17

Deal each of four players a 13-card hand at random. What is the probability that at least one of the four hands is missing at least one suit?

Let $A_i$ mean that player $i$ is missing at least one suit.

I can compute the probability that player 1 is missing at least one suit using the inclusion-exclusion principle: $P(A_1) = \dfrac{4 {39 \choose 13} - 6 {26 \choose 13} + 4{13 \choose 13}}{52\choose 13}$

The four players are identical so that $P(A_1)=P(A_2)=P(A_3)=P(A_4)$.

Again using the inclusion-exclusion principle, the probability that at least one of the four players is missing at least one suit is $P(\cup_{i=1}^4A_i)=4*P(A_1)-{4 \choose 2}P(A_1\cap A_2) + {4\choose 3} P(A_1\cap A_2 \cap A_3) - P(A_1\cap A_2 \cap A_3 \cap A_4)$

Now we know that $P(A_1 \cap A_2)= P(A_2)P(A_1|A_2)$. Here is where I am stuck. So far as I can tell, $A_1$ and $A_2$ are not independent, and I can't figure out how to calculate the conditional probability $P(A_1|A_2)$. Anyone know how to do it?

RobPratt
  • 45,619
  • 1
    See http://math.stackexchange.com/questions/98257/a-bridge-hand-void-in-one-suit and http://math.stackexchange.com/questions/98671/non-routine-application-of-inclusion-exclusion – Henry Sep 04 '14 at 18:55
  • Thanks @Henry. So this is a similar question, but a little different because we are calculating the probability that at least one of the suits is missing, so I think the $P(A_1)$ that I calculated is correct. I still don't see how to calculate the conditional probability. – Bridge counter Sep 04 '14 at 19:08
  • 1
    After some more thinking, I imagine that I should be making use of the multinomial counting formula. There are ${52 \choose 13,13,13,13}$ ways to distribute the cards to the four players. – Bridge counter Sep 04 '14 at 21:24
  • I know that it is not correct because it is not even between zero and one, but what is wrong with the following solution:

    $1- \dfrac{{13 \choose 1,1,1,1,9}^4 {36\choose 9,9,9,9}}{52 \choose 13,13,13,13}$

    To get this "solution," find the probability that no player is missing a suit. The total number of ways to deal 13 cards each to 4 players is ${52 \choose 13,13,13,13}$. To have no player missing a suit, choose four of the 13 cards from each suit to give to each of the players, then deal 9 cards each to each of the four players. Why precisely is this wrong?

    – Bridge counter Sep 05 '14 at 00:32
  • 1
    http://en.wikipedia.org/wiki/Stirling_numbers_of_the_second_kind

    You are trying to divide 4 suits of 13 objects each into 4 subsets such that each subset gets atleast one. So ${4!^4 \cdot S(13,4)}^4$ should be the number of ways to do it. I don't understand combinatorics much but sometime back I had asked a question which was similar. The answer was given by @André Nicolas. http://math.stackexchange.com/questions/804088/probability-of-getting-k-different-coupons-in-a-set-of-s-coupons-from-an-infini So I am just trying to reapply the same principle here.

    – Durin Sep 05 '14 at 11:09
  • Thanks @Durin. I think that is correct. So the answer is $\dfrac{(4!S(13,4))^4}{{52 \choose 13,13,13,13}}$. I'm still trying to wrap my mind around how the formula for the Sterling numbers of the second kind (which is vaguely inclusion-exclusion-esque) can be interpreted in terms of this problem. – Bridge counter Sep 05 '14 at 16:51
  • This doesn't give a solution, but does include some numerical information: http://www.rpbridge.net/7z79.htm – user84413 Sep 06 '14 at 21:52
  • I attempted the direct approach of identifying each individual ordered $16$-tuplet of spades hearts diamonds clubs to each of Person-1, Person-2, Person-3, Person-4. I was going to then compute the probability of each satisfying $16$-tuplet occurring and then add them together. I abandoned this approach when I discovered that (assuming that I have not made any programming error) there are 2,309,384 such ordered $16$-tuplets. – user2661923 Oct 13 '23 at 23:40
  • According to this page https://en.wikipedia.org/wiki/Contract_bridge_probabilities the probability of one hand getting a void is .0512, so the probability that at least 1 hand has at least one void is 1 minus the probability that no hand has a void = $1 - (1-.0512)^4$ = $1-.9488^4 \approx 1 - .81$. So the probability that at least one hand has at least one void is about 19%. – Nate Oct 21 '23 at 00:29
  • @Nate Close, but the events are not independent. Simulation gives the probability in question to be about 18.4% – Daniel Mathias Oct 21 '23 at 04:07
  • @ Daniel Mathias , @Nate Close As 18.3766327...% and 18.9601359% are very close, the four events are very weakly dependent, which may be an interesting observation. – Amir Nov 10 '23 at 09:37
  • @Daniel Mathias I realized that the void probability 0.0512 given in wikipedia needed to be corrected as 0.051065521. Could you check it? – Amir Nov 10 '23 at 18:13
  • @Amir You are correct. The sum of the given probabilities is 1.0001, indicating the error. – Daniel Mathias Nov 10 '23 at 23:52

