Problem 21.1 Easy Free 8 variations

Reroll a Die Once

A fair six-sided die is rolled once. You may either accept the payout equal to the value shown, or discard it and roll the die a second time, in which case you are paid the value of the second roll instead. Assuming you play optimally, find the fair value (expected payout) of this game.

Work it out, then compare with the answer.

SolutionShow the worked solution

After the first roll, the decision is final. Rerolling gives the value of a fresh roll, whose expectation is

E[X2]=1+2+⋯+66=72.\mathbb{E}[X_2]=\frac{1+2+\dots+6}{6}=\frac72 .

Seeing the first roll X1=xX_1=x, you therefore keep it if x>72x>\tfrac72 and reroll if x<72x<\tfrac72. So you keep 4,5,64,5,6 and reroll 1,2,31,2,3. The value of the game is

E ⁣[max⁡ ⁣(X1,72)]=16(3⋅72+4+5+6)=16⋅512=174.\mathbb{E}\!\left[\max\!\left(X_1,\tfrac72\right)\right] =\frac16\left(3\cdot\frac72+4+5+6\right) =\frac16\cdot\frac{51}{2}=\frac{17}{4}.
174=4.25\boxed{\dfrac{17}{4}=4.25}

Key idea. With one decision left, the value is E[max⁡(current payout, continuation value)]\mathbb{E}[\max(\text{current payout},\ \text{continuation value})].

Variations

The same idea with a twist. Each solution gives only the changed step.

Variation 21.1a

A fair 88-sided die is rolled once. You may keep the payout equal to the value shown, or instead roll a fair 1212-sided die once and be paid its value. Find your expected payout under optimal play.

Work it out, then compare with the answer.

Solution

The continuation value is now the mean of a 1212-sided die, 132\frac{13}{2}. Keep the 88-sided roll only if it shows 77 or 88; otherwise roll the 1212-sided die. The value is

7+88+68⋅132=158+398=274.\frac{7+8}{8}+\frac68\cdot\frac{13}{2}=\frac{15}{8}+\frac{39}{8}=\frac{27}{4}.
274=6.75\boxed{\dfrac{27}{4}=6.75}
Variation 21.1b

A fair 66-sided die is rolled. If the face shown is odd, the payout equals the face value; if it is even, the payout equals twice the face value. You may keep this payout or reroll the die once and accept the payout (under the same odd/even rule) from the new roll. Find the expected payoff under optimal play.

Work it out, then compare with the answer.

Solution

The possible payouts are 1,4,3,8,5,121,4,3,8,5,12. Their mean is 336=112\frac{33}{6}=\frac{11}{2}, which is the continuation value. Keep only the payouts 88 and 1212 (faces 44 and 66) and reroll otherwise:

8+126+46⋅112=206+226=7.\frac{8+12}{6}+\frac46\cdot\frac{11}{2}=\frac{20}{6}+\frac{22}{6}=7.
7\boxed{7}
Variation 21.1c

A number is drawn uniformly at random from (0,1)(0,1). You may keep this number as your payout, or discard it and draw a new uniform number from (0,1)(0,1) once, in which case you are paid that new number. Find your expected payout under optimal play.

Work it out, then compare with the answer.

Solution

A fresh draw is worth 12\frac12 on average, so keep U1U_1 if and only if U1>12U_1>\frac12. The value is

E ⁣[max⁡ ⁣(U1,12)]=∫01/212 du+∫1/21u du=14+38=58.\mathbb{E}\!\left[\max\!\left(U_1,\tfrac12\right)\right]=\int_0^{1/2}\tfrac12\,du+\int_{1/2}^1 u\,du=\frac14+\frac38=\frac58.
58\boxed{\dfrac58}
Variation 21.1d

Alice rolls a fair 66-sided die and observes the value shown. She may stop and receive that value as her payout, or pay $1\$1 to roll the die a second time and receive the value of that second roll instead. Assuming Alice plays optimally, find her expected payout.

Work it out, then compare with the answer.

