condorcet_stationary_weight_restriction_obstruction
There exist an enumeration of all56 three-candidate subsets of eight candidates, an outside challenger for every triple, bijective member enumerators, and the insert-challenger/evict-one-member transition function, such that every nonnegative integer weighting stationary for this three-transition chain has total weight at most twice its identity-ranking claim score. Nevertheless this selector has unrestricted nonnegative integer weights and a natural-valued subset potential satisfying every local removal inequality with twice its full-set value strictly below the total weight.
Statement
theorem condorcet_stationary_weight_restriction_obstruction : ∃ (S : Fin 56 → Finset (Fin 8)) (f : Fin 56 → Fin 8) (d : Fin 56 → Fin 3 → Fin 8) (next : Fin 56 → Fin 3 → Fin 56), (∀ i, (S i).card = 3 ∧ f i ∉ S i ∧ Function.Injective (d i) ∧ S i = univ.image (d i)) ∧ (∀ T : Finset (Fin 8), T.card = 3 → ∃ i, S i = T) ∧ (∀ i j, S (next i j) = insert (f i) ((S i).erase (d i j))) ∧ (∀ w : Fin 56 → ℕ, (∀ j, 3 * w j = ∑ i, ∑ k : Fin 3, if next i k = j then w i else 0) → (∑ i, w i) ≤ 2 * ∑ i, if ∀ b ∈ S i, f i < b then w i else 0) ∧ (∃ (w : Fin 56 → ℕ) (H : Finset (Fin 8) → ℕ), (∀ A : Finset (Fin 8), ∀ a ∈ A, (∑ i, if f i = a ∧ S i ⊆ A.erase a then w i else 0) + H (A.erase a) ≤ H A) ∧ 2 * H univ < ∑ i, w i)
- Source
- IN-339 auxiliary restriction diagnostic for OP-19. Exact integer data from experiments8322 and8326; related strict election definition from Condorcet winning sets, Social Choice and Welfare44(3):493-517 (2015). This refutes an extra stationarity restriction, not OP-19, and carries no research advance.
- Verified
- 29 Oct 2026
- Axioms
Classical.choiceQuot.soundpropext- Built on 3
- condorcet_ordering_score_le_subset_potentialFor any finite family of challenger claims and nonnegative integer claim weights, suppose a natural-valued potential on remaining…
- condorcet_stationary_telescoping_boundOn any 3-regular directed transition graph on 56 vertices, if a potential vector p and positive bound K satisfy K + 3 * p i <= C * score i…
- condorcet_stationary_certificate_existsExplicit constructive witnesses (S, f, d, next, p) on 8 candidates and 56 triples, verified by finite decision, satisfying the subset…
The proof, explained
The theorem establishes an obstruction between stationary Markov chain weightings and general subset potentials for a Condorcet tournament selection process on 8 candidates with 56 three-candidate triples. An explicit transition system and a potential function on triples are verified computationally to satisfy a local drift inequality at every state. For any stationary weighting on this transition graph, the Markov balance condition cancels the potential terms, implying that the total weight is bounded by twice the identity-ranking claim score. Finally, an explicit 3-element weight vector and subset indicator potential are verified to satisfy every local removal inequality while having twice the potential value on the full candidate set strictly below the total weight.
At every state $i \\in \\{0, \\dots, 55\\}$, the potential $p$ satisfies $3 p(i) + 70 \\le \\sum_{k=0}^2 p(\\text{next}(i, k)) + 139 \\cdot \\text{score}(i)$. automation
This local drift inequality is verified by exhaustive kernel computation (`decide +kernel`) over all 56 states, evaluating the explicit tables for the potential $p$, the transition function $\\text{next}$, and the identity-ranking claim score.
decide +kernelEach subset $S_i$ contains exactly 3 candidates, the challenger $f_i$ is not in $S_i$, and $d_i$ is an injective enumeration of $S_i$. automation
Verified by exhaustive kernel computation (`decide +kernel`) across all 56 entries of the tables defining the candidate triples $S$, the challengers $f$, and member enumerators $d$.
