© Hindawi Publishing Corp. PARRONDO’S PARADOX

Since coming to the attention of the general news media several 
years ago, the paradoxical combination of two losing games into a 
winning game by J. M. R. Parrondo has been the subject of 
numerous numerical investigations and simulations. This note 
provides a clear statement of the nature of the paradox together 
with a straightforward analysis and resolution that is accessible 
to a wide audience.


Introduction.
A recently discovered paradox by J. M. R. Parrondo involves combining two losing games into a winning game [1, 3,4,5,6,7].In this note, we set Parrondo's games in a continuum of such games, find an explicit expression for the difference between the losing and winning games, and maximize this difference, thereby eliminating the probabilistic components and explaining the paradox simply in deterministic terms.
2. Parrondo's paradox.Parrondo's paradox involves two games, A and B, each based on tossing biased coins, with the player winning $1 on heads and losing $1 on tails.In game A, the coin (which we call coin 1) has probability 1/2 − c of heads, and so probability 1/2 + c of tails, where c is a small positive number such as 0.005.For c = 0, game A is fair, so c represents the bias against the player.Game B uses two coins: one (coin 2) has probability 1/10−c of heads and the other (coin 3) has probability 3/4−c of heads.The choice of which coin to use is determined by the following rule: use coin 2 if the current winnings are a multiple of 3, and otherwise use coin 3.For c = 0, game B is also fair.To see this, we represent the game as a Markov chain (see Figure 2.1), with the states representing the winnings being 0,1, or 2(mod 3) and the transition probabilities on the arrows giving the probabilities of winning or losing $1 with the appropriate coin, so that the transition matrix is Standard methods (see, e.g., [8, pages 479-480]) allow one to calculate the "stationary distribution" d with dT = d, giving the long-run probabilities of the various states, obtaining d = (5/13, 2/13, 6/13).Multiplying these probabilities by the expected winnings for those states, (−4/5, 1/2, 1/2), gives 0, showing that for c = 0, the game is fair.For c > 0, the game is unfavorable to the player since the probabilities for heads are reduced.
Parrondo's paradox is the observation that if games A and B (with c > 0 to make them unfavorable) are played in certain combinations, the result is favorable in that the long-run expected winnings for each play, or the "value" of the game, are positive.The games can be played in strict alternation (ABABAB . ..), by repeating blocks like AAABB, or by randomly choosing game A with a probability p and B with probability 1−p.For example, if games A and B are played with c = 0.005, their long-run expected winnings per play are, respectively, −1 cent and about −0.87 cent, while if they are combined by choosing each randomly with probability 1/2, the value is about 1.57 cents (found by representing the combined game as another Markov chain and multiplying the new long-run probabilities by the expected gains).Taking instead c = 0 raises the long-run value of the combined game to about 2.54 cents.Other combinations are discussed (along with simulations) in [2] and elsewhere on the world wide web.

Analyzing the paradox by coin selection.
Since two unfavorable games combining into a favorable game is no more paradoxical than two fair games combining into an "even more" favorable game, we take c = 0, both for simplicity and because it increases the difference between the value of the favorable game and the larger of the two values of games A and B. A paradox is a seemingly contradictory occurrence, and Parrondo's paradox may be analyzed as follows.There are three coins: coin 1 is fair (probability of heads is 1/2), coin 2 is unfavorable (probability of heads is 1/10), and coin 3 is favorable to the player (probability of heads is 3/4).Game B uses the unfavorable and favorable coins in proportions that make the game fair, using the coins 0, 5/13 ≈ 38%, and 8/13 ≈ 62% of the time, respectively.In the combined game, however, they are used 1/2 = 50%, 245/1418 ≈ 17%, and 232/709 ≈ 33% of the time, and the favorable outcome should not be unexpected since the unfavorable coin is used only about half as often.
where w(x The value of the combined game is maximized at x = 0, meaning that in game B coins 2 and 3 have probabilities 0 and 1, respectively, for heads.This makes game B deterministic: if your winnings are a multiple of 3 you lose $1, and otherwise you win $1. Game B is still fair since ultimately it oscillates between states of winning or losing $1, as shown in Figure 4  Combining this extreme version of game B with game A leads to a positive expected value because game A, with equal chances of winning and losing $1, will occasionally lead to the middle state above, resulting in the gain of $1 before resuming the oscillation.

