Let $p, q$ and $r$ be relatively prime odd positive integers satisfying $$pq + pr - qr = 1$$ and $$1<p<q<r.$$
Define $$f(a,n)= -(q + r)a^2 + 4aqn - 4(q - p)n^2 - 4n$$ where $-p \leq a \leq p+2$ and $1 \leq n \leq r-1$.
Under these conditions, can we prove that
$$\mathrm max \{ f(a,n) \} =f(1,1)$$
Or can one produce a counterexample?
For example, I proved the above idendity for triplets $$(p,q,r)=(2k+1,4k+1,4k+3)$$ and $$(p,q,r)=(4k+3,6k+5,12k+7)$$ by showing $$-f(a,n) + f(1,1) \geq 0$$ with the help of basic number theoretic arguments.
Any contribution (proof suggestion or a counterexample) will make me extremely happy...
EDIT
The above numbers have meaning in topology. For example, consider Pretzel knots $K(-p,q,r)$ where $p, q$ and $r$ are relatively prime odd positive integers satisfying $$pq + pr - qr = 1.$$
Then the Alexander polynomial associated to these knots is $1$, i.e,. $$\Delta_K(K(-p,q,r))=1,$$ see pg. 66, Example 6.9, W. B. Raymond Lickorish - An Introduction to Knot Theory.
Following Freedman's excellent paper, one can conclude that $K(-p,q,r)$ are topologically slice.
demanding $n \geq 1$ cuts out a few counterexamples...