decide +kernelThe family $S$ contains all $\\binom{8}{3} = 56$ three-element subsets of candidates. automation
Verified by exhaustive kernel computation (`decide +kernel`) checking that every 3-element subset of 8 candidates matches some $S_i$.
decide +kernelEach transition replaces candidate $d(i, j)$ in triple $S_i$ with the outside challenger $f(i)$. automation
Verified by exhaustive kernel computation (`decide +kernel`) across all $56 \\times 3 = 168$ transitions.
decide +kernelThere exist unrestricted weights $w$ and a subset potential $H$ satisfying every local removal inequality such that $2 H(\\text{univ}) < \\sum_i w_i$. automation
We define $w$ to have weight 1 on states 18, 19, and 20 (and 0 elsewhere), and $H(A)$ to be 1 if $\\{5, 6, 7\\} \\subseteq A$ and 0 otherwise. Kernel decision (`decide +kernel`) confirms that $H$ satisfies the local removal inequalities for all subsets $A$ and elements $a \\in A$, and that $2 H(\\text{univ}) = 2 < 3 = \\sum_i w_i$.
let w : Fin 56 → ℕ := fun i => if i = 18 ∨ i = 19 ∨ i = 20 then 1 else 0 let H : Finset (Fin 8) → ℕ := fun A => if ({5, 6, 7} : Finset (Fin 8)) ⊆ A then 1 else 0 refine ⟨w, H, ?_, ?_⟩ · decide +kernel · decide +kernel
For any stationary weight vector $w$, the potential balance identity $3 \\sum_i w_i p_i = \\sum_i w_i \\sum_{k=0}^2 p(\\text{next}(i, k))$ holds.
We expand $3 \\sum_i w_i p_i = \\sum_j (3 w_j) p_j$, substitute the Markov stationarity hypothesis $3 w_j = \\sum_{i, k : \\text{next}(i, k) = j} w_i$, and swap the order of summation. Summing over $j$ collapses the indicator condition $\\text{next}(i, k) = j$, leaving $\\sum_i w_i \\sum_{k=0}^2 p(\\text{next}(i, k))$.
uses
Finset.mul_sum,Nat.mul_assoc,sum_congr,Finset.sum_mul,Finset.sum_comm,ite_mul,zero_mul,Finset.sum_ite_eq,Finset.mem_univ,if_truehave hbalance : 3 * (∑ i, w i * p i) = ∑ i, w i * (∑ k : Fin 3, p (next i k)) := by calc 3 * (∑ i, w i * p i) = ∑ j, (3 * w j) * p j := by simp only [Finset.mul_sum, Nat.mul_assoc] _ = ∑ j, (∑ i, ∑ k : Fin 3, if next i k = j then w i else 0) * p j := by apply sum_congr rfl intro j _ rw [hstat j] _ = ∑ i, w i * (∑ k : Fin 3, p (next i k)) := by simp only [Finset.sum_mul] rw [Finset.sum_comm] apply sum_congr rfl intro i _ rw [Finset.sum_comm] simp only [ite_mul, zero_mul] simp only [Finset.sum_ite_eq, Finset.mem_univ, if_true] rw [Finset.mul_sum]
The weighted sum of the local potential inequalities holds: $\\sum_i w_i (3 p_i + 70) \\le \\sum_i w_i (\\sum_{k=0}^2 p(\\text{next}(i, k)) + 139 s_i)$.
Multiplying each local inequality by the nonnegative weight $w_i$ and summing over all states $i$ yields the combined inequality.
uses
Finset.sum_le_sum,Nat.mul_le_mul_lefthave hsum := Finset.sum_le_sum (fun i (_ : i ∈ (univ : Finset (Fin 56))) => Nat.mul_le_mul_left (w i) (hlocal i))
The left-hand side expands as $\\sum_i w_i (3 p_i + 70) = 3 \\sum_i w_i p_i + 70 \\sum_i w_i$.
