Given are $n$ points with different abscissas in the plane. Through every pair points is drawn a parabola - a graph of a square trinomial with leading coefficient equal to $1$. A parabola is called $good$ if there are no other marked points on it, except for the two through which it is drawn, and there are no marked points above it (i.e. inside it). What is the greatest number of $good$ parabolas?
Problem
Source: St Petersburg 2021 9.4
Tags: algebra
cadaeibf
23.12.2021 15:35
Isn't it trivially $\binom{n}{2}$ by choosing points inductively so that they don't lie on the previous parabolas?
VicKmath7
23.12.2021 15:49
Indeed, there's an additional condition, but I can't really understand it, now I have corrected it with my interpretation. Here's the Russian text, if someone can translate it better: На плоскости отмечено n точек с разными абсциссами. Через каждую пару точек провели параболу — график квадратного трехчлена с единичным старшим коэффициентом. Парабола называется хорошей, если ни на ней, ни выше неё нет отмеченных точек, кроме тех двух, через которые она проведена. Какое наибольшее количество хороших парабол могло получиться?
Tintarn
12.01.2022 21:31
$n-1$
Replace the points $(x,y)$ by $(x,y-x^2)$. What happens?
We do the replacement as suggested in the hint. This replaces the parabolas between the old points by the lines between the new points.
So we get the same problem as before, but now with parabolas replaced by straight lines everywhere, and this is of course much simpler:
If the convex hull of the $n$ points is a $k$-gon, then obviously at most the $k$ boundary lines are good, since every other line has points on both sides of it or a third point on it.
But among the $k$ boundary lines, at most $k-1$ of them are good, since at least one of them has the convex hull above it instead of below (take the line containing the lowest point of the convex hull).
Since $k \le n$, we therefore have at most $n-1$ good lines, and hence at most $n-1$ good parabolas.
Clearly, this is sharp as one can see from just choosing the points on the $x$-axis (before the change of variables).