** IPD/TOUREXEC REFERENCE MANUAL ** PRE-LAUNCH · ALL VALUES PROPOSED **
Documentation
The manual.
Everything the front page leaves out: parameters, formulas, the protocol and the risks. Nothing is deployed yet, so every value here is a proposal and may change before launch.
Overview
| Part | What it does |
|---|---|
| Vault | Collects your share of trading fees. Visible at all times, locked at all times. It can only be staked at the Table. |
| Wallet | Where Table payouts land. Money in your wallet is claimed and spendable. |
| The Table | An iterated prisoner's dilemma against a random holder. One chip per round, unknown number of rounds. |
| Chip | The stake per round: 1% of the smaller of the two vaults, fixed for the whole match. |
| The Commons | A shared pool. Filled by 20% of fees and by every mutual steal; pays the co-op bonus. |
| Dossier | Your public record, computed from every move you've sealed. Shown to each opponent before the match. |
| Style | A label you earn, not a setting you pick. When your record fits one of the ten classic strategies, you get its name. Every move is played by hand; there is no autopilot. |
Fee engine & vaults
| Parameter | Value | Notes |
|---|---|---|
| Supply | 1,000,000,000 $IPD | Fixed. |
| Trade fee | 4.25% | On every buy and sell, swapped to the base asset (SOL by default). |
| Fee split | 80 / 20 | 80% to holder vaults, 20% to the Commons. |
| Accrual | Per block | Pro-rata to $IPD balance, weighted by seasoning. |
| Seasoning | 72 h | A new balance ramps linearly from 0 to full accrual. The unseasoned share goes to the Commons. |
| Forfeit | Pro-rata | Sell or transfer X% of your $IPD and X% of your unclaimed vault goes to the Commons. |
| Withdrawal | None | Vault funds leave only as chips staked at the Table. |
Vault accrual
For holder i with balance bi, total supply held B, and block fees F:
The loop, as statements
IF (DRAW .LT. .90): 90% of the time there's another round, 10% the match ends.Parameters, as punched
The match
| Step | Rule |
|---|---|
| Queue | Paired uniformly at random among queued holders whose vault is within 2× of yours. You can't choose or see your opponent before pairing. |
| Chip | c = 0.01 · min(VA, VB), measured at pairing and fixed for the whole match. |
| Round | Both players stake c from their vaults and choose COOPERATE or DEFECT (steal) at the same time, by commit–reveal (D9). |
| Settle | Payouts per D4 go straight to wallets. Stakes leave vaults permanently. |
| Continue | With probability w = 0.90 another round follows; otherwise the match ends. Hard cap: 50 rounds (D5). |
| Move clock | 30 seconds per move. If it runs out you forfeit: scored ROBBED (your chip goes to your opponent) and the match ends. There is no autopilot. |
| Vault floor | If a vault can't cover the next chip, the match ends after the current round. |
Payoffs
Each player stakes one chip per round. β is the co-op bonus (D6), between 0 and 0.5 chip.
| Outcome | You | Them | You receive | They receive | Commons | Points (β = .5) |
|---|---|---|---|---|---|---|
| CO-OP (R) | C | C | 1 + β | 1 + β | − 2β | 3 |
| STEAL (T) | D | C | 2 | 0 | 0 | 4 |
| ROBBED (S) | C | D | 0 | 2 | 0 | 0 |
| STANDOFF (P) | D | D | 0 | 0 | + 2 | 0 |
In a single round, stealing never pays less than trusting (4 vs 3 if they cooperate; 0 vs 0 if they steal), so defection weakly dominates. But mutual cooperation is the only outcome that creates value (2R = 6 > T + S = 4), which makes this a social dilemma.
Because P = S, this is the weak form of the prisoner's dilemma. Getting robbed and a standoff pay you the same: nothing. The temptation to steal is pure greed, not self-defense, and mutual defection is as harsh as the game can make it.
Compared with 1980
| Game | T | R | P | S | Why |
|---|---|---|---|---|---|
| Axelrod (1980) | 5 | 3 | 1 | 0 | Classic tournament values. |
| $IPD | 4 | 3 | 0 | 0 | A steal is a straight transfer (no subsidy to thieves); a standoff loses both chips. |
Match length
If you knew round 10 was the last, you'd steal on round 10. So would they. Knowing that, you'd both steal on round 9, then 8, all the way back to round 1. Game theorists call it unraveling.
So the Table never tells you. After every round, a draw seeded by both players' sealed salts decides whether there's another: 90% yes. The draw is memoryless. Round 1 and round 30 look exactly the same from the inside.
Matches are capped at 50 rounds (P ≈ .006). Axelrod's second tournament used the same device with w = .99654, an expected 200 moves.