Distribute the multiplication by $w_i$ over addition, split the sum into two separate sums using `sum_add_distrib`, and factor out the constant multipliers 3 and 70 using `mul_sum`.
uses
sum_congr,sum_add_distrib,mul_sumhave hleft : (∑ i, w i * (3 * p i + 70)) = 3 * (∑ i, w i * p i) + 70 * (∑ i, w i) := by calc _ = ∑ i, (3 * (w i * p i) + 70 * w i) := by apply sum_congr rfl intro i _ ring _ = _ := by rw [sum_add_distrib, ← mul_sum, ← mul_sum]
The right-hand side expands as $\\sum_i w_i (\\sum_k p(\\text{next}(i, k)) + 139 s_i) = \\sum_i w_i \\sum_k p(\\text{next}(i, k)) + 139 \\sum_i w_i s_i$.
Distribute $w_i$ across the sum, split the sum using `sum_add_distrib`, and factor out 139 from the second term using `mul_sum`.
uses
sum_congr,sum_add_distrib,mul_sumhave hright : (∑ i, w i * ((∑ k : Fin 3, p (next i k)) + 139 * s i)) = (∑ i, w i * (∑ k : Fin 3, p (next i k))) + 139 * (∑ i, w i * s i) := by calc _ = ∑ i, (w i * (∑ k : Fin 3, p (next i k)) + 139 * (w i * s i)) := by apply sum_congr rfl intro i _ ring _ = _ := by rw [sum_add_distrib, ← mul_sum]
Any stationary weighting $w$ satisfies $\\sum_i w_i \\le 2 \\sum_i w_i s_i$. automation
Substituting the left and right expansions into the weighted inequality and canceling the equal balance terms $3 \\sum w_i p_i = \\sum w_i \\sum_k p(\\text{next}(i, k))$ leaves $70 \\sum_i w_i \\le 139 \\sum_i w_i s_i$. Since $2 \\times 70 = 140 > 139$, linear integer arithmetic (`omega`) deduces $\\sum_i w_i \\le 2 \\sum_i w_i s_i$.
rw [hleft, hright] at hsum have hbound : 70 * (∑ i, w i) ≤ 139 * (∑ i, w i * s i) := by omega omega
The explicit candidate triples, challengers, member enumerators, and transitions satisfy all structural conditions and the obstruction certificate exists.
We instantiate the existential quantifiers with the explicit tables (`in339_s`, `in339_f`, `in339_d`, `in339_next`) and supply the verified properties `in339_family`, `in339_complete`, `in339_transition`, and `in339_certificate`.
uses
in339_family,in339_complete,in339_transition,in339_certificaterefine ⟨in339_s, in339_f, in339_d, in339_next, in339_family, in339_complete, in339_transition, ?_, in339_certificate⟩Every stationary weighting $w$ satisfies $\\sum_i w_i \\le 2 \\sum_i (\\text{if } \\forall b \\in S_i, f_i < b \\text{ then } w_i \\text{ else } 0)$.
Given any stationary weighting $w$, applying the general bound `in339_bound` with `in339_local` yields $\\sum w_i \\le 2 \\sum w_i \\cdot \\text{score}(i)$. Simplifying with the definition of `in339_score` rewrites the product $w_i \\cdot \\text{score}(i)$ as an indicator on whether $f_i < b$ for all $b \\in S_i$.
uses
in339_bound,in339_local,mul_ite,mul_one,mul_zerointro w hstat have hb := in339_bound in339_next in339_p in339_score in339_local w hstat simpa only [in339_score, mul_ite, mul_one, mul_zero] using hb
Finset.mul_sum- Factoring a scalar into a finite sum: `a * ∑ i, f i = ∑ i, a * f i`.
Nat.mul_assoc- Associativity of natural multiplication: `(n * m) * k = n * (m * k)`.
sum_congr- Congruence of finite sums: equal index sets and pointwise equal summands produce equal sums.
Finset.sum_mul- Distributing multiplication by a scalar on the right over a finite sum: `(∑ i, f i) * a = ∑ i, f i * a`.
Finset.sum_comm- Fubini-type interchange of finite sums: `∑ x, ∑ y, f x y = ∑ y, ∑ x, f x y`.
ite_mul- Distributing multiplication on the right into a conditional: `(if P then a else b) * c = if P then a * c else b * c`.
zero_mul- Zero multiplied by anything is zero: `0 * a = 0`.
Finset.sum_ite_eq- Evaluating an indicator sum against an equality condition: `∑ x ∈ s, (if a = x then b x else 0)` collapses to `b a` when `a ∈ s`.
