Throwback sequences

TheoremKozar's conjecture on the throwback sequence#✓fully proved

For a sequence \(Y = (Y_0, Y_1, Y_2, \dots)\) of natural numbers, the throwback step produces the sequence \(Y'\) with \(Y'_i = Y_{i+1}\) for \(i < Y_0\), \(Y'_{Y_0} = Y_0\), and \(Y'_i = Y_i\) for \(i > Y_0\): the leading term \(Y_0\) moves to index \(Y_0\), the terms \(Y_1, \dots, Y_{Y_0}\) shift one place to the left, and all later terms stay in place. Let \(X\) be the sequence with \(X_i = i + 3\) (that is, \(3, 4, 5, \dots\)), and let \(T_k\) be the leading term of the sequence obtained from \(X\) by \(k\) throwback steps. Then for every natural number \(n \ge 3\) and every natural number \(N\) there is \(k \ge N\) with \(T_k = n\); that is, every \(n \ge 3\) equals \(T_k\) for infinitely many \(k\).

This is Kozar's conjecture for OEIS A357081. Darnell and Dribus prove a more general recurrence theorem for every sequence of distinct positive integers.

theorem kozar_throwback_recurrence :
    let throwbackStep : (ℕ → ℕ) → (ℕ → ℕ) := fun Y i =>
      if i < Y 0 then Y (i + 1) else if i = Y 0 then Y 0 else Y i
    let X : ℕ → ℕ := fun i => i + 3
    let T : ℕ → ℕ := fun k => (Nat.iterate throwbackStep k X) 0
    ∀ n : ℕ, 3 ≤ n → ∀ N : ℕ, ∃ k : ℕ, N ≤ k ∧ T k = n
Sources

DefinitionThrowback operation#✓fully proved

Let \(s = (s_0, s_1, s_2, \dots)\) be a sequence of elements of a set \(A\), each element \(a\) carrying a weight \(w(a) \in \mathbb{N}\). The throwback \(\tau_w(s)\) moves the leader \(s_0\) to index \(w(s_0)\), moves the entries at indices \(1, \dots, w(s_0)\) one place left and leaves later entries in place: \(\tau_w(s)_i = s_{i+1}\) for \(i < w(s_0)\), \(\tau_w(s)_{w(s_0)} = s_0\), and \(\tau_w(s)_i = s_i\) for \(i > w(s_0)\).

With \(A = \mathbb{N}\) and \(w\) the identity, this is the throwback \(\tau\) of Darnell and Dribus on sequences of positive integers. With \(A = \mathbb{N}\), the initial arrangement \(s_i = i\) and weights \(w(i) = x_i\) for a sequence \(X = (x_0, x_1, \dots)\), it tracks the original index of each term, so repeated values stay distinguishable.

public def throwback {α : Type*} (w : α → ℕ) (s : ℕ → α) : ℕ → α := fun i =>
  if i < w (s 0) then s (i + 1) else if i = w (s 0) then s 0 else s i
Sources
  • Jesiah Darnell and Benjamin F. Dribus, Throwback Sequences of Positive Integers, Journal of Integer Sequences 28 (2025), Article 25.5.1, Section 2.
Used by 6

DefinitionThrowback sequence#✓fully proved

For a sequence \(X = (x_0, x_1, \dots)\) of natural numbers, the throwback sequence \(T(X) = (t_0, t_1, \dots)\) records the successive leaders under the throwback operation: \(t_n\) is the leader (the entry at index \(0\)) of \(\tau^n(X)\), where \(\tau\) is the throwback operation in which each entry's weight is its own value.

public def throwbackSequence (X : ℕ → ℕ) (n : ℕ) : ℕ :=
  ((throwback id)^[n] X) 0
Sources
  • Jesiah Darnell and Benjamin F. Dribus, Throwback Sequences of Positive Integers, Journal of Integer Sequences 28 (2025), Article 25.5.1, Section 2.
Uses 1 · Used by 6

TheoremThrowback of the integers from 3 on#✓fully proved

Let \(X = (3, 4, 5, \dots)\), that is, \(x_i = i + 3\). The first five terms of its throwback \(\tau(X)\) are \(4, 5, 6, 3, 7\), and the first six terms of its throwback sequence \(T(X)\) are \(3, 4, 5, 6, 3, 7\) (OEIS A357081).