Solution

Rolling again is worth 72−1=52\frac72-1=\frac52 net of the cost. Alice therefore keeps 3,4,5,63,4,5,6 and rerolls on 11 or 22. Her expected net payout is

3+4+5+66+26⋅52=186+56=236.\frac{3+4+5+6}{6}+\frac26\cdot\frac52=\frac{18}{6}+\frac56=\frac{23}{6}.
236≈3.83\boxed{\dfrac{23}{6}\approx 3.83}
Variation 21.1e

Ten cards numbered 11 through 1010 lie face down. You draw one card at random and look at its value. You may take a fixed payout of $3.50\$3.50, or instead take the face value of the card. Find the fair value of this game under optimal play.

Work it out, then compare with the answer.

Solution

Take the face value when it exceeds 3.53.5, which happens for cards 44 to 1010. Take the fixed $3.50\$3.50 for cards 11 to 33. The value is

310⋅72+4+5+⋯+1010=2120+4910=11920.\frac{3}{10}\cdot\frac72+\frac{4+5+\dots+10}{10}=\frac{21}{20}+\frac{49}{10}=\frac{119}{20}.
$5.95\boxed{\$5.95}
Variation 21.1f

Two fair 66-sided dice are rolled, and you observe the product of the two values. You may keep this product as your payout, or pay $4\$4 to reroll both dice once and take the product of the new roll instead. Find your expected payout under optimal play.

Work it out, then compare with the answer.

Solution

By independence, E[XY]=(72)2=494\mathbb{E}[XY]=\left(\frac72\right)^2=\frac{49}{4}, so rerolling is worth 494−4=334=8.25\frac{49}{4}-4=\frac{33}{4}=8.25. You therefore reroll exactly when the product is at most 88.

Count the pairs with product at most 88 by fixing the first die:

  • first die 11: 66 pairs, with products summing to 2121;
  • first die 22: 44 pairs, summing to 2020;
  • first die 33: 22 pairs, summing to 99;
  • first die 44: 22 pairs, summing to 1212;
  • first die 55: 11 pair, summing to 55;
  • first die 66: 11 pair, summing to 66.

That gives 1616 pairs whose products sum to 7373. Over all 3636 pairs the products sum to 212=44121^2=441, so the kept products sum to 441−73=368441-73=368. The value is

36836+1636⋅334=368+13236=1259.\frac{368}{36}+\frac{16}{36}\cdot\frac{33}{4}=\frac{368+132}{36}=\frac{125}{9}.
1259≈13.89\boxed{\dfrac{125}{9}\approx 13.89}
Variation 21.1g

Four fair coins are flipped, and you are paid $1\$1 for each head shown. You also have the option, before being paid, to pay $1\$1 to reflip all four coins once and be paid based on the new outcome instead. Suppose you follow the rule of rerolling whenever the first flip shows 00 or 11 heads. Find the fair value of this game under that rule.

Work it out, then compare with the answer.

Solution

Let H∼Bin(4,12)H\sim\mathrm{Bin}(4,\tfrac12) be the number of heads. Then P(H=k)=(4k)/16\mathbb{P}(H=k)=\binom4k/16.

  • On H≤1H\le1 (probability 516\frac5{16}) you reflip, worth 2−1=12-1=1 on average.
  • Otherwise you keep HH, which contributes 2⋅6+3⋅4+4⋅116=2816\frac{2\cdot6+3\cdot4+4\cdot1}{16}=\frac{28}{16}.

The fair value is

2816+516⋅1=3316.\frac{28}{16}+\frac{5}{16}\cdot1=\frac{33}{16}.
3316≈2.06\boxed{\dfrac{33}{16}\approx 2.06}
Variation 21.1h

Eight fair coins are flipped simultaneously. You may then choose, independently for each coin, whether to reflip it once more. After all chosen reflips, you are paid $1\$1 for every coin showing heads. Assuming optimal play, find the fair value of this game.

Work it out, then compare with the answer.

Solution

The payout is a sum over coins, and each reflip is independent of everything else. So it is optimal to decide for each coin separately.

  • A head is worth 11 if kept and 12\frac12 if reflipped, so keep it.
  • A tail is worth 00 if kept and 12\frac12 if reflipped, so reflip it.

Each coin then ends on heads unless it shows tails twice, so it ends on heads with probability 1−14=341-\frac14=\frac34. By linearity the value is

8⋅34=6.8\cdot\frac34=6.
6\boxed{6}