2019 points are chosen at random, independently, and distributed
uniformly in the unit disc
{(x,y)∈R2:x2+y2≤1}. Let C be the
convex hull of the chosen points. Which probability is larger:
that C is a polygon with three vertices, or a polygon with four
vertices?
Proposed by Fedor Petrov, St. Petersburg State University
Solution (official)
We will show that the quadrilateral has larger probability.
Let D={(x,y)∈R2:x2+y2≤1}. Denote
the random points by X1,…,X2019 and let
pq=P(C is a triangle with vertices X1,X2,X3),=P(C is a convex quadrilateral with vertices X1,X2,X3,X4).
By symmetry we have
P(C is a triangle)=(32019)p,
P(C is a quadrilateral)=(42019)q and we need to prove that
(42019)q>(32019)p, or equivalently
p<42016q=504q.
Note that p is the average over X1,X2,X3 of the following
expression:
u(X1,X2,X3)=P(X4∈△X1X2X3)⋅P(X5,X6,…,X2019∈△X1X2X3),
and q is not less than the average over X1,X2,X3 of
v(X1,X2,X3)=P(X1,X2,X3,X4 form a convex quad.)⋅P(X5,X6,…,X2019∈△X1X2X3).
Thus it suffices to prove that
u(X1,X2,X3)≤500v(X1,X2,X3) for all
X1,X2,X3. It reads as
area(△X1X2X3)≤500area(Ω), where
Ω={Y:X1,X2,X3,Y form a convex quadrilateral}. Assume the contrary, i.e.,
area(△X1X2X3)>500area(Ω).
Let the lines X1X2, X1X3, X2X3 meet the boundary of
D at A1,A2,A3,B1,B2,B3; these lines divide D into
7 regions as shown in the picture;
Ω=D4∪D5∪D6.
By our indirect assumption,
area(D4)+area(D5)+area(D6)=area(Ω)<5001area(D0)<5001area(D)=500π.
From △X1X3B3⊂Ω we get
X3B3/X3X2=area(△X1X3B3)/area(△X1X2X3)<1/500, so
X3B3<5001X2X3<2501. Similarly, the
lengths segments
A1X1,B1X1,A2X2,B2X2,A3X2 are less than
2501.
The regions D1,D2,D3 can be covered by disks with radius
2501, so
area(D1)+area(D2)+area(D3)<3⋅2502π.
Finally, it is well-known that the area of any triangle inside the
unit disk is at most 433, so
area(D0)≤433.
But then
i=0∑6area(Di)<433+3⋅2502π+500π<area(D),
contradiction.
How the field did
contestants scored
360
average (of 10)
0.62
solved (≥ 80%)
2.2%
near-0 (≤ 10%)
88.6%
discrimination
0.36
Score distribution (field cohort)
Computed on contestants with a meaningful total (field cohort); discrimination is the corrected item–total correlation.