Problem 313: Sliding Game
View on Project EulerProject Euler Problem 313 Solution
EulerSolve provides an optimized solution for Project Euler Problem 313, Sliding Game, with C++, Python, Java, and a step-by-step mathematical explanation.
Problem Summary Let \(S(m,n)\) be the minimum number of moves needed to move the red counter from the top-left corner of an \(m\times n\) board to the bottom-right corner, with the empty cell initially at the bottom-right corner. The problem asks for the number of grids such that $S(m,n)=p^2,\qquad p<10^6,\ p\text{ prime}.$ The key fact is that the sliding game has an affine move formula, so the problem becomes a pure lattice-point count for each prime square. Mathematical Approach 1) Symmetry and small data. Because rotating the board swaps \(m\) and \(n\) without changing the puzzle, $S(m,n)=S(n,m).$ A brute-force BFS on small boards gives $S(2,2)=5,\quad S(3,2)=9,\quad S(4,2)=15,\quad S(5,2)=21,$ $S(4,3)=17,\quad S(5,4)=25,\quad S(6,5)=33.$ The pattern is clear: $S(m,2)=6m-9\qquad (m\ge 3),$ and every extra row beyond \(2\) adds exactly two more moves. 2) Closed form for non-square boards. Assume \(m>n\ge 2\). Starting from the \(m\times 2\) strip, each extra row gives two extra blank-cell shuffles, so $S(m,n)=S(m,2)+2(n-2)=6m+2n-13.$ This formula matches the sample from the statement: $S(5,4)=6\cdot 5+2\cdot 4-13=25.$ 3) What about square boards?...
Detailed mathematical approach
Problem Summary
Let \(S(m,n)\) be the minimum number of moves needed to move the red counter from the top-left corner of an \(m\times n\) board to the bottom-right corner, with the empty cell initially at the bottom-right corner.
The problem asks for the number of grids such that
$S(m,n)=p^2,\qquad p<10^6,\ p\text{ prime}.$
The key fact is that the sliding game has an affine move formula, so the problem becomes a pure lattice-point count for each prime square.
Mathematical Approach
1) Symmetry and small data.
Because rotating the board swaps \(m\) and \(n\) without changing the puzzle,
$S(m,n)=S(n,m).$
A brute-force BFS on small boards gives
$S(2,2)=5,\quad S(3,2)=9,\quad S(4,2)=15,\quad S(5,2)=21,$
$S(4,3)=17,\quad S(5,4)=25,\quad S(6,5)=33.$
The pattern is clear:
$S(m,2)=6m-9\qquad (m\ge 3),$
and every extra row beyond \(2\) adds exactly two more moves.
2) Closed form for non-square boards.
Assume \(m>n\ge 2\). Starting from the \(m\times 2\) strip, each extra row gives two extra blank-cell shuffles, so
$S(m,n)=S(m,2)+2(n-2)=6m+2n-13.$
This formula matches the sample from the statement:
$S(5,4)=6\cdot 5+2\cdot 4-13=25.$
3) What about square boards?
For \(m=n\ge 3\), brute force gives
$S(3,3)=13,\quad S(4,4)=21,\quad S(5,5)=29,$
so the square case follows
$S(n,n)=8n-11.$
But for an odd prime \(p\),
$p^2\equiv 1\pmod 8,$
whereas
$8n-11\equiv 5\pmod 8.$
So no square board can satisfy \(S(n,n)=p^2\). The case \(p=2\) also fails. Therefore only the non-square formula matters for prime squares.
4) Convert the move equation into an integer interval.
Set
$q=p^2.$
We must count integer pairs \((m,n)\) with \(m>n\ge2\) such that
$6m+2n-13=q.$
Divide by \(2\):
$3m+n=\frac{q+13}{2}=:h.$
Now solve for \(n\):
$n=h-3m.$
The conditions become:
From \(n\ge 2\),
$m\le \left\lfloor\frac{h-2}{3}\right\rfloor.$
From \(m>n\),
$m>h-3m\iff 4m>h\iff m\ge \left\lfloor\frac{h}{4}\right\rfloor+1.$
So the valid values of \(m\) lie in the interval
$m_{\min}=\left\lfloor\frac{h}{4}\right\rfloor+1,\qquad m_{\max}=\left\lfloor\frac{h-2}{3}\right\rfloor.$
5) Why the final count is multiplied by \(2\).
The interval above counts only boards with \(m>n\). But each such board has a transposed partner \(n\times m\), and both are distinct grids in the problem statement. Therefore the contribution of \(q=p^2\) is
$f(q)=2\cdot \max\bigl(0,\ m_{\max}-m_{\min}+1\bigr).$
This is exactly the formula implemented in the C++ solver.
6) Simplify the formula for odd primes \(p>3\).
Every odd prime larger than \(3\) satisfies
$p^2\equiv 1\pmod{24},$
so write
$p^2=24k+1.$
Then
$h=\frac{p^2+13}{2}=12k+7,$
and therefore
$m_{\min}=\left\lfloor\frac{12k+7}{4}\right\rfloor+1=3k+2,$
$m_{\max}=\left\lfloor\frac{12k+5}{3}\right\rfloor=4k+1.$
The number of admissible \(m\) values is
$m_{\max}-m_{\min}+1=(4k+1)-(3k+2)+1=k,$
so
$f(p^2)=2k=\frac{p^2-1}{12}\qquad (p>3).$
The exceptional prime \(p=3\) gives
$q=9,\quad h=11,\quad m_{\min}=3,\quad m_{\max}=3,\quad f(9)=2.$
And \(p=2\) contributes \(0\).
7) Final counting formula.
Hence
$\#\{(m,n):S(m,n)=p^2,\ p<P\} = 2+\sum_{\substack{3<p<P\\ p\text{ prime}}}\frac{p^2-1}{12},$
with the leading \(2\) coming from \(p=3\).
Worked Examples
The sample \(p<10\). The primes are \(2,3,5,7\).
\(p=2\) contributes \(0\), \(p=3\) contributes \(2\), \(p=5\) contributes \(2\), and \(p=7\) contributes \(4\). So the total is
$0+2+2+4=8,$
which matches the checkpoint in the code.
The sample \(p<100\). Summing the same formula over all primes below \(100\) gives
$5482,$
exactly as stated in the problem.
Algorithm
1) Generate all primes \(p<10^6\) with the sieve of Eratosthenes.
2) For each prime, compute \(q=p^2\).
3) Evaluate
$h=\frac{q+13}{2},\qquad m_{\min}=\left\lfloor\frac{h}{4}\right\rfloor+1,\qquad m_{\max}=\left\lfloor\frac{h-2}{3}\right\rfloor.$
4) Add
$2\cdot \max(0,m_{\max}-m_{\min}+1).$
This is \(O(1)\) work per prime.
Complexity Analysis
The sieve costs \(O(P\log\log P)\) time and \(O(P)\) memory. After that, the summation is linear in the number of primes, with constant-time arithmetic for each one.
Checks And Final Result
The implementation verifies
$\text{solve}(10)=8,\qquad \text{solve}(100)=5482.$
The full answer is
$2057774861813004.$
Further Reading
- Problem page: https://projecteuler.net/problem=313
- Sieve of Eratosthenes: https://en.wikipedia.org/wiki/Sieve_of_Eratosthenes
- Diophantine equations: https://en.wikipedia.org/wiki/Diophantine_equation