1
$\begingroup$

I was working on creating a card game with a friend, and we wanted to understand how card draws would play out.

We have a set of 18 distinct cards. Each card participates four times in some pre-defined combination of three cards (order doesn't matter) for a total of 24 combinations. How many cards do we need to draw in our opening hand, to ensure that one of the combinations exists in our opening hand?

We thought this had to do with combinatorics, maybe related to this question, but we get stuck thinking about it, because each of our cards is distinct, so we don't actually have 24 identical cards to choose from, but rather 24 different sets of 3 cards out of 18 unique cards.

$\endgroup$
2
  • $\begingroup$ Sounds a bit like Dobble, but not quite that... presumably though any two combinations of cards share at most $1$ card? $\endgroup$
    – Joffan
    Commented Apr 14, 2021 at 23:01
  • $\begingroup$ no, I think two combinations of cards are not restricted to only sharing at most 1 card. Up to 4 combinations of cards can share up to 2 cards (if those two cards appear in only those 4 combinations, seeing as each card participates exactly four times in any of the combinations). For example, we could have 4 of the combinations be 'abc', 'abd', 'abe', 'abf'. The remaining 20 combinations would then have to be composed of 'c' through 'r' though, since 'a' and 'b' would be done. Note, I don't know ahead of time what the 24 pre-defined combinations actually are. $\endgroup$ Commented Apr 15, 2021 at 0:04

2 Answers 2

0
$\begingroup$

Presumably any possible two 3-card combinations of cards share at most $1$ card. So if you take an initial card, all the combinations involving that card would be another $4\cdot 2 = 8$ cards. So adding the initial card, that's half the pack - is there an "anti-card" to our initial choice for which all of its combinations complete the pack? In which case you could certainly hold $1+4+1+4=10$ cards without a combination & need at least $11$ cards to have a combination.

Coming from the other side, each card can only break $4$ combinations, so you'd need to omit $6$ cards to have any chance of breaking all combinations. That would imply that $13$ cards would always be enough, depending on how the combinations are organized. The requirement for $13$ cards can be firm if a lot of combinations have some common cards, for example with combinations:
$\{abm, abn, abo, abp, $ $cdq, cdr, cdm, cdn, $ $efo, efp, efq, efr, $ $ghm, ghn, gho, ghp, $ $ijq, ijr, ijm, ijn, $ $klo, klp, klq, klr \}$
you could have zero combinations unless at least one of $m,n,o,p,q,r$ are held.

$\endgroup$
3
  • $\begingroup$ If the card combinations are somehow set dynamically, the card/anti-card configuration is still possible, so the answer is unchanged as a lower bound, that is you need at least $10+1=11$ cards. $\endgroup$
    – Joffan
    Commented Apr 15, 2021 at 0:46
  • $\begingroup$ After also reading Mike Earnest's answer, the second part of your answer makes sense to me now, and I think is correct; the 13th card will ensure that we have a combo in our hand. More generally, it sounds like to guarantee that we draw a combo, we need to draw this many cards (does it seem correct?): (number of unique cards) - ((number of combos) / (number of combos a card participates in)) + 1 But sorry I still don't get the first part. Are you saying 10 is the absolute minimum you need to draw to get a combo, like, 10 is the optimistic answer, while 13 is the pessimistic one? $\endgroup$ Commented Apr 20, 2021 at 20:53
  • $\begingroup$ I guess the first part was my initial thoughts, if the combinations were designed such that any two combinations share at most one card. Then the general answer is the second part. $\endgroup$
    – Joffan
    Commented Apr 20, 2021 at 20:57
0
$\begingroup$

The answer depends on what the particular $24$ combinations are. Here are two examples where the required number of cards in the deck are different. In order to answer your question, you would have to enumerate all possible ways to choose $24$ combinations, find the minimum number of cards which contains at least one combination for each, and take the maximum of these minima. I do not know how to do this.

  • Suppose the cards are labeled $A_1,A_2,\dots,A_6$, $B_1,B_2,\dots,B_6$, and $C_1,C_2,\dots,C_6$. The following $8$ subsets contain each of the $A$ cards exactly four times: $$ \{A_1,A_2,A_3\},\{A_4,A_5,A_6\}\\\{A_1,A_2,A_4\},\{A_3,A_5,A_6\}\\\{A_1,A_2,A_5\},\{A_3,A_4,A_6\}\\\{A_1,A_2,A_6\},\{A_3,A_4,A_5\}\\\ $$ If you repeat that construction with $A$ replaced with $B$ and $C$, you get $24$ subsets total which contain all cards exactly four times. For this choice of subsets, you need $\boxed{10}$ cards to guarantee having one of the $24$ combinations. Indeed, if you have $10$ cards, then there must exist four cards which have the same letter (either all $A$, all $B$, or all $C$), and you can directly check that four of $A_1,A_2,\dots,A_6$ will contain one of the above subsets, while any three will not suffice (e.g. $\{A_1,A_3,A_4\}$).

  • For the other example, we first construct $9$ cards and $12$ subsets of size $3$ which contain these cards four times each. The nine cards are the squares of a tic-tac-toe board, while the $12$ subsets consist of

    • The three rows.
    • The three columns.
    • The three downward sloping broken diagonals.
    • The three upward sloping broken diagonals.

    You can check that any five cards will contain one of these 12 subsets, but there exist subsets of size $4$ which do not, like the set of $4$ corners.

    Now, if you use two copies of this deck, you get a deck of $18$ cards and $24$ combinations of these cards containing each card four times. In order to guarantee getting one of these subsets in a set of cards, you need at least $\boxed 9$ cards in that set, while $8$ does not suffice.

As you see, the answer can be either $9$ or $10$. It is possible there are other answers.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .