Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated


Use the definitions in L29. Each of the I(P)I(P) triples extends in n−3n-3 ways to a bad four-set. Each four-set contains at most four such triples. Therefore

(n−3)I(P)≤4D(P).(n-3)I(P)\le4D(P).

In every bad four-set, choose two distinct unordered pairs of equal length. If they share an endpoint, their three endpoints are one of the triples counted by I(P)I(P). All four-sets containing any such triple number at most (n−3)I(P)(n-3)I(P). Otherwise their four endpoints are distinct, and the equality is counted by Q(P)Q(P). Each ordered quadruple has only one underlying four-set, so the number of four-sets of this second type is at most Q(P)Q(P). Overlap between the types only enlarges this upper bound.