public theorem throwback_examples :
    (List.range 5).map (throwback id (fun i => i + 3)) = [4, 5, 6, 3, 7] ∧
      (List.range 6).map (throwbackSequence (fun i => i + 3)) = [3, 4, 5, 6, 3, 7]
Sources
  • Jesiah Darnell and Benjamin F. Dribus, Throwback Sequences of Positive Integers, Journal of Integer Sequences 28 (2025), Article 25.5.1, Table 1.
Uses 2

DefinitionPartial placement sequences#✓fully proved

Let \(X = (x_0, x_1, \dots)\) be a sequence of natural numbers; the source takes \(X\) strictly increasing with positive terms. The partial placement sequences \(P_n(X)\), for \(n \ge 0\), are indexed by positions \(z \in \mathbb{Z}\), and each position is either empty or holds a term of \(X\). \(P_0(X)\) holds \(x_0\) at the multiples of \(x_0 + 1\) and is empty elsewhere. \(P_{n+1}(X)\) agrees with \(P_n(X)\) at the filled positions of \(P_n(X)\) and fills some of its empty (open) positions with \(x_{n+1}\), as follows. Let \(s\) be the least natural number that is an open position of \(P_n(X)\) (\(s = 0\) if there is none), and let \(d = x_{n+1} - (n + 1) + 1 = x_{n+1} - n\), an integer. An open position \(z\) of \(P_n(X)\) receives \(x_{n+1}\) exactly when \(d\) divides the signed number of open positions of \(P_n(X)\) between \(s\) and \(z\), where open positions in \([s, z)\) count \(+1\) and those in \([z, s)\) count \(-1\); all other open positions stay empty. When \(d > 0\) and \(P_n(X)\) has an open position \(\ge 0\), this places \(x_{n+1}\) at \(s\) and then, in both directions, at every \(d\)-th open position.

public noncomputable def partialPlacement (X : ℕ → ℕ) : ℕ → ℤ → Option ℕ
  | 0 => fun z => if ((X 0 : ℤ) + 1) ∣ z then some (X 0) else none
  | n + 1 => fun z =>
      let P := partialPlacement X n
      let s : ℤ := (sInf {p : ℕ | P p = none} : ℕ)
      let openCount : ℤ := (((Finset.Ico s z).filter fun w => P w = none).card : ℤ) -
        (((Finset.Ico z s).filter fun w => P w = none).card : ℤ)
      if P z = none ∧ ((X (n + 1) : ℤ) - (n + 1) + 1) ∣ openCount then some (X (n + 1))
      else P z
Sources
  • Jesiah Darnell and Benjamin F. Dribus, Throwback Sequences of Positive Integers, Journal of Integer Sequences 28 (2025), Article 25.5.1, Section 4.
Used by 3

TheoremFirst partial placement sequence of the integers from 3 on#✓fully proved

Let \(X = (3, 4, 5, \dots)\), that is, \(x_i = i + 3\). The partial placement sequence \(P_0(X)\) holds the term \(3\) at positions \(0\), \(4\) and \(-4\), and its positions \(1\), \(2\) and \(3\) are empty.

public theorem partialPlacement_zero_examples :
    partialPlacement (fun i => i + 3) 0 0 = some 3 ∧
      partialPlacement (fun i => i + 3) 0 4 = some 3 ∧
      partialPlacement (fun i => i + 3) 0 (-4) = some 3 ∧
      ∀ z ∈ ({1, 2, 3} : Finset ℤ), partialPlacement (fun i => i + 3) 0 z = none
Sources
  • Jesiah Darnell and Benjamin F. Dribus, Throwback Sequences of Positive Integers, Journal of Integer Sequences 28 (2025), Article 25.5.1, Table 4.
Uses 1

TheoremEach term eventually settles or leads infinitely often#●ready to prove

Let \(X = (x_0, x_1, \dots)\) be a sequence of positive integers, with positions numbered from \(0\), the leader being at position \(0\). Track the terms of \(X\) by their original indices: start from the arrangement \((0, 1, 2, \dots)\) and apply the throwback repeatedly, where the leader, with original index \(i\), moves to position \(x_i\). Then for every index \(j\), either there are a position \(p > 0\) and a number \(N\) such that the term with original index \(j\) is at position \(p\) after \(k\) throwbacks for every \(k \ge N\), or there are infinitely many \(k\) such that this term is the leader after \(k\) throwbacks.

