magicciv/.project/reports/simulation/README.md

Simulation Report Index

Structure

simulation-report/
├── README.md                    <- this file (index + top-line verdict)
├── mechanics.md                 <- what the simulation models vs. designed-but-unwired
├── baseline/                    <- latest iteration baseline (current: iter 7o)
│   ├── report.md                    findings + balance verdict
│   ├── stats.md                     raw numbers + tables
│   └── story.md                     evolutionary narrative
└── scenarios/                   <- 5-scenario sweep at current balance
    ├── 0ai/                         pure ecology, no civilization
    │   ├── report.md
    │   ├── stats.md
    │   └── story.md
    ├── 1ai/                         solo player vs ecology
    │   ├── report.md
    │   ├── stats.md
    │   └── story.md
    ├── 2ai/                         two players competing
    │   ├── report.md
    │   ├── stats.md
    │   └── story.md
    ├── 3ai/                         three players (canonical bench)
    │   ├── report.md
    │   ├── stats.md
    │   └── story.md
    ├── 4ai/                         four players, max pressure
    │   ├── report.md
    │   ├── stats.md
    │   └── story.md
    └── comparison.md            <- cross-scenario comparison + unified story

Reading order

1. baseline/report.md -- current ecology balance state and what it means
2. mechanics.md -- what the simulation actually models vs. what's designed but not yet wired
3. scenarios/comparison.md -- how the world plays differently at 0/1/2/3/4 AI
4. scenarios/Nai/story.md -- dive into individual scenario narratives

Latest baseline

Iter 7o -- 589 named species, 13 lair-forming species, T10 apex across all maps. Rat Snake + Caiman lineages dominate. All balance targets met across all 5 scenarios.

Scenario Sweep Results

Post-iter-7o sweep (2026-04-09). 5 scenarios run (0-4 AI players), map scaled by player count (48 + 16 * (n-1)), 50K ticks, 500 turns, seed=42. First sweep where ALL scenarios meet ALL balance targets cleanly.

Scenario	Map	Lairs	T4-T6 Kill Rate	Target 10-30%	T7-T10 Kill Rate	Target 40-70%	Encounters	Deaths	Elapsed
0AI (ecology only)	48x48	168	--	--	--	--	0	0	564.0s
1AI (militarist)	48x48	168	27%	pass	62%	pass	9,222	5,211	641.7s
2AI (mil+exp)	64x64	282	20%	pass	60%	pass	19,618	11,087	870.8s
3AI (canonical)	80x80	484	22%	pass	63%	pass	24,855	14,459	1,107.5s
4AI (max pressure)	96x96	672	24%	pass	63%	pass	31,136	18,334	1,641.4s

10/10 balance targets met. No edge cases, no warning flags.

Previous sweep (Phase 7 / iter 7d) -- historical comparison

Scenario	T7-T10 KR (old)	T7-T10 KR (new)	T4-T6 KR (old)	T4-T6 KR (new)
1AI	73% over	62% pass	7% under	27% pass
2AI	71% edge	60% pass	17% pass	20% pass
3AI	81% FAIL	63% pass	13% pass	22% pass
4AI	73% over	63% pass	11% pass	24% pass

The 3AI scenario -- previously the worst miss at 81% T7-T10 kill rate -- now lands at 63%, cleanly inside the 40-70% window. The spatial index optimization from iter 7f combined with balance tuning from iter 7d and encounter probability scaling brought all scenarios into target.