1 Answers1

4

The problem seems difficult that has remained unanswered for about 9 years. I could find two solutions to it, while they are both long.

First method:

Let A, B, C, and D denote the event of that the $j\text{th}$ hand, $j=1,2,3,4$, misses at least one suit. Then, using the De Morgan's Law, the probability of event is given by

$$ P(A\cup B\cup C\cup D)=1- P(A'\cap B' \cap C'\cap D')=1-\dfrac{ N }{\binom{52}{13,13,13,13} }$$

where $N$ is the number of cases that no hand misses any suit. $N$ can be obtained by summing all the following terms:

$$\binom{13}{i_1} \binom{13}{i_2}\binom{13}{i_3}\binom{13}{13-i_1-i_2-i_3}\times$$ $$\binom{13-i_1}{j_1}\binom{13-i_2}{j_2}\binom{13-i_3}{j_3}\binom{13-(13-i_1-i_2-i_3)}{13-j_1-j_2-j_3}\times$$ $$\binom{13-i_1-j_1}{l_1}\binom{13-i_2-j_2}{l_1}\binom{13-i_3-j_3}{l_1}$$ $$\binom{13-(13-i_1-i_2-i_3)-(13-j_1-j_2-j_3)}{13-j_1-j_2-j_3}\times$$ $$\binom{13-(13-i_1-i_2-i_3)-(13-j_1-j_2-j_3)-(13-l_1-l_2-l_3)}{13-j_1-j_2-j_3}$$

where the indices $i_1,i_2,_, j_1,j_2, j_3, l_1,l_2,l_3$ vary in the following domain:

$$1\le i_1,i_2,i_3\le10, 13-i_1-i_2-i_3\ge1$$ $$1\le j_1,j_2,j_3\le10, 13-j_1-j_2-j_3\ge1$$ $$1\le l_1,l_2,l_3\le10, 13-l_1-l_2-l_3\ge1$$ $$13-i_1-j_1-l_1\ge1, 13-i_2-j_2-l_2\ge1, 13-i_3-j_3-l_3\ge1, 13-(13-i_1-i_2-i_3)-(13-j_1-j_2-j_3)-(13-l_1-l_2-l_3)\ge1.$$

Using a Python code, I could compute $$N=4.378664133339871e+28.$$ Hence, the probability is $$1- \frac{4.378664133339871e+28}{5.36447377654888e+28}=0.18376632718730682$$

This probability is very close to one obtained by assuming that all the four events A, B, C, and D are independent, that is, $$1-(1-0.051065521)^4=0.189141811,$$ which shows that the events A, B, C, and D are very weakly dependent (note that the void probability given in Wikipedia 1, i.e. 0.0512, is not accurate; see 2 for more details and other related problems).

Second method:

For each $i=1,2,3,4$, let us define $A_i$,$B_i$,$C_i$, and $D_i$ as the event of that the $j\text{th}$, $j=1,2,3,4$,hand has no card of the $i\text{th}$ suit, respectively.

