2nd/10_Wiki/Topics/Architecture/Machinations.md

id, title, category, status, canonical_id, aliases, duplicate_of, source_trust_level, confidence_score, verification_status, tags, raw_sources, last_reinforced, github_commit, tech_stack

id	title	category	status	canonical_id	aliases	duplicate_of	source_trust_level	confidence_score	verification_status	tags	raw_sources	last_reinforced	github_commit	tech_stack
wiki-2026-0508-machinations	Machinations	10_Wiki/Topics	verified	self	Machinations Diagram Game Economy Modeling	none	A	0.9	applied	game-design economy-modeling simulation balancing		2026-05-10	pending	language framework typescript machinations

Machinations

매 한 줄

"매 game economy 의 visual flow simulation". Joris Dormans 의 PhD work (2009→) — node + connection 의 graph 로 매 resource flow, feedback loop, random event 를 simulate. 2026 의 Machinations.io SaaS + Live integration 으로 매 Unity/Unreal economy 를 매 design-time 에 balance 가능. F2P, gacha, idle 게임의 매 standard tool.

매 핵심

매 핵심 nodes

• Pool: resource 저장 (gold, energy, XP).
• Source: resource 생성.
• Drain: resource 소비.
• Converter: resource 변환 (예: 5 wood → 1 plank).
• Trader: bidirectional exchange.
• Gate: 매 conditional flow router.
• End condition: simulation 종료 trigger.

매 connection types

• Resource connection: rate (per tick) + label.
• State connection: source pool 의 value 가 매 다른 node 의 rate modifier.
• Trigger: event-based fire.
• Activator: enable/disable based on threshold.

매 simulation modes

• Deterministic: fixed rates → predictable.
• Stochastic: random with distribution → Monte Carlo.
• Interactive: player input 의 click on Source/Drain.

매 응용

1. F2P economy balancing (gacha rates, soft/hard currency).
2. Idle game progression curves.
3. RTS resource gathering loop.
4. Battle-pass / season pass tuning.
5. Player retention model (LTV simulation).
6. Educational economic modeling (supply/demand).

💻 패턴

1. Idle game core loop (Machinations-style pseudo)

[Source: idle] --1/sec--> [Pool: Gold]
[Pool: Gold] --click 10/click--> [Drain: Spend]
[Pool: Gold] --on >= 100--> [Converter: BuyUpgrade]
[Converter] --1--> [Pool: Multiplier]
[Pool: Multiplier] -- state * 0.5 --> [Source: idle].rate
# Feedback loop: more multiplier → faster gold → more upgrades

2. JS simulation harness (build your own)

type Pool = { name: string; value: number };
type Edge = {
  from: string;
  to: string;
  rate: (state: Record<string, number>) => number;
};

class Simulator {
  pools: Pool[] = [];
  edges: Edge[] = [];
  tick = 0;

  step() {
    const state = Object.fromEntries(this.pools.map(p => [p.name, p.value]));
    const deltas: Record<string, number> = {};
    for (const e of this.edges) {
      const r = e.rate(state);
      deltas[e.from] = (deltas[e.from] ?? 0) - r;
      deltas[e.to]   = (deltas[e.to]   ?? 0) + r;
    }
    for (const p of this.pools) p.value = Math.max(0, p.value + (deltas[p.name] ?? 0));
    this.tick++;
  }
}

const sim = new Simulator();
sim.pools = [{ name: "gold", value: 0 }, { name: "mult", value: 1 }];
sim.edges = [
  { from: "_source", to: "gold", rate: s => 1 * s.mult },
  { from: "gold", to: "mult", rate: s => s.gold >= 100 ? 100 : 0 }, // buy upgrade
];

3. Monte Carlo balance test

function runMontecarlo(n = 1000) {
  const completionTimes: number[] = [];
  for (let i = 0; i < n; i++) {
    const sim = buildEconomy({ seed: i });
    while (!sim.done && sim.tick < 100_000) sim.step();
    completionTimes.push(sim.tick);
  }
  completionTimes.sort((a, b) => a - b);
  return {
    p50: completionTimes[Math.floor(n * 0.5)],
    p95: completionTimes[Math.floor(n * 0.95)],
    p99: completionTimes[Math.floor(n * 0.99)],
  };
}
// Use to validate "median player reaches L20 in ~2 hours".

