Erdős Problem #1133 — Let C>0C>0.

About 59 years old · traced to

Let C>0C>0. There exists ϵ>0\epsilon>0 such that if nn is sufficiently large the following holds. For any x1,…,xn∈[−1,1]x_1,\ldots,x_n\in [-1,1] there exist y1,…,yn∈[−1,1]y_1,\ldots,y_n\in [-1,1] such that, if PP is a polynomial of degree m<(1+ϵ)nm<(1+\epsilon)n with P(xi)=yiP(x_i)=y_i for at least (1−ϵ)n(1-\epsilon)n many 1≤i≤n1\leq i\leq n, then max⁡x∈[−1,1]∣P(x)∣>C.\max_{x\in [-1,1]}\lvert P(x)\rvert >C.

References

Progress summary

Refreshed
Open

The conjecture remains open: an older weaker theorem is known, and a proposed argument has not been verified.

Erdős’s problem asks whether every sufficiently large node set in [−1,1][-1,1] admits bounded labels such that any polynomial of degree below (1+ε)n(1+\varepsilon)n fitting almost all labels must have arbitrarily large uniform norm. No proposer date beyond Erdős’s attribution is recorded.

Known results

  • Erdős, reported in 1967: for every C>0C>0, some ε>0\varepsilon>0 works so that, for sufficiently large nn, every set of m=loor(1+ε)nm=loor{(1+\varepsilon)n} nodes admits a degree-nn polynomial bounded by 11 on the nodes but exceeding CC on [−1,1][-1,1].
  • Erdős reported that he could not prove even the case m=nm=n.

Unverified density-principle argument

A proposed argument using Beurling’s interpolation theorem, the substitution x=cos⁡θx=\cos\theta, and a block-pigeonhole principle claims to yield the robust obstruction. It has not been verified or published as a solution, and the problem remains marked open.

Current status (as of August 2026): Erdős’s weaker result is known, but the stated robust interpolation conjecture remains open and has no verified proof or counterexample.

Sources

Solutions 0

No solutions have been posted yet.