Then, the probability of interest is

$$ P\{(\cup_{i=1}^4A_i) \cup (\cup_{i=1}^4B_i) \cup (\cup_{i=1}^4C_i) \cup (\cup_{i=1}^4D_i)\}. $$

By the inclusion-exclusion principle, the above probability is equal to

$$4\times4P(A_1)$$ $$-4{4 \choose 2}P(A_1\cap A_2) -{4 \choose 2}4 \times 3 P(A_1\cap B_2) - {4 \choose 2}4 P(A_1\cap B_1)$$

$$4{4\choose 3} P(A_1\cap A_2 \cap A_3) + {4\choose 2} {4\choose 2} 2 P(A_1\cap A_2 \cap B_3)+ {4\choose 2} {4\choose 2} 2 P(A_1\cap A_2 \cap B_1)+ {4\choose 3}4\times3\times2 P(A_1\cap B_2 \cap C_3)+{4\choose 3}4\times3 P(A_1\cap B_1 \cap C_2)+{4\choose 3}4 P(A_1\cap B_1 \cap C_1) $$

$$-{4\choose 2}{4\choose 3}P(A_1\cap A_2 \cap A_3\cap B_4)-{4\choose 2}{4\choose 3}3P(A_1\cap A_2 \cap A_3\cap B_1)- {4\choose 2}{4\choose 2}P(A_1\cap A_2 \cap B_3\cap B_4)-{4\choose 2}{4\choose 2}P(A_1\cap A_2 \cap B_1\cap B_2)-{4\choose 2}{4\choose 2}2\times 2P(A_1\cap A_2 \cap B_1\cap B_3)-{4\choose 3}{4\choose 2}2P(A_1\cap A_2 \cap B_1\cap C_1)-{4\choose 3}{4\choose 2}2P(A_1\cap A_2 \cap B_3\cap C_4)-{4\choose 3}{4\choose 2}2P(A_1\cap A_2 \cap B_3\cap C_3)-4!P(A_1\cap B_2 \cap C_3\cap D_4)-{4\choose 2}4\times 3 \times 2 P(A_1\cap B_2 \cap C_2\cap D_4)-4\times 4 \times 3P(A_1\cap B_2 \cap C_2\cap D_2)$$

$$+\dots - 4! P(A_1\cap A_2 \cap A_3\cap B_1\cap B_2 \cap B_4 \cap C_1\cap C_3 \cap C_4 \cap D_2\cap D_3 \cap D_4).$$

The terms including more than 4 events can be generated similarly. Combinations related to different situations need to be separated, while they may finally have the same probability (c.f. $P(A_1\cap B_1)$ and $P(A_1\cap B_2)$ below). Many combinations of 4 or more events can be ignored as their intersections are empty. A useful rule is that in a reasoanable combination each hand or each suit cannot appear more than three times, e.g.,

$$P(A_1\cap A_2 \cap A_3\cap A_4)=P(A_1\cap B_1 \cap C_1\cap D_1)=0.$$

Using this rule, one can see that the maximum number of events whose intersection is not empty is 12.

Each element of the above expansion can be computed considering how the corresponding subset of 16 events defines a subset of suits for each hand ($A_1A_2B_1$ means that the first hand includes cards only from the 3rd and 4th suits, the second includes cards only from the 2nd, 3rd, and 4th suits, and the third and forth hands can have cards from all suits). For example,

$$ P(A_1)= \dfrac{ \binom{39}{13}\binom{39}{13,13,13} }{ \binom{52}{13,13,13,13} } $$

$$P(A_1\cap A_2)=\dfrac{ \binom{26}{13}\binom{39}{13,13,13} }{ \binom{52}{13,13,13,13} }$$

$$P(A_1\cap B_1)=\dfrac{ \binom{39}{13}\binom{26}{13}\binom{26}{13,13} }{ \binom{52}{13,13,13,13} }$$

$$P(A_1\cap B_2)=\dfrac{ \binom{39}{13}\binom{26}{13}\binom{26}{13,13} }{ \binom{52}{13,13,13,13} }$$

$$P(A_1\cap A_2 \cap A_3)=\dfrac{ \binom{13}{13}\binom{39}{13,13,13} }{ \binom{52}{13,13,13,13} }$$

$$P(A_1\cap A_2 \cap B_3)=\dfrac{ \sum_{i=1}^{13}\binom{13}{i}\binom{13}{13-i}\binom{26+i}{13}\binom{26}{13,13} }{ \binom{52}{13,13,13,13} }$$

$$ \dots$$

$$P(A_1\cap A_2 \cap A_3\cap B_1\cap B_2 \cap B_4 \cap C_1\cap C_3 \cap C_4 \cap D_2\cap D_3 \cap D_4)=\dfrac{ 1 }{ \binom{52}{13,13,13,13} }.$$

Amir
  • 4,305