Free Parrondo Paradox Calculator

The Core Idea of Parrondo's Paradox

The Parrondo Paradox Calculator is a free online tool that demonstrates a surprising result from game theory: two games, each guaranteed to lose money over many rounds, can be combined in a specific way to produce a net profit. This counterintuitive phenomenon was discovered by Spanish physicist Juan Parrondo while studying the Brownian ratchet — a thought experiment from statistical physics.

Using this calculator, you can run a Monte‑Carlo simulation based on biased coin tosses. You are free to choose different sequences (or a random order) and instantly see how the capital evolves. The tool also provides a Markov‑chain analysis that explains mathematically why the paradox is not a violation of probability theory but rather a consequence of the way the two games are coupled.

A crucial caveat: this effect does not occur in casino games. It requires a specific dependency between the games — in the coin‑tossing model the dependency is through the player’s current capital.

The Two Losing Games

Imagine you are given two games, A and B. In each round you bet $1; a win adds$1 to your capital, a loss subtracts $1.

• Game A uses an almost fair coin. The win probability is $P(A)=0.495$, just below 0.5. Over the long run the expected value is negative ($-0.01$$ per game), so you are guaranteed to lose money if you only play A.

• Game B is more complex and depends on your current capital.

  • If the capital is an exact multiple of 3, you must play variant B1, which has a very low win probability $P(B_1)=0.095$.
  • If the capital is not a multiple of 3, you play variant B2, which has a high win probability $P(B_2)=0.745$.

  Even though B2 is strongly favorable, the overall expected value of game B alone is still negative, because a nontrivial fraction of the time you are forced into the unfavorable B1. Hence both A and B are losing games when played in isolation.

Winning by Alternating

Now consider what happens when you switch between A and B instead of sticking to one. The calculator simulates thousands of independent trials, each consisting of 100 rounds. The results are striking:

• Game A alone: capital steadily declines with the number of rounds.
• Game B alone: the same downward trend appears — a sure loss.
• Sequence AABB (two rounds of A, then two of B, repeating): capital increases as the rounds progress.
• Random alternation (a fair coin decides each round whether to play A or B): capital also increases on average.

Both the deterministic AABB pattern and the random mixture turn two losing games into a winning strategy. The reason is that the alternation changes how often the player’s capital falls into each of the three residue classes modulo 3, making the favorable variant B2 appear much more often than it would under pure‑B play.

A Deeper Look with Markov Chains

To understand why mixing changes the outcome, model the system as a Markov chain. The state is the remainder when the current capital is divided by 3 — that is, 0, 1, or 2. The transition probabilities are fully determined by the rules of the games.

Game B Alone

When only game B is played, the one‑step transition matrix (columns correspond to the current state, rows to the next state) is:

$$P_B = \begin{bmatrix} 0 & 0.255 & 0.745 \\ 0.095 & 0 & 0.255 \\ 0.905 & 0.745 & 0 \end{bmatrix}$$

From state 0 (capital a multiple of 3) the player plays B1: winning (probability 0.095) leads to state 1, losing (0.905) leads to state 2. From state 1 or 2 the player plays B2: winning (0.745) moves to the next higher state (2 or 0), losing (0.255) moves to the next lower state.

The stationary distribution $\pi = (\pi_0,\pi_1,\pi_2)^T$ satisfies $\pi = P_B\,\pi$. Solving gives

$$\pi_0 \approx 0.384,\qquad \pi_1 \approx 0.308,\qquad \pi_2 \approx 0.308.$$

Thus about 38.4 % of the time the capital is a multiple of 3, forcing the disadvantageous B1. The overall probability of winning a single round of game B is

$$P(\text{win}\mid B) = 0.384\times0.095 + 0.308\times0.745 + 0.308\times0.745 \approx 0.4955,$$

which is below 0.5, confirming that game B alone is a losing game. (The Perron–Frobenius theorem guarantees that such a stochastic matrix converges to a unique steady state in the long run.)

Mixed Strategy (Random Alternation)

Now consider a strategy where each round a fair coin decides with probability 0.5 whether to play game A or game B. The resulting transition probabilities combine the two games:

• From state 0: win probability = $0.5\times0.495 + 0.5\times0.095 = 0.295$; a win moves to state 1, a loss (0.705) to state 2.
• From state 1: win probability = $0.5\times0.495 + 0.5\times0.745 = 0.62$; win → state 2, lose → state 0.
• From state 2: win probability = 0.62; win → state 0, lose → state 1.

Hence the transition matrix is:

$$P_{\text{mix}} = \begin{bmatrix} 0 & 0.38 & 0.62 \\ 0.295 & 0 & 0.38 \\ 0.705 & 0.62 & 0 \end{bmatrix}.$$

The stationary distribution becomes approximately

$$\pi_0 \approx 0.345,\quad \pi_1 \approx 0.254,\quad \pi_2 \approx 0.401.$$

The overall win probability per round is then

$$0.345\times0.295 + 0.254\times0.62 + 0.401\times0.62 \approx 0.508 > 0.5.$$

Even a random mixture of the two losing games therefore yields a positive expected gain in the long run — this is the heart of Parrondo’s paradox.

Generalising the Paradox

The same phenomenon can be reproduced with other numeric values. If we introduce a small parameter $\varepsilon$ and set

$$P(A) = 0.5 - \varepsilon,\qquad P(B_1) = 0.1 - \varepsilon,\qquad P(B_2) = 0.75 - \varepsilon,$$

then for sufficiently small $\varepsilon$ the paradox persists. This generality shows that the effect is robust and not an artefact of the particular numbers chosen.

Parrondo’s paradox has implications beyond coin‑tossing; similar coupling mechanisms appear in evolutionary biology, financial market modelling, and the study of nanoscale motors. Nevertheless, it does not apply to conventional casino games because those games are independent of the player’s capital history.