In a fish shop with 28 kinds of fish, there are 28 fish sellers. In every seller, there exists only one type of each fish kind, depending on where it comes, Mediterranean or Black Sea. Each of the $k$ people gets exactly one fish from each seller and exactly one fish of each kind. For any two people, there exists a fish kind which they have different types of it (one Mediterranean, one Black Sea). What is the maximum possible number of $k$?
Problem
Source: Turkey 2021 IMO TST Problem 4
Tags: Turkey, combinatorics
electrovector
26.05.2021 14:48
Let's number fish sellers with $1, 2 \dots 28$ and fish kinds with $1, 2 \dots 28$. Number the columns and rows of a $28 \times 28$ chart with the numbers of fish sellers and fish kinds, respectively. Let the intersection square of the $i$th row and $j$th column be $(i, j)$ and if we put $B$ into squares $(1, 1), (2, 2) \dots (28, 28)$ and $M$ into the rest, we will end up with the answer $2^{28}-28$ by using induction.
gnoka
29.06.2021 17:47
here is a video solution in Chinese Language
hakN
10.08.2021 17:33
Note that the problem is equivalent to this:
Rephrased and Generalized Problem wrote:
We are given an $n \times n$ board with every cell being $0$ or $1$. We choose exactly $n$ cells such that all the chosen cells are in different rows and different columns. Then, we create a binary string of length $n$ as follows: For each row, we append the value of the cell that is chosen from that row into our string. For every choice of these $n$ cells, what is the maximum number of different binary string that we can create?
Here, $28$ is replaced by $n$ and $k$ is the number of different binary strings that we created.
We claim that the answer is $2^n - n$. Let $s_i$ be the complement string of column $i$. (See the figure) It is easy to see that we cannot create the strings $s_1 , s_2 , \dots , s_n$. Thus, $k \leq 2^n - n$.
Also, post #2 already gave a construction and that construction is easy to prove by induction as #2 said. Thus we are done.
Yeah @Tintarn, you're right, thanks for correction.
Attachments:

Tintarn
31.12.2023 14:51
hakN wrote: Let $s_i$ be the complement string of column $i$. (See the figure) It is easy to see that we cannot create the strings $s_1 , s_2 , \dots , s_n$. Thus, $k \leq 2^n - n$. This is not quite true: If some of the strings $s_i$ are equal, you do not necessarily get $k \le 2^n-n$ from here. However, it is easy to see that the final conclusion still holds: If two $s_i$ are equal, we cannot create strings that differ by at most one from $s_i$. But this is already more than $n$, so we are still done.