Skip to content
Omnikalc

Parrondo's paradox calculator: can two losing games make a winning one?

Enter the bias, the share of plays of game A and the number of plays to get the expected gain per play of each game and of the random mix.

Updated on
Your inputs
%

Value out of range

%

Value out of range

Value out of range

Results

Expected gain per play of the random mix of A and B

0.0157

Parrondo's paradox: two losing games combine into a winning one
Expected gain per play of game A alone
-0.0100
Expected gain per play of game B alone
-0.0087
Expected capital change after the plays, random mix
15.7
Share of time with capital at a multiple of 3 (mix)
34.5 %

Parrondo’s paradox shows that combining two losing strategies can win. The Parrondo’s paradox calculator takes the bias, the share of plays of game A and the number of plays, solves the Markov chain of the capital modulo 3, and gives the expected gain per play of game A, of game B and of the random mix, the expected change in capital and the share of time spent at multiples of 3.

How the calculation works

  • Enter the bias ε, a small advantage that the house takes in every game. The original paradox uses ε = 0.5 %.
  • Game A wins with probability 1/2 − ε. Game B depends on your capital: if it is a multiple of 3 you win with probability 1/10 − ε, otherwise with 3/4 − ε.
  • Enter the share of plays that use game A when you pick A or B at random for each play, and the number of plays.
  • The calculator solves the long-run behaviour of the capital modulo 3 and gives the expected gain per play of A, B and the mix.

Example

With ε = 0.5 %, game A alone loses 0.0100 per play. Game B alone loses 0.0087 per play, because it spends too much time at multiples of 3 where its chance is only 9.5 %. Mixing them at random, half of the plays each, wins 0.0157 per play. Over 1,000 plays that is about +15.7 in capital. The mix puts you at a multiple of 3 for 34.5 % of the time instead of 38.4 %, which is enough to turn the sign.

Why it matters

The paradox is a standard example in game theory, physics and finance of how interactions change a result that looks obvious. The exact calculation avoids long simulations and shows how sensitive the effect is to the bias and to the mixing proportion.

Good to know

The calculation uses the original version of the games with random mixing. Deterministic patterns, such as alternating two plays of A with two of B, give other results that are not computed here. The expected gain is an average: a single run of plays can end well above or below it.

The Markov chain calculator solves long-run behaviour of a chain, the gambler’s ruin calculator covers the risk of going broke, and the expected value calculator handles single bets.

Frequently asked questions

What is Parrondo's paradox?

It is a result in game theory where two games that each lose in the long run produce a winning game when they are alternated or mixed. It was devised by the physicist Juan Parrondo in 1996 to explain how the ratchet effect can extract directed motion from random fluctuations.

Why does mixing the games help?

Game B is a trap: it has a poor chance of winning when the capital is a multiple of 3 and a good one otherwise, and its own dynamics put you at multiples of 3 too often. Playing game A from time to time shuffles your capital so that you land on multiples of 3 less often, and game B then wins more of the time.

Is this a way to beat the casino?

No. The paradox needs a specific structure: game B depends on the capital, and the games have a specific interaction. Real casino games are independent plays with a fixed edge, and mixing independent losing bets only averages their losses. The effect is also tiny for small advantages.

What do the numbers show?

The expected gain per play of the mix is a weighted result of a Markov chain on the capital modulo 3, solved exactly here. You can change the bias to see that the effect disappears when the bias becomes large, and the share of game A to see that the best mix is not at the extremes.

Was this calculator helpful?

Thanks for your feedback!

Results are estimates provided for information purposes. They do not replace professional advice.