The Commons
The Commons makes cooperation positive-sum. It fills from fees and from failure, and drains only through the co-op bonus. Cooperation is what drains it, so the bonus is paid out more slowly than the Commons fills: at most 65% of what came in, with the rest kept as a reserve. Once the reserve passes three weeks of inflow, a slice of the excess is released too.
| Entry | Source | Flow |
|---|---|---|
| CR | 20% of every trade fee | In |
| CR | Standoff: both chips | In |
| CR | Sell / transfer forfeits | In |
| CR | Unseasoned accrual | In |
| DR | Co-op bonus, ≤ 0.5 chip per player | Out |
| — | Daily budget: 65% of avg. inflow + 10% of reserve above 21 days | Limit |
The bonus budget
Cd is the Commons balance at the start of day d; Bd is that day's bonus budget; Kd−1 is the total staked in CO-OP rounds the day before (both players' chips). If CO-OP volume runs ahead of forecast, the rest of the day's bonuses are scaled down so total payout never exceeds Bd. When bonuses are the only outflow, the reserve settles near 24 days of inflow. The Data page (§4.7) compares this rule with the alternatives.
The bonus creates a population cycle. Widespread stealing fills the Commons and raises β, which pays cooperators more. Widespread cooperation spreads the budget thin and lowers β, which makes stealing more tempting.
Dossier metrics
For a player with moves mt and opponent moves ot, pooled over all their matches:
| Field | Definition |
|---|---|
| COOP RATE | Share of all moves that were C. |
| NICE | Share of matches in which they never defected first. Defecting in the same round as the opponent's first defection counts as first. |
| RETALIATES | Pr(mt = D | ot−1 = D). How often a steal is answered with a steal. |
| FORGIVES | Pr(mt = C | ot−2 = D, ot−1 = C). How often they return to cooperating after the opponent does. |
| STYLE | The label of the classic style (D8) your record fits. It's the nearest style in Euclidean distance over the four numbers above, each reference computed from 120 simulated matches. Awarded once you have at least 20 moves on file and sit within 30 points of a style; otherwise NO CLEAR STYLE. FIT = 100 − distance. |
Styles
Players can't pick one of these; they're the patterns the Table recognises in your play (D7, STYLE). They also drive the tournament and the simulated opponents. Notation: mt is the style's move in round t, ot the opponent's, u a fresh uniform random number.
Protocol & settlement
STEP WHO ACTION DETAIL WINDOW
1 BOTH COMMIT h = KECCAK(MOVE ‖ SALT) 30 S
2 BOTH REVEAL (MOVE, SALT); CHECK AGAINST h 30 S
3 TABLE SETTLE PAYOUTS PER D4; STAKES LEAVE VAULTS
4 TABLE DRAW KECCAK(SALT_A ‖ SALT_B ‖ t) MOD 100
.LT. 90 → GO TO 1 ELSE END MATCH
| Case | Handling |
|---|---|
| No commit | Treated like a missed reveal: the player is scored ROBBED and the match ends. Nothing plays on a player's behalf. |
| No reveal | The non-revealer is scored ROBBED (their chip goes to the opponent) and the match ends. |
| Randomness | The continuation draw mixes both players' salts, so neither can bias it after committing. The last revealer can see the outcome before revealing, but aborting costs them the chip. |
| Settlement | Moves are signed off-chain (EIP-712 or ed25519). The full transcript settles on-chain in one transaction carrying both signatures, followed by a dispute window. |
| Accounting | Vault balances use a magnified dividend-per-share tracker, so per-holder accrual is O(1) to update. |
Risks & limits
| Risk | Description | Mitigation |
|---|---|---|
| Sybil collusion | One person runs two wallets that always cooperate with each other to farm the co-op bonus. | Random global pairing (no choosing opponents), seasoning, a minimum holding age to queue, and the daily bonus cap. Private duels, if added, pay no bonus. |
| Last-revealer abort | The second player to reveal can see the continuation draw first. | Aborting is scored ROBBED and ends the match, so it always costs the chip. |
| Commons drain | Every cooperative round is paid from the Commons, so sustained cooperation, or a slump in trading, could outrun it. | Bonuses are paid from at most 65% of recent inflow; the rest builds a reserve that is drawn down only above three weeks of inflow (D6.1–D6.3). |
| Vault size gaps | Large and small vaults risking very different amounts. | Pairing within 2×, and the chip is set by the smaller vault. |
| Losses | You can lose staked vault funds. Outcomes depend on other players. | Expected exposure per match is about 10% of the smaller vault (10 rounds × 1%). |
Pre-launch: no contract is deployed. The vault page is a browser-only demo with simulated opponents.
Glossary
| β (beta) | The co-op bonus: extra chips per player per CO-OP round, paid by the Commons. Between 0 and 0.5. |
| Chip | The per-round stake: 1% of the smaller vault in the match. |
| Defect | Steal. The D move. |
| ρ (rho) | Payout ratio: the share of the Commons' recent inflow that may be paid out as bonus each day. 0.65. |
| Iterated | Played repeatedly against the same opponent, so today's move shapes tomorrow's. |
| Nice | Never the first to defect (Axelrod, 1984). |
| Seasoning | The 72-hour ramp before a new balance earns full vault accrual. |
| Shadow of the future | Axelrod's term for how much future rounds matter. Here, w = 0.90. |
| Standoff | Both players steal. Both chips go to the Commons. |
| w | Continuation probability: the chance another round follows the current one. |