Skip to content

Erdős Problem #388

Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: open (not automatically the live result).
Subject areas: number theory.

Read this status correctly

The status above is reproduced from GCL's immutable external-catalogue snapshot. It is not a new claim of current mathematical resolution, a source-semantic review, an independent proof check, or GCL certification. A published proof and a GCL replay are separate events.

The question in plain language

Erdős Problem 388 asks about collisions between two products of consecutive positive integers. Write

\[ P(m,k)=(m+1)(m+2)\cdots(m+k). \]

The main question is whether there are only finitely many quadruples \((m_1,k_1,m_2,k_2)\) such that \(k_1,k_2>3\), the second block starts after the first ends, \(m_1+k_1\leq m_2\), and \(P(m_1,k_1)=P(m_2,k_2)\). “Classify” asks for more: describe every such quadruple, not merely prove that their number is finite.

The displayed problem on the maintained page does not spell out the domains of \(m_1\) and \(m_2\). The original source discusses positive integers, while the current formal interface chooses natural numbers. That choice matters: allowing negative starting points introduces sign and symmetry phenomena that are not the intended separated positive-block question. This article therefore treats the nonnegative-natural-number interface as the working normalization, while recording the domain choice as a formalization decision rather than silently attributing it to the source.

The same page gives a broader variant with fixed positive multipliers \(a\) and \(b\), but it does not specify all domain, length, or separation hypotheses. That variant is useful context, not a substitute for the exact main statement.

A small example

There is a genuine separated collision:

\[ 8\cdot9\cdot10\cdot11\cdot12\cdot13\cdot14 =17{,}297{,}280 =63\cdot64\cdot65\cdot66. \]

It has \((m_1,k_1,m_2,k_2)=(7,7,62,4)\). The left block ends at \(14\), so \(m_1+k_1=14\leq62=m_2\), and both lengths exceed \(3\). An exact-arithmetic Python check using integer prod returned 17297280 17297280 True; a second check returned True True True for equality, separation, and the two length conditions. This verifies one solution, not its uniqueness and not finiteness of the full four-parameter family.

One elementary restriction is already informative. Under the separation condition, every right-hand factor is larger than every left-hand factor. If \(k_2\geq k_1\), the first \(k_1\) right-hand factors already have product strictly larger than the left-hand product. Therefore any solution must have \(k_1>k_2\). This is a proved local lemma for the normalized positive setting, not a solution of the problem.

What is known

The primary source is Erdős’s 1976 paper, Problems and results on number theoretic properties of consecutive integers and related questions. In its discussion around equation (8), printed on the early pages of the paper, Erdős asks whether separated products of consecutive integers have only finitely many solutions. The following paragraphs discuss a rational-ratio generalization and then a related “same prime factors” question. The paper presents these as conjectural questions and records substantial difficulty; it does not provide a resolution of #388.

The maintained page associates the question with Erdős’s 1976 paper, the Erdős–Graham 1980 monograph, and Erdős’s 1992 survey. Those references establish provenance for the question, but a bibliography entry is not a proof of the modern formulation. The live discussion has ten user comments and no proof exposition. One comment reports an attempted fixed-pair finiteness argument using polynomial Diophantine-equation machinery; another explicitly says that an automated check judged the proof incorrect. These are useful leads for a future source-critical review, not accepted results. In particular, they do not settle the global question where both lengths vary.

The exact example above is an independently checked finite calculation. Search reports in the discussion are not treated as exhaustive certificates here because their code, bounds, and reproducibility record are not part of the protected source packet. Likewise, a plausible smooth-number obstruction—large prime factors on the upper block cannot occur in an equality—is a useful observation, but it does not by itself prove that only finitely many blocks can occur.

GCL's current state

The bounded assignment is original issue #1658, assignment ERDOS-CATALOGUE-0388-S01. The issue’s frozen packet pins teorth/erdosproblems commit 536408716724a2402bdfa7c1d764bf83d64a8c89 and the data/problems.yaml blob 4e60263a80af6207aca494d2e1388e44fa372889. At that revision, #388 is open, last updated 2025-08-31, without a prize; its formal-status field says unformalized.

The live problem page, accessed 2026-10-13, still presents the question as open and was edited 2026-04-11. It now displays a “formalised statement” link. The live formal-conjectures file was added at commit b7933eb0a4826be80e2d917f3393d12530d36896, file blob 40b7576b0c0cbbf4d2b357468435133731bf60ee. It contains three declarations: finiteness, exact classification, and a restricted general variant. All three end with sorry; the declarations are formal statement interfaces, not formal proofs. The issue packet separately says no statement file was present in the admitted FC-GDM-85F86371 snapshot, so the newer live file must not be retroactively called part of that frozen snapshot.

GCL has not supplied mathematical certification, Lean replay, or a protected proof result for this problem. Documentary acceptance is separate: the protected V2 ledger alone decides accepted compendium coverage.

What remains

The first need is a source-locked semantic review: compare the original equation (8), the maintained transcription, and the natural-number formalization, including whether “positive integers” and the separation inequality are intended in every variant. The broader \(aP(m_1,k_1)=bP(m_2,k_2)\) statement needs its missing hypotheses recorded before any theorem is transferred to it.

A bounded mathematical successor should then test one fixed pair of lengths, with a primary theorem in hand and a complete check that its hypotheses match the polynomial products here. A separate exact-arithmetic search can enumerate a declared finite box and emit all hits with independently recomputed products. Either outcome is useful: a verified reduction would isolate a theorem-sized residual, while a counterexample would falsify any uniqueness claim. Neither task should be described as resolving the varying-length problem unless it actually controls all lengths.

Sources and claim boundaries

The source trail is: Erdős 1976, primary PDF, especially the discussion of equations (8)–(10); the maintained #388 page and discussion; and the catalogue’s pinned YAML. The page’s bibliographic records are Er76d, ErGr80, and Er92e. The formal statement is the pinned-by-commit file.

The catalogue status, live-page status, forum reports, exact example, and sorry-based Lean declarations have different evidentiary roles. None is a GCL mathematical certification, novelty decision, or publication decision. The original SOURCE RESULT/1 citation is authenticated original RESULT/1. Any later claim of accepted coverage must be read from the protected V2 ledger and its exact manuscript evidence, not inferred from this explanatory article.

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #1658 — protected issue/assignment identity; its labels and results are live, not reproduced automatically here.
  • GCL triage: STATEMENT_CAPTURE_REQUIRED.
  • Formal-statement coverage: No statement was confirmed in GCL's pinned formal-provider snapshot. Presence does not mean the statement or proof has been checked.
  • External mathematical status: consult current problem page and its dated bibliography.

Back to catalogue · Compendium introduction