRITICAL
Social triangles	8/10	7/10	6/10	MEDIUM

Bottom Line

The v0.1 plan proves the two-character concept beautifully. It does NOT yet prove the replayability concept.

Playthroughs 1 and 2 (smuggler vs detective) will deliver. The architecture is sound, the information boundaries are well-designed, and the character-as-lens approach creates genuinely inverted experiences.

But playthroughs 3 and beyond — same characters, different seeds — are where The Settled Reach's long-term value lives. And that's where the ticket plan has its biggest gap. Seed-based variation (#178) at LOW priority, entanglement configuration (#175) at MEDIUM, NPC pool generation (#176) at MEDIUM — these are the tickets that turn a two-play demo into a replayable game.

The DECISIONS are right. D-023 (pool-based modules), D-029 (variable entanglement), the variable attachment ratios — these are brilliant anti-metagaming designs. But the tickets that IMPLEMENT those decisions are deprioritized relative to their importance.

My single most important recommendation: promote #178 (seed-based variation) from LOW to HIGH. Everything else is incremental improvement. This one change is the difference between "play it twice" and "play it eight times."

And it nails replayability without us engineering it! The SYSTEMS do the work. We just need to make sure the seed reaches deep enough into those systems.