#Magic Vault 5×3 — Draft Math Design

Game ID: magic-vault-5x3

Artifact type: complete game pack using a reusable mechanic

Lifecycle target: draft-v0.1

Deployment status: Dev only / Not for production / Not for certification

Approved design direction: 2026-08-10

#Purpose

Create a simple 5-reel × 3-row, 10-payline template game that demonstrates the approved hold_and_win/v1 runtime in a complete game pack. This Draft is for math authoring, simulation and future frontend integration. It must not be frozen, published to production, enabled for a real operator or described as production-ready.

#Identity and ownership

• Workspace: games/magic-vault-5x3/.
• Canonical backend draft config: libs/mathengine/games/magic_vault_5x3/magic-vault-5x3.yaml.
• The new game owns its base symbols, reels, paytable, trigger distribution and whole-game RTP.
• It reuses hold_and_win/v1; it does not fork or rewrite the mechanic state machine, RNG draw order, replay, persistence or settlement behavior.
• hold_and_win_reference remains an immutable reference pack. Its economic and state-machine parameters are reused deliberately, but its base reels and whole-game RTP are not evidence for Magic Vault. Magic Vault uses the approved 5-column × 3-row presentation rather than the reference pack's 3-column × 5-row presentation; both contain 15 cells, but runtime parity must bind the Magic-specific cell ordering explicitly.

#Presentation

Theme: a fantasy magic vault with heroes, gems and prize coins.

Existing art source: /Users/admin/frontmvp/template_symbols.

symbol_bonus.png, symbol_scatter.png, symbol_mystery.png and fs-background.png are out of scope for Draft v0.1. No new art is required.

#Base game contract

• Grid: 5 reels × 3 rows.
• Evaluation: 10 fixed paylines, left-to-right.
• Winning combinations: consecutive matching symbols starting on reel 1.
• WILD substitutes for H1–H4 and L1–L4; it does not substitute for COIN.
• COIN appears on the base reels and has no line payout.
• Six or more visible COIN symbols trigger Hold & Win.
• The triggering COIN symbols become the initially locked Hold & Win cells in their corresponding grid positions; the feature does not redraw or replace them when it opens.
• Round max-win cap: 500× total bet, applied once at the existing runtime cap stage.

Initial payline geometry:

1. middle-middle-middle-middle-middle;
2. top-top-top-top-top;
3. bottom-bottom-bottom-bottom-bottom;
4. top-middle-bottom-middle-top;
5. bottom-middle-top-middle-bottom;
6. top-top-middle-bottom-bottom;
7. bottom-bottom-middle-top-top;
8. middle-top-top-top-middle;
9. middle-bottom-bottom-bottom-middle;
10. top-middle-top-middle-top.

Base reel weights and line payouts are Draft tuning parameters. Symbol ordering, Wild behavior, COIN trigger behavior, line count and payline geometry are design-locked for Draft v0.1.

#Hold & Win contract

Reuse the accepted hold_and_win/v1 economic and state-machine parameters unchanged. The only layout adaptation is the already approved 5-column × 3-row grid:

• Grid: 5×3, 15 cells.
• Trigger: 6+ COIN in the base outcome; all triggering coins carry into the feature as its initial locked cells.
• Initial respins: 3.
• Draw only empty cells in canonical cell order.
• Any new landing locks and resets respins to 3.
• A miss decrements respins by one.
• Settle at zero respins or a full grid.
• No additional wager during respins.
• Landing chance: 43,000 PPM per empty cell.
• Regular coin values and weights:
  • 1×: 30;
  • 2×: 25;
  • 3×: 18;
  • 5×: 15;
  • 10×: 10;
  • 20×: 2.
• Prize-kind weights:
  • regular: 1000;
  • Mini: 20 at 10×;
  • Minor: 5 at 25×;
  • Major: 1 at 100×.
• Grand: 500× only on a full grid; it replaces the ordinary coin sum.
• Ordinary settlement is capped once at 500×.

The accepted reference conditional mean is approximately 32.4461×. Draft verification must confirm that Magic Vault uses the same mechanic distribution rather than silently retuning it.

#Math targets

The game intentionally puts most value in a frequent feature.

The 55.44% base target is a consequence of the approved decision to keep the reference mechanic unchanged while triggering it about twice as often. It is not a generic recommendation for future production games.

Acceptance bands for the first Draft evidence:

• trigger interval: 1 in 75–85;
• total RTP: 96.0% ±0.3 percentage points;
• max award: never above 500× and 500× reachable;
• exact base/trigger results where tractable;
• seeded feature and whole-game simulation with rounds, seeds, build/config hash, sigma, SE and CI;
• RTP decomposition reconciled within reported numerical/sampling uncertainty.

If no reel/paytable solution satisfies trigger, base and total targets without changing a locked rule, stop at AWAITING_USER_APPROVAL — DRAFT MATH with measured trade-offs. Do not silently change the mechanic.

#Draft artifacts

The implementation plan will create:

• games/magic-vault-5x3/README.md;
• games/magic-vault-5x3/gdd/magic-vault-5x3-gdd.md;
• games/magic-vault-5x3/math/magic-vault-5x3-math-spec.md;
• versioned Draft/tuning changelog;
• evidence reports under games/magic-vault-5x3/evidence/;
• canonical backend Draft pack/config and focused tests.

The authored workspace must record backend config path, Git SHA and config SHA-256. It must not duplicate an uncontrolled canonical config.

#Gates and exclusions

Draft v0.1 requires reference execution, canonical config identity, exact calculations where available and seeded simulation. It does not require live routing, deployment or frontend E2E.

Explicitly excluded:

• Final/Freeze/Handoff;
• production or stage publication;
• real-operator flags or wallet use;
• feature buy;
• free spins;
• Mystery/Scatter mechanics;
• collectors, multipliers or additional wagers;
• changes to hold_and_win/v1;
• new art.

After Draft evidence is produced, stop at AWAITING_USER_APPROVAL — DRAFT MATH. Tuning, runtime handoff and any Final version require later explicit approval.