Commit–reveal

Commit–reveal is the first-class fairness primitive for hidden values such as dice hands. Related patterns for boards (commit once, open only some cells) use Merkle roots and proofs — e.g. Battleship-style ship placement.

Scheme

commitment = SHA-256(message ‖ nonce)

• commit(message) samples a 32-byte nonce and returns hex commitment, nonce, and message.
• verify(commitment, message, nonce) recomputes the digest and compares.

Dice helpers treat faces as raw bytes:

• Single face: one byte ([face])
• Hand: PackedByteArray of faces (e.g. five dice for Liar’s Dice–shaped play)

Protocol shape

1. Commit — each player samples a nonce, commits to hidden values (faces ‖ nonce), and publishes only commitment_hex.
2. Play — game rules use commitments as binding claims (bids, challenges, etc.).
3. Open — players reveal message + nonce; peers call verify_* and reject bad openings.

GDScript example

var api = ClassDB.instantiate("FairPlayApi")
var faces := PackedByteArray([1, 2, 3, 4, 5])
var c: Dictionary = api.commit_dice_hand(faces)

# Wire: public only
var public_msg := {
    "type": "commitment",
    "label": "dice_hand",
    "commitment_hex": c["commitment_hex"],
}

# Later open
var ok: bool = api.verify_commitment(
    c["commitment_hex"],
    faces,
    c["nonce_hex"]
)

Wrong faces or wrong nonce → verification fails (bait-and-switch is detectable).

What it guarantees

Property	Yes?
Binding after publish	Yes — different open fails
Detect wrong open	Yes
Force fair random dice by itself	No — combine with shared RNG / mental poker if needed
Hide values forever	No — open reveals the message

Full honesty bounds: Threat model.