** JOB 0417 ** IPD/TOUREXEC ** THE ITERATED PRISONER'S DILEMMA ** PRE-LAUNCH **
$IPD · a token you have to play
Your money is in the vault. To get it out, you have to trust a stranger.
Every $IPD trade pays a fee into holder vaults. You can see your vault. You can't withdraw it. The only way out is the Table: an iterated prisoner's dilemma against another holder, one chip a round. Each round you both secretly choose to cooperate or steal.
FUNCTION K92R(J,M,K,L,R,JA)
C BY ANATOL RAPOPORT.
C COOPERATE FIRST. AFTER THAT,
C DO WHATEVER THEY DID LAST.
C J = THEIR LAST MOVE (0=C, 1=D)
K92R = J
RETURN
END
Pre-launch. Every figure on this page is simulated.
The rules fit in a box
Four cells.
Each round you and your opponent both put one chip on the table, then choose at the same time. Neither of you sees the other's move until both are sealed.
FIG. 1 Payoff per chip staked, at the full co-op bonus. Big numbers are tournament points. Full payoff table and math in the docs.
In 1980 they submitted programs. Now you play by hand.
The tournament.
In 1980, Robert Axelrod invited game theorists to submit programs to play the iterated dilemma against each other, everyone against everyone. Fourteen entered, plus RANDOM. The winner was the simplest program in the field: Anatol Rapoport's TIT FOR TAT. Axelrod ran it again with 62 entries, and TIT FOR TAT won again.
On the Table there are no presets and no autopilot: you play every move yourself. These ten classic styles are how the Table reads you. Once your record fits one of them, you get its label, and every opponent sees it on your file. Run them against each other below to see which labels you'd want.
Clusters beat individuals
Defectors win the fight. The bonus decides the map.
One on one, a defector always beats a cooperator. But cooperators who find each other outscore everyone around them, and their neighbours copy what works. How far that spreads comes down to one number, the temptation b = T ÷ R. On the Table, the co-op bonus sets it.
A world of strangers, one neighbourhood at a time.
FIG. 2 Spatial prisoner's dilemma, after Nowak & May (1992). Each square plays its eight neighbours, then copies whichever neighbour earned the most. Temptation b = 2 ÷ (1 + bonus); mutual defection pays 0, exactly as on the Table. Drag the bonus to zero and defectors take the map. Drag it up and they shrink to scattered holdouts.
Reputation is on-chain
You will be remembered.
Every holder has a file. Before each match you get your opponent's, and they get yours. Nothing in it is self-reported. It's computed from every move you've ever sealed.
AXELROD, 1984 The traits shared by the programs that scored well. The dossier measures the first three directly; definitions in the docs.
Questions, answered under oath
Cross-examination.
** END OF JOB 0417 ** 6 FORMS PRINTED ** 0 ERRORS **