public theorem_wanted throwback_final_position_or_frequently_leader (X : ℕ → ℕ)
    (hX : ∀ n, 0 < X n) (j : ℕ) :
    (∃ p, 0 < p ∧ ∃ N, ∀ k ≥ N, ((throwback X)^[k] id) p = j) ∨
      {k | ((throwback X)^[k] id) 0 = j}.Infinite
Sources
  • Jesiah Darnell and Benjamin F. Dribus, Throwback Sequences of Positive Integers, Journal of Integer Sequences 28 (2025), Article 25.5.1, Lemma 1 (jis-2025-25.5.1-lem-1).
Uses 1
Statement usesThrowback operation

TheoremEvery term recurs in the throwback sequence#●ready to prove

Let \(X = (x_0, x_1, \dots)\) be a sequence of pairwise distinct positive integers, and let \(T(X) = (t_0, t_1, \dots)\) be its throwback sequence. Then every term \(x_j\) of \(X\) appears infinitely often in \(T(X)\): there are infinitely many \(k\) with \(t_k = x_j\).

public theorem_wanted throwbackSequence_infinite_eq (X : ℕ → ℕ) (hX : ∀ n, 0 < X n)
    (hinj : Function.Injective X) (j : ℕ) :
    {k | throwbackSequence X k = X j}.Infinite
Sources
  • Jesiah Darnell and Benjamin F. Dribus, Throwback Sequences of Positive Integers, Journal of Integer Sequences 28 (2025), Article 25.5.1, Theorem 2 (jis-2025-25.5.1-thm-2).
Uses 1
Statement usesThrowback sequence

TheoremRecursive bound on the time to become leader#●ready to prove

Let \(X = (x_0, x_1, \dots)\) be a sequence of pairwise distinct positive integers, with positions numbered from \(0\), the leader being at position \(0\). Let \(\tau\) be the throwback and \(T(X) = (t_0, t_1, \dots)\) the throwback sequence, so \(t_k\) is the leader of \(\tau^k(X)\). Let \(n \ge 1\), and let \(N\) be the number of indices \(i\) with \(x_i < n\). Suppose \(a\) and \(b\) are natural numbers such that for every \(m \ge 0\), the term at position \(n - 1\) of \(\tau^m(X)\) equals \(t_{m+k}\) for some \(k \le a\), and the term at position \(N\) of \(\tau^m(X)\) equals \(t_{m+k}\) for some \(k \le b\). Then for every \(m \ge 0\), the term at position \(n\) of \(\tau^m(X)\) equals \(t_{m+k}\) for some \(k \le a + b + 1\).

This is the bound \(M(n) \le M(n - 1) + M(N) + 1\) of the source, stated for upper bounds, where \(M(n)\) is the largest number of throwbacks the term at position \(n\) of any \(\tau^m(X)\) needs to become leader.

public theorem_wanted throwback_leader_time_le (X : ℕ → ℕ) (hX : ∀ n, 0 < X n)
    (hinj : Function.Injective X) (n : ℕ) (hn : 0 < n) (a b : ℕ)
    (ha : ∀ m, ∃ k ≤ a, throwbackSequence X (m + k) = ((throwback id)^[m] X) (n - 1))
    (hb : ∀ m, ∃ k ≤ b,
      throwbackSequence X (m + k) = ((throwback id)^[m] X) (Set.ncard {i | X i < n})) :
    ∀ m, ∃ k ≤ a + b + 1, throwbackSequence X (m + k) = ((throwback id)^[m] X) n
Sources
  • Jesiah Darnell and Benjamin F. Dribus, Throwback Sequences of Positive Integers, Journal of Integer Sequences 28 (2025), Article 25.5.1, Lemma 3 (jis-2025-25.5.1-lem-3).
Uses 2

TheoremLimiting frequencies in throwback sequences#●ready to prove

Let \(X = (x_0, x_1, \dots)\) be a sequence of pairwise distinct positive integers with throwback sequence \(T(X) = (t_0, t_1, \dots)\), and for a natural number \(w\) let \(\sigma(w)\) be the number of indices \(i\) with \(x_i < w\). Then for every index \(j\), writing \(x = x_j\), the number of \(k < N\) with \(t_k = x\), divided by \(N\), converges as \(N \to \infty\) to

\[f(x) = \frac{1}{x - \sigma(x) + 1} \prod_{w} \frac{w - \sigma(w)}{w - \sigma(w) + 1},\]

where the product runs over the values \(w < x\) taken by \(X\).

