How this instrument works
Parrondo's paradox, first described by physicist Juan Parrondo and formally analyzed by Harmer and Abbott in 1999, shows it's possible to construct two games that each lose money in expectation when played on their own, yet produce a winning game when mixed together. This calculator reproduces the classic version: Game A is a simple biased coin flip with win probability pA, while Game B's win probability depends on your current capital — pB1 when capital is a multiple of 3, and pB2 otherwise.
The mechanism behind the paradox is that Game B's win probability isn't fixed — it depends on capital, which the combined strategy changes in a way that shifts how often you're playing Game B's better state versus its worse state. Playing Game A occasionally resets which capital-mod-3 state you're in, which can push the combined strategy's time spent in Game B's favorable state up relative to what Game B alone would experience, tipping an overall-losing setup into a net winner.
This calculator solves for each game's expected profit per round using closed-form Markov chain analysis — finding each game's long-run stationary distribution over capital-mod-3 states via linear algebra, not by simulating thousands of rounds — so the results are exact given the input probabilities, not subject to simulation noise. With the classic textbook parameters (pA = 0.495, pB1 = 0.095, pB2 = 0.745), both games lose alone but the 50/50-randomized combination wins, reproducing the actual published result.
- Enter Game A's win probability into pA — the classic losing setup uses a value just under 0.5.
- Enter Game B's win probability when capital is a multiple of 3 into pB1 — typically the much worse of Game B's two states.
- Enter Game B's win probability otherwise into pB2 — typically the much better state.
- Read profitA and profitBAlone — each game's expected profit per round played on its own.
- Read profitCombined — the expected profit per round when Games A and B are mixed 50/50; compare its sign against the two solo results.
Worked example — the classic Harmer & Abbott parameters
Use the textbook parameters from Harmer and Abbott's original analysis: pA = 0.495, pB1 = 0.095, pB2 = 0.745. Enter those three values.
Game A alone: profitA = 2 × 0.495 − 1 = -0.01 exactly — a losing game, since pA is just under 0.5. Game B alone: profitBAlone ≈ -0.008695, also a losing game once its own capital-mod-3 stationary distribution is accounted for. Combined, mixing the two games 50/50: profitCombined ≈ +0.015704 — a winning game, even though both components lose on their own. This is the actual published paradox, not a rounding artifact: two losing games, combined, produce a net gain.
Questions
Is this real math, or a trick with the numbers?
It's genuine, peer-reviewed mathematics — Parrondo's paradox was formally analyzed by Harmer and Abbott and published in Nature in 1999, and it's a well-studied result in game theory and Markov chain analysis. The 'paradox' label reflects that the outcome is counterintuitive, not that anything is invalid: because Game B's win probability depends on capital, and the combined strategy changes how often you're in each capital state, the two games interact in a way a naive intuition, losing plus losing should still lose, doesn't predict.
Does this calculator use simulation, or exact math?
Exact math — each game's long-run expected profit is computed by solving for its stationary distribution over the three capital-mod-3 states using linear algebra (Gaussian elimination), not by simulating rounds and averaging results. The results here are deterministic and reproducible for a given set of input probabilities, unlike a Monte Carlo simulation, which would introduce sampling noise.
Does the paradox always appear, for any input probabilities?
No — the paradox's direction depends heavily on the specific parameters entered. With some inputs, the combined game can still lose, or the individual games might not both be losing to begin with. That's exactly why this calculator computes the result rather than asserting a universal rule; try the classic textbook parameters first to see the paradox in its most famous published form, then experiment with other values to see how sensitive the effect is to the specific probabilities.
What's the real-world relevance of Parrondo's paradox?
It's primarily of theoretical and pedagogical interest — a striking illustration that combining two unfavorable processes doesn't always produce an unfavorable result, which has drawn interest in fields well beyond gambling, including physics, biology, and control theory. Despite occasional popular claims, it does not translate into a reliable real-world trading or investment strategy; the specific capital-dependent structure needed rarely exists in real financial markets in a form that could actually be exploited.