4. Gacha drop rate modeling

type Rarity = "common" | "rare" | "sr" | "ssr";
const rates: Record<Rarity, number> = {
  common: 0.79,
  rare: 0.17,
  sr: 0.035,
  ssr: 0.005,  // 0.5%
};

function pull(): Rarity {
  const r = Math.random();
  let acc = 0;
  for (const [k, p] of Object.entries(rates) as [Rarity, number][]) {
    acc += p;
    if (r < acc) return k;
  }
  return "common";
}

function pityCounter() {
  let pulls = 0;
  return () => {
    pulls++;
    if (pulls >= 90) { pulls = 0; return "ssr" as Rarity; }
    return pull();
  };
}
// Simulate 10k players → expected SSR per player + variance.

5. Feedback loop classification

Positive feedback: A → ... → A+ (rich-get-richer, snowball)
Negative feedback: A → ... → A- (rubber-banding, balance)
Multiplayer dynamics: Mario Kart blue shell (negative for leader)

// Detect runaway: if d(pool)/dt grows super-linearly → rebalance
function detectRunaway(history: number[]): boolean {
  const ratios = history.slice(1).map((v, i) => v / Math.max(history[i], 1));
  return ratios.slice(-10).every(r => r > 1.1);  // 10% growth/tick sustained
}

6. Progression pacing curve

// Target: L = sqrt(XP / 50) → quadratic XP curve
function xpForLevel(L: number) { return 50 * L * L; }

// Sim: drop rate × kill rate × time = level after T
function levelAtTime(t: number, xpPerSec = 2) {
  const totalXP = xpPerSec * t;
  return Math.floor(Math.sqrt(totalXP / 50));
}
// Plot levelAtTime over 0..7200 (2h) and check pacing.

7. Machinations.io live data hook (2026)

// Fetch live economy params from Machinations.io
// (designers tune in browser; game pulls latest)
async function fetchEconomyConfig(diagramId: string) {
  const r = await fetch(`https://api.machinations.io/v2/diagrams/${diagramId}/config`, {
    headers: { Authorization: `Bearer ${process.env.MACH_TOKEN}` },
  });
  return r.json();  // { goldPerSec, ssRate, levelCurve, ... }
}
// Update server tunables without redeploy.

매 결정 기준

상황	Approach
F2P economy balance	Machinations.io diagram + Monte Carlo.
Idle / incremental	Custom JS simulator (lighter).
MMO economy stability	Agent-based simulation (more nodes).
Single-player RPG progression	Spreadsheet + simple sim.
Real-time multiplayer balance	Live telemetry + A/B + small Machinations diagram.
PVP balance (rock-paper-scissors)	Game theory tool, not Machinations.

기본값: 매 currency-flow heavy game 의 design phase 에 매 Machinations diagram. Pure combat balance 는 매 다른 tool.

🔗 Graph

• 부모: Game Design
• Adjacent: Monte Carlo Simulation · Feedback Loops

🤖 LLM 활용

언제: economy diagram review, drop rate sanity check, progression curve drafting, feedback loop identification. 언제 X: 매 combat balance, 매 narrative pacing, 매 art/UX concerns.

❌ 안티패턴

• Diagram only, no simulation: 매 visual 만 그리고 매 numeric Monte Carlo 의 X. 매 deterministic only 는 stochastic player 의 outlier 을 miss.
• Over-detailed model: 매 50+ nodes — designer 가 매 mental model 잃음. 매 module 단위 분리.
• Static rates in live game: 매 Machinations 에서 balanced 후 hardcode. 매 LiveOps tunable 로 expose.
• Ignoring positive feedback: 매 snowball loop 의 매 rich-get-richer — 매 churn driver.
• Single-seed simulation: 매 1 run 으로 결정 — 매 N=1000+ Monte Carlo 필수.

🧪 검증 / 중복

• Verified (Dormans, Engineering Emergence: Applied Theory of Game Design PhD thesis 2012, Machinations.io docs).
• 신뢰도 A.

🕓 Changelog

날짜	변경
2026-05-08	Phase 1
2026-05-10	Manual cleanup — Machinations primitives + Monte Carlo + gacha/idle pattern