For a term \(w\) of \(X\), \(\sigma(w)\) is the index of \(w\) in the increasing rearrangement of \(X\).

public theorem_wanted throwbackSequence_limiting_frequency (X : ℕ → ℕ) (hX : ∀ n, 0 < X n)
    (hinj : Function.Injective X) (j : ℕ) :
    Filter.Tendsto
      (fun N : ℕ =>
        ((((Finset.range N).filter fun k => throwbackSequence X k = X j).card : ℕ) : ℝ) / N)
      Filter.atTop
      (nhds (1 / ((X j : ℝ) - Set.ncard {i | X i < X j} + 1) *
        ∏ w ∈ Finset.range (X j),
          if w ∈ Set.range X then
            ((w : ℝ) - Set.ncard {i | X i < w}) / ((w : ℝ) - Set.ncard {i | X i < w} + 1)
          else 1))
Sources
  • Jesiah Darnell and Benjamin F. Dribus, Throwback Sequences of Positive Integers, Journal of Integer Sequences 28 (2025), Article 25.5.1, Theorem 4 (jis-2025-25.5.1-thm-4).
Uses 1
Statement usesThrowback sequence

TheoremEvery term has a nonzero limiting frequency#●ready to prove

Let \(X = (x_0, x_1, \dots)\) be a sequence of pairwise distinct positive integers, and let \(T(X) = (t_0, t_1, \dots)\) be its throwback sequence. Then for every \(j\), the proportion \(\#\{k < N : t_k = x_j\} / N\) of terms equal to \(x_j\) among the first \(N\) terms of \(T(X)\) converges, as \(N \to \infty\), to a nonzero real number.

public theorem_wanted throwbackSequence_limiting_frequency_ne_zero (X : ℕ → ℕ)
    (hX : ∀ n, 0 < X n) (hinj : Function.Injective X) (j : ℕ) :
    ∃ c : ℝ, c ≠ 0 ∧ Filter.Tendsto
      (fun N : ℕ =>
        ((((Finset.range N).filter fun k => throwbackSequence X k = X j).card : ℕ) : ℝ) / N)
      Filter.atTop (nhds c)
Sources
  • Jesiah Darnell and Benjamin F. Dribus, Throwback Sequences of Positive Integers, Journal of Integer Sequences 28 (2025), Article 25.5.1, Corollary 5 (jis-2025-25.5.1-cor-5).
Uses 1
Statement usesThrowback sequence

TheoremPartial placement sequences are periodic#●ready to prove

Let \(X = (x_0, x_1, \dots)\) be a strictly increasing sequence of positive integers and let \(n \ge 0\). Then the partial placement sequence \(P_n(X)\) is periodic: there is an integer \(p > 0\) such that for every position \(z \in \mathbb{Z}\), the positions \(z + p\) and \(z\) of \(P_n(X)\) are either both empty or hold the same term.

public theorem_wanted partialPlacement_periodic (X : ℕ → ℕ) (hX : StrictMono X)
    (hX0 : 0 < X 0) (n : ℕ) :
    ∃ p : ℤ, 0 < p ∧ ∀ z, partialPlacement X n (z + p) = partialPlacement X n z
Sources
  • Jesiah Darnell and Benjamin F. Dribus, Throwback Sequences of Positive Integers, Journal of Integer Sequences 28 (2025), Article 25.5.1, Lemma 6 (jis-2025-25.5.1-lem-6).
Uses 1

TheoremThrowback sequences of increasing sequences are uniformly recurrent#●ready to prove

Let \(X = (x_0, x_1, \dots)\) be a strictly increasing sequence of positive integers with throwback sequence \(T(X) = (t_0, t_1, \dots)\). Then every block of consecutive terms of \(T(X)\) recurs with bounded gaps: for all natural numbers \(i\) and \(L\) there is a natural number \(G\) such that for every natural number \(m\) there is an index \(j\) with \(m \le j \le m + G\) and \(t_{j+s} = t_{i+s}\) for all \(s < L\).

public theorem_wanted throwbackSequence_uniformly_recurrent (X : ℕ → ℕ) (hX : StrictMono X)
    (hX0 : 0 < X 0) (i L : ℕ) :
    ∃ G, ∀ m, ∃ j, m ≤ j ∧ j ≤ m + G ∧
      ∀ t < L, throwbackSequence X (j + t) = throwbackSequence X (i + t)
Sources
  • Jesiah Darnell and Benjamin F. Dribus, Throwback Sequences of Positive Integers, Journal of Integer Sequences 28 (2025), Article 25.5.1, Theorem 7 (jis-2025-25.5.1-thm-7).
Uses 1
Statement usesThrowback sequence

