2nd/10_Wiki/Topics/4X_Strategy.md

---
id: wiki-2026-0508-4x-strategy
title: 4X Strategy
category: 10_Wiki/Topics
status: verified
canonical_id: self
aliases: [4X 전략, eXplore eXpand eXploit eXterminate]
duplicate_of: none
source_trust_level: A
confidence_score: 0.9
verification_status: applied
tags: [gaming, strategy, game-design, ai]
raw_sources: []
last_reinforced: 2026-05-10
github_commit: pending
tech_stack:
  language: python
  framework: gym/pettingzoo
---

# 4X Strategy

## 매 한 줄
> **"매 eXplore, eXpand, eXploit, eXterminate — 매 long-horizon empire-building loop."** 매 1993 *Master of Orion* 의 Alan Emrich 가 coined. 매 Civilization, Stellaris, Endless Legend, Old World 의 cornerstone genre — 매 2026 에 AI agents 의 benchmark domain (Pyrrho, Civ-Bench).

## 매 핵심

### 매 4 pillars
- **eXplore**: fog-of-war 의 map reveal, scout units, sensor range, anomaly detection.
- **eXpand**: settler/colony placement, 영토 claim, infrastructure rollout.
- **eXploit**: resource extraction, tech tree, trade routes, population growth.
- **eXterminate**: military doctrine, alliance/diplomacy, conquest victory.

### 매 design tensions
- **Wide vs tall**: many small cities vs few mega-cities.
- **Snowball vs catchup**: 초기 lead 의 compound 의 X — 매 modern designs (Old World, Humankind) 의 catchup mechanics.
- **Turn-based vs RTS**: Civ (TBS) · Stellaris (real-time pausable).

### 매 응용
1. Multi-agent RL benchmark — 매 long horizon, partial observability, sparse reward.
2. Game theory testbed — 매 negotiation, alliances, betrayal.
3. Economic simulation — 매 supply chain, monetary policy.

## 💻 패턴

### Civ-style turn loop
```python
# Pseudocode of 4X turn structure
class Empire:
    def turn(self):
        self.explore()   # move scouts, reveal tiles
        self.expand()    # settle, build improvements
        self.exploit()   # collect yields, research, grow
        self.exterminate()  # military / diplomacy actions

def game_loop(empires, world):
    for turn in range(max_turns):
        for empire in empires:
            empire.turn()
        world.tick()  # natural events, barbarians
        if any(empire.victory() for empire in empires):
            return
```

### Tile yield model
```python
# Civ VI-style tile yields
def tile_yield(tile, improvements, adjacency):
    base = tile.terrain.yields  # food/production/gold
    base += sum(imp.bonus for imp in improvements)
    base += adjacency_bonus(tile, adjacency)
    return base * tile.modifier  # govt, religion, etc.
```

### Tech tree (DAG)
```python
import networkx as nx

tech = nx.DiGraph()
tech.add_edge("Pottery", "Writing")
tech.add_edge("Writing", "Currency")
tech.add_edge("Bronze Working", "Iron Working")

def available(researched):
    return [t for t in tech.nodes
            if t not in researched
            and all(p in researched for p in tech.predecessors(t))]
```

### Combat resolution (Lanchester)
```python
def lanchester_combat(att_strength, def_strength, terrain_mod):
    # Square law for ranged + simultaneous combat
    a, d = att_strength, def_strength * terrain_mod
    ratio = a / d
    att_loss = d**2 / (a + d)
    def_loss = a**2 / (a + d)
    return att_loss, def_loss
```

### MCTS for 4X agent
```python
# Minimal MCTS for turn decisions
def mcts(state, n_sims=1000):
    root = Node(state)
    for _ in range(n_sims):
        leaf = root.select()        # UCB1
        child = leaf.expand()
        reward = child.rollout()    # heuristic playout
        child.backprop(reward)
    return root.best_child().action
```

### Diplomacy as repeated game
```python
class Diplomat:
    def __init__(self):
        self.history = {}  # opponent -> [actions]

    def respond(self, opponent, their_action):
        # Tit-for-tat with forgiveness
        past = self.history.get(opponent, [])
        if not past or their_action == "cooperate":
            return "cooperate"
        return past[-1]  # mirror last
```

## 매 결정 기준
| 상황 | Approach |
|---|---|
| Early game | eXplore 우선 — map info 의 highest marginal value |
| Mid game | eXpand + eXploit — economy snowball |
| Late game | eXterminate or science/cultural victory |
| Tight map | 매 tall play (few strong cities) |
| Open map | 매 wide play (many cities) |

**기본값**: 매 explore-first, 매 expand-until-economy-saturates, 매 specialize.

## 🔗 Graph
- 응용: [[MCTS]]

## 🤖 LLM 활용
**언제**: 매 narrative event generation, flavor text, 매 dynamic dialogue with diplomats.
**언제 X**: 매 hard rule resolution (combat math, yields) — 매 deterministic engine 사용.

## ❌ 안티패턴
- **Snowball lock-in**: 매 early lead 의 unwinnable game — 매 catchup mechanics 의 추가.
- **Micromanagement hell**: 매 50+ cities 의 manual ordering — 매 governors, automation.
- **Tech rush dominance**: 매 single optimal path — 매 multiple viable trees 의 design.
- **Combat spam**: 매 stack-of-doom — 매 1UPT or limited stacks.

## 🧪 검증 / 중복
- Verified (Sid Meier 인터뷰, *Master of Orion* 1993 design notes).
- 매 [[4X 전략]] redirect.
- 신뢰도 A.

## 🕓 Changelog
| 날짜 | 변경 |
|---|---|
| 2026-05-08 | Phase 1 |
| 2026-05-10 | Manual cleanup — FULL content with patterns, design tensions |