Extreme Parrondo.
It should now be clear that we may push matters still further by replacing oscillations with absorbing states and then switching between games so as to always win.Such deterministic games can be disguised behind a screen of probabilistic rules for choosing which game is played next to achieve Parrondo-like effects.
The "ultimate" Parrondo paradox would be two fair games that when played in some combination result in winning $1 on each play.We describe such a situation below, beginning with one where the combined game wins $0.50 on each play, and then a deterministic version with winnings of $1 on each play.The diagram for game B is similar to Figure 5.1 but "odd" leads to "even" with probability 1 and "even" is the absorbing state.Clearly, both games are fair in that after at most one step you are in the absorbing state with winnings of 0. If we now combine games A and B by playing each randomly, with probability 1/2 each, we have the situation shown in Figure 5.2.By symmetry, the long-term invariant distribution for this chain is (1/2, 1/2), and multiplying by the expected payoff vector (1/2, 1/2) gives expected winnings of 1/2.
Even more favorable, simply alternate ABABAB . . ., with no probabilities in the game at all.Ultimately, the chain is as shown in Figure 5.3, in which you win $1 each play.
This can be made into a more Parrondo-like game by introducing probabilities, but at a cost of decreasing the value of the game.For example, consider two games, C and D, with each play of each based on a toss of a coin that comes up heads with probability 9/10.In Game C, if your current winnings are even, you win $1 on heads and lose $1 on tails; while if your current winnings are odd, you win nothing on heads and lose $1 on tails.Modeling this as a Markov chain with states even and odd, the invariant vector is (1/11, 10/11), and multiplying by the payoffs (4/5, −1/10) gives a negative value of −1/55.Game D is the same as game C but reversing even and odd, and therefore has the same negative value.However, playing these in strict alternation CDCDCD . . .and beginning with 0 winnings, each play wins $1 with probability 9/10 and loses $1 with probability 1/10, so the combined game has a value of +4/5.

Call for Papers
Thinking about nonlinearity in engineering areas, up to the 70s, was focused on intentionally built nonlinear parts in order to improve the operational characteristics of a device or system.Keying, saturation, hysteretic phenomena, and dead zones were added to existing devices increasing their behavior diversity and precision.In this context, an intrinsic nonlinearity was treated just as a linear approximation, around equilibrium points.
Inspired on the rediscovering of the richness of nonlinear and chaotic phenomena, engineers started using analytical tools from "Qualitative Theory of Differential Equations," allowing more precise analysis and synthesis, in order to produce new vital products and services.Bifurcation theory, dynamical systems and chaos started to be part of the mandatory set of tools for design engineers.
This proposed special edition of the Mathematical Problems in Engineering aims to provide a picture of the importance of the bifurcation theory, relating it with nonlinear and chaotic dynamics for natural and engineered systems.Ideas of how this dynamics can be captured through precisely tailored real and numerical experiments and understanding by the combination of specific tools that associate dynamical system theory and geometric tools in a very clever, sophisticated, and at the same time simple and unique analytical environment are the subject of this issue, allowing new methods to design high-precision devices and equipment.
Authors should follow the Mathematical Problems in Engineering manuscript format described at http://www .hindawi.com/journals/mpe/.Prospective authors should submit an electronic copy of their complete manuscript through the journal Manuscript Tracking System at http:// mts.hindawi.com/according to the following timetable:

Figure 2
Figure 2.1.Game B as a Markov chain.

4 .
Improving Parrondo.How can we modify game B to maximize the value of the combined game?The probabilities 1/10 and 3/4 of heads for coins 2 and 3 that make game B fair are not unique-they can be replaced by any other probabilities x and (1 − x − √ x − x 2 )/(1 − 2x), such as 1/5 and 2/3, and game B will remain fair.We could also vary the proportions with which games A and B are played, from 50:50 to any probability p for game A and 1−p for game B. The graph in Figure 4.1 shows the values V (x,p) of the combined games for different values of x and p:

Figure 4 . 1 .
Figure 4.1.Values of combined games with probability x of winning with coin 2 and probability p of playing game A (a darker shade means a smaller value).The maximum value occurs at x = 0, p = 1 − 13 − 4 √ 10. .

First
Round of Reviews March 1, 2009