TheoremMinimal complete blocks of partial placements match configurations#●ready to prove

Let \(X = (x_0, x_1, \dots)\) be a strictly increasing sequence of positive integers, let \(\tau\) be the throwback, let \(n \ge 0\), and write \(P_n(X)_z\) for the entry, possibly empty, of the partial placement sequence \(P_n(X)\) at position \(z \in \mathbb{Z}\). The minimal complete block at \(z\) is the word \(P_n(X)_z, P_n(X)_{z+1}, \dots, P_n(X)_{z+L-1}\), empty positions included, where \(L\) is the least length for which these positions contain every \(x_i\) with \(i \le n\) (\(L = 0\) if there is no such length). Then for all natural numbers \(k\) and \(l\), the minimal complete blocks at \(k\) and at \(l\) are equal if and only if each of \(x_0, \dots, x_n\) occupies the same positions in \(\tau^k(X)\) as in \(\tau^l(X)\). Moreover, for every \(z \in \mathbb{Z}\) the minimal complete block at \(z\) equals the minimal complete block at some natural number \(k\). So minimal complete blocks of \(P_n(X)\) correspond bijectively to the configurations of \(x_0, \dots, x_n\) in the sequences \(\tau^k(X)\) with \(k \ge 0\).

public theorem_wanted partialPlacement_minimal_complete_blocks_iff_configurations
    (X : ℕ → ℕ) (hX : StrictMono X) (hX0 : 0 < X 0) (n : ℕ) :
    let len : ℤ → ℕ := fun z =>
      sInf {L | ∀ i ≤ n, ∃ t < L, partialPlacement X n (z + t) = some (X i)}
    let block : ℤ → List (Option ℕ) := fun z =>
      (List.range (len z)).map fun t => partialPlacement X n (z + t)
    (∀ k l : ℕ, block k = block l ↔
      ∀ i ≤ n, ∀ p, (((throwback id)^[k] X) p = X i ↔ ((throwback id)^[l] X) p = X i)) ∧
      ∀ z : ℤ, ∃ k : ℕ, block z = block k
Sources
  • Jesiah Darnell and Benjamin F. Dribus, Throwback Sequences of Positive Integers, Journal of Integer Sequences 28 (2025), Article 25.5.1, Lemma 8 (jis-2025-25.5.1-lem-8).
Uses 2

TheoremPerfect mixing for consecutive integers#●ready to prove

Let \(m \ge 1\) and \(n \ge 0\), let \(X = (m, m + 1, m + 2, \dots)\), and let \(\tau\) be the throwback. Then the positions of the terms \(m, m + 1, \dots, m + n\) in \(\tau^k(X)\) are different for different \(k\) with \(0 \le k < (m + 1)^{n + 1}\): for \(0 \le k < l < (m + 1)^{n + 1}\), some \(m + i\) with \(i \le n\) occupies different positions in \(\tau^k(X)\) and \(\tau^l(X)\). Moreover, each of \(m, m + 1, \dots, m + n\) occupies the same positions in \(\tau^{(m+1)^{n+1}}(X)\) as in \(X\).

public theorem_wanted throwback_consecutive_perfect_mixing (m n : ℕ) (hm : 0 < m) :
    let Y : ℕ → ℕ → ℕ := fun k => (throwback id)^[k] (fun i => m + i)
    (∀ i ≤ n, ∀ p, (Y ((m + 1) ^ (n + 1)) p = m + i ↔ Y 0 p = m + i)) ∧
      ∀ k l, k < l → l < (m + 1) ^ (n + 1) →
        ¬ ∀ i ≤ n, ∀ p, (Y k p = m + i ↔ Y l p = m + i)
Sources
  • Jesiah Darnell and Benjamin F. Dribus, Throwback Sequences of Positive Integers, Journal of Integer Sequences 28 (2025), Article 25.5.1, Theorem 9 (jis-2025-25.5.1-thm-9).
Uses 1
Statement usesThrowback operation
6 fully proved9 ready to prove

The dependency explorer needs JavaScript and its graph data.

Download the complete graph data.