Let $C=V(f)\subset \mathbb{P}^2$ be a smooth plane algebraic curve of degree $d$.
For all $x\in \mathbb{P}^2\setminus C$ there are at most $d(d-1)$ tangent lines to $C$ passing through $x$.
There exists an $x$ such that we have an equality in the statement above.
For the first part I have tried to do this: After a translation we can assume that $x=(0:0:1)$, then as an arbitrary line passing through $x$ has exactly one intersection with the line at infinity we have an unique parametric representation in the form $l_{(a:b)}=\{(a\mu:b\mu:\lambda); \lambda,\mu \in k\}$ for some $(a:b)\in \mathbb{P}^1$.
Now, $l_{(a:b)}$ is tangent to $C$ $\iff$ one of the equations $f(ax,bx,1)=0$ or $f(a,b,x)=0$ has a multiple root $\iff$ one of the associated discriminants $\Delta_{(a:b)}$ or $\hat{\Delta}_{(a:b)}$ vanished.
If $(a:b)\neq (1:0)$ we can take $b=1$ and studying $P(a)=\Delta_{(a:1)}$ and $Q(a)=\hat{\Delta}_{(a:1)}$ as polynomials in $a$ so it would be enough to prove the following
$\deg(P)+\deg(Q)+\#(\{\Delta_{(1:0)}\}\cap \{0\})+\#(\{\hat{\Delta}_{(1:0)}\}\cap \{0\})\leq d(d-1)$
But I don't know how to compute $\deg(P)$ or $\deg(Q)$ so I don't know hot to continue.
For the second part perhaps I could take $x=(c:0:1)$ and make computations similar to those above to obtain expressions of the form $\Delta_{c,(a:b)}$ and $\hat{\Delta}_{c,(a:b)}$. Then if we define $P^c(a)=\Delta_{c,(a:1)}$ and $Q^c(a)=\hat{\Delta}_{c,(a:1)}$ maybe we can choose $c\in k$ such that $P^c$ and $Q^c$ have different roots and $\Delta_{c,(1:0)},\hat{\Delta}_{c,(1:0)}\neq 0$. Then $\deg(P^c)+\deg(Q^c)$ would be equal to the number of tangent lines, and a good understanding of the first part should give us a way to compute this.
As noticed here assuming Hurwitz's formula this problem is equivalent to Plücker's Formula and that's where my interest comes from.
