One-shot dominance does not transfer unchanged to repeated tournaments.
Precise claimIn a one-shot prisoner's dilemma, defection strictly dominates. In this fixed eight-strategy round-robin, every strategy that never defects first outscored both defect-first strategies at 10, 50 and 200 rounds per pairing for each of 50 tested seeds; the exact winning strategy still changed with horizon and seed.
Applies
The page computes a fixed pool of Always Cooperate, Always Defect, Tit for Tat, Suspicious Tit for Tat, Grim Trigger, Tit for Two Tats, Pavlov and Random; payoffs T=5, R=3, P=1, S=0; self-play included; no move noise; Mulberry32 reset to seed 20260720 for each visible table. The group claim was audited for seeds 1 through 50 at exactly 10, 50 and 200 rounds.
Does not prove
This is not a universal ranking of strategies. It does not cover other pools, payoff matrices, random generators, self-play rules, noisy moves, unknown or probabilistic horizons, elimination formats or evolutionary dynamics. A known final round also permits an endgame defection. The result concerns scores under stated rules, not whether a person is good or selfish.
Portable rule
If a conclusion comes from a game with repetition, then ask how long the game runs, who else is playing, and what the rules reward before treating the winner as a law.
Declare the discussion context, then copy a stable link. No account or personal data is attached.
repeated-game
Does Being Selfish Actually Win?
Play ten rounds against a copycat, then keep the strategy pool, payoffs and seed fixed while changing only how many rounds each tournament pairing lasts.
Your opponent is Copycat
It cooperates first. Then it copies your previous move.
You know the match lasts exactly ten rounds. Higher total score wins.
Points per round — rows are you, columns are Copycat. The highlighted cell is the temptation.
You ↓Copycat →
It cooperates
It defects
You cooperate
3 · 3both eat
0 · 5you feed it
You defect
5 · 0the temptation
1 · 1both starve
Round1 / 10
Your score0
Copycat0
Copycat opens with cooperation. Your move.
You
Copycat
Read the diagonal: from round 2 on, Copycat's cell is your previous cell.
RoundYouCopycat
The reveal · your ten-round result
The one-round winner can lose the sequence.
In a one-shot prisoner's dilemma, defection strictly dominates. In this fixed eight-strategy round-robin, every strategy that never defects first outscored both defect-first strategies at 10, 50 and 200 rounds per pairing for each of 50 tested seeds; the exact winning strategy still changed with horizon and seed.
Your plan—
Play ten rounds to see your result.
Your score—
Benchmarks appear here.
You—
Always cooperate30
Always defect14
9 cooperate + final defect32
Defection still pays five points in the round where it happens. The missing cost arrives one round later, when Copycat answers it. Only the first or final defection can create an unanswered five-point edge.
A move is not only a payoff. In a repeated game, it is also the next round's message.
The rebuilt model
You assumed:
Defection pays more in every single exchange, so the strategies that defect first should also lead a repeated tournament.
The actual model:
A move in a repeated game changes both today's payoff and the opponent's next response. How much the response matters depends on the horizon, opponent pool and scoring rules. One-shot dominance is real inside its one-shot boundary; tournament rankings are properties of the full setup, not laws attached to a strategy name.
The variable that failed you:
The response horizon — how many future rounds are available for today's move to change what happens next.
Change one variable
How long does the future last?
Eight fixed strategies play every other strategy and themselves. The pool, payoff table, self-play rule and random seed stay fixed. Only rounds per pairing changes.
The result will compare your prediction with the computed table.
What this proves: in this fixed pool, every strategy that never defects first beats both defect-first strategies at the audited 10, 50 and 200-round horizons for seeds 1–50. What it does not prove: that any named strategy always wins, or that cooperation is a universal moral law. Pool, horizon, noise and scoring rules can change the ranking.
For the stated prisoner's-dilemma payoff ordering, defection strictly dominates cooperation in the one-shot game: it pays more whether the other player cooperates or defects.
In laboratory infinitely repeated games, the strategies participants choose vary systematically with the parameters of the game.
paper — checked 2026-07-20.
Scope: The page computes a fixed pool of Always Cooperate, Always Defect, Tit for Tat, Suspicious Tit for Tat, Grim Trigger, Tit for Two Tats, Pavlov and Random; payoffs T=5, R=3, P=1, S=0; self-play included; no move noise; Mulberry32 reset to seed 20260720 for each visible table. The group claim was audited for seeds 1 through 50 at exactly 10, 50 and 200 rounds.
Does not prove: This is not a universal ranking of strategies. It does not cover other pools, payoff matrices, random generators, self-play rules, noisy moves, unknown or probabilistic horizons, elimination formats or evolutionary dynamics. A known final round also permits an endgame defection. The result concerns scores under stated rules, not whether a person is good or selfish.
If a conclusion comes from a game with repetition, then ask how long the game runs, who else is playing, and what the rules reward before treating the winner as a law.
Supplier relationships
Squeezing a long-term vendor may win one invoice while changing the price, service and trust in every delivery that follows.
Pricing decisions
A price cut cannot be valued as a one-round gain if competitors can answer it in the next round and sustain a price war.
Benchmark claims
A headline that says a strategy or tool wins is incomplete until the comparison pool, horizon and scoring rule are visible.