Finset.mem_univ- Every element of a finite type belongs to its universal set `univ`.
if_true- A conditional on a true proposition simplifies to the then-branch: `(if True then t else e) = t`.
Finset.sum_le_sum- Monotonicity of finite sums: if `f i ≤ g i` for all `i`, then `∑ i, f i ≤ ∑ i, g i`.
Nat.mul_le_mul_left- Multiplying both sides of an inequality of natural numbers by a factor preserves the inequality: `n ≤ m → k * n ≤ k * m`.
sum_add_distrib- Splitting a sum of sums: `∑ x, (f x + g x) = ∑ x, f x + ∑ x, g x`.
mul_sum- Factoring a scalar into a finite sum: `a * ∑ i, f i = ∑ i, a * f i` (unqualified form).
in339_family- For all 56 states, the candidate triple has 3 elements, the challenger is outside the triple, and the index map is a bijection onto the triple.
in339_complete- Every 3-element subset of 8 candidates appears as some triple in the family.
in339_transition- Every transition replaces the specified member of the triple with the outside challenger.
in339_certificate- An explicit 3-element weight vector and subset indicator potential satisfy all local removal inequalities while having `2 * H univ < ∑ w i`.
in339_bound- Any stationary weighting on a Markov chain with a local potential inequality satisfies `∑ w i ≤ 2 * ∑ w i * s i`.
in339_local- The explicit potential `p` satisfies the local inequality `3 * p i + 70 ≤ ∑ k, p (next i k) + 139 * score i` at every state.
mul_ite- Distributing multiplication on the left into a conditional: `a * (if P then b else c) = if P then a * b else a * c`.
mul_one- Multiplying by one leaves a value unchanged: `a * 1 = a`.
mul_zero- Multiplying by zero on the right yields zero: `a * 0 = 0`.
- The local potential inequality across all 56 triples, the structural properties of the candidate triples, the transition correctness, and the 256 subset removal inequalities of the certificate are all verified by exhaustive kernel computation (`decide +kernel`).
- The final deduction in `in339_bound` deriving `∑ w i ≤ 2 * ∑ w i * s i` from `70 * ∑ w i ≤ 139 * ∑ w i * s i` is automated by linear integer arithmetic (`omega`).
Written by a model (gemini-3.8-flash) from the Lean proof and the goal states Lean reported at each step. Lean verified the proof; this explanation is a reading aid and is not itself machine-checked.
Lean source view module on GitHub
Choose a tactic — a marked line — to see the goals Lean had before and after it, as in Lean’s infoview. Step with ← → or j k.
open Finsetopen scoped BigOperatorsset_option maxHeartbeats 8000000set_option maxRecDepth 100000private def in339_data_s : Array (Finset (Fin 8)) := #[{0, 1, 2}, {0, 1, 3}, {0, 1, 4}, {0, 1, 5}, {0, 1, 6}, {0, 1, 7}, {0, 2, 3}, {0, 2, 4}, {0, 2, 5}, {0, 2, 6}, {0, 2, 7}, {0, 3, 4}, {0, 3, 5}, {0, 3, 6}, {0, 3, 7}, {0, 4, 5}, {0, 4, 6}, {0, 4, 7}, {0, 5, 6}, {0, 5, 7}, {0, 6, 7}, {1, 2, 3}, {1, 2, 4}, {1, 2, 5}, {1, 2, 6}, {1, 2, 7}, {1, 3, 4}, {1, 3, 5}, {1, 3, 6}, {1, 3, 7}, {1, 4, 5}, {1, 4, 6}, {1, 4, 7}, {1, 5, 6}, {1, 5, 7}, {1, 6, 7}, {2, 3, 4}, {2, 3, 5}, {2, 3, 6}, {2, 3, 7}, {2, 4, 5}, {2, 4, 6}, {2, 4, 7}, {2, 5, 6}, {2, 5, 7}, {2, 6, 7}, {3, 4, 5}, {3, 4, 6}, {3, 4, 7}, {3, 5, 6}, {3, 5, 7}, {3, 6, 7}, {4, 5, 6}, {4, 5, 7}, {4, 6, 7}, {5, 6, 7}]private def in339_s : Fin 56 → (Finset (Fin 8)) := fun i => in339_data_s[i.val]!private def in339_data_f : Array (Fin 8) := #[7, 7, 7, 7, 7, 6, 7, 7, 7, 7, 6, 7, 7, 7, 6, 7, 7, 6, 7, 6, 5, 0, 0, 0, 0, 6, 0, 0, 0, 6, 0, 0, 6, 0, 6, 5, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 3, 3, 3, 4]private def in339_f : Fin 56 → (Fin 8) := fun i => in339_data_f[i.val]!private def in339_data_d : Array (Fin 3 → Fin 8) := #[![0, 1, 2], ![0, 1, 3], ![0, 1, 4], ![0, 1, 5], ![0, 1, 6], ![0, 1, 7], ![0, 2, 3], ![0, 2, 4], ![0, 2, 5], ![0, 2, 6], ![0, 2, 7], ![0, 3, 4], ![0, 3, 5], ![0, 3, 6], ![0, 3, 7], ![0, 4, 5], ![0, 4, 6], ![0, 4, 7], ![0, 5, 6], ![0, 5, 7], ![0, 6, 7], ![1, 2, 3], ![1, 2, 4], ![1, 2, 5], ![1, 2, 6], ![1, 2, 7], ![1, 3, 4], ![1, 3, 5], ![1, 3, 6], ![1, 3, 7], ![1, 4, 5], ![1, 4, 6], ![1, 4, 7], ![1, 5, 6], ![1, 5, 7], ![1, 6, 7], ![2, 3, 4], ![2, 3, 5], ![2, 3, 6], ![2, 3, 7], ![2, 4, 5], ![2, 4, 6], ![2, 4, 7], ![2, 5, 6], ![2, 5, 7], ![2, 6, 7], ![3, 4, 5], ![3, 4, 6], ![3, 4, 7], ![3, 5, 6], ![3, 5, 7], ![3, 6, 7], ![4, 5, 6], ![4, 5, 7], ![4, 6, 7], ![5, 6, 7]]private def in339_d : Fin 56 → (Fin 3 → Fin 8) := fun i => in339_data_d[i.val]!private def in339_data_next : Array (Fin 3 → Fin 56) := #[![25, 10, 5], ![29, 14, 5], ![32, 17, 5], ![34, 19, 5], ![35, 20, 5], ![35, 20, 4], ![39, 14, 10], ![42, 17, 10], ![44, 19, 10], ![45, 20, 10], ![45, 20, 9], ![48, 17, 14], ![50, 19, 14], ![51, 20, 14], ![51, 20, 13], ![53, 19, 17], ![54, 20, 17], ![54, 20, 16], ![55, 20, 19], ![55, 20, 18], ![55, 19, 18], ![6, 1, 0], ![7, 2, 0], ![8, 3, 0], ![9, 4, 0], ![45, 35, 24], ![11, 2, 1], ![12, 3, 1], ![13, 4, 1], ![51, 35, 28], ![15, 3, 2], ![16, 4, 2], ![54, 35, 31], ![18, 4, 3], ![55, 35, 33], ![55, 34, 33], ![26, 22, 21], ![27, 23, 21], ![28, 24, 21], ![29, 25, 21], ![30, 23, 22], ![31, 24, 22], ![32, 25, 22], ![33, 24, 23], ![34, 25, 23], ![35, 25, 24], ![40, 37, 36], ![41, 38, 36], ![42, 39, 36], ![43, 38, 37], ![44, 39, 37], ![45, 39, 38], ![49, 47, 46], ![50, 48, 46], ![51, 48, 47], ![54, 53, 52]]private def in339_next : Fin 56 → (Fin 3 → Fin 56) := fun i => in339_data_next[i.val]!private def in339_data_p : Array (ℕ) := #[0, 6, 12, 20, 20, 20, 13, 18, 24, 24, 24, 28, 33, 33, 33, 44, 44, 44, 55, 55, 55, 29, 33, 37, 37, 28, 38, 42, 42, 36, 48, 48, 45, 54, 55, 55, 56, 59, 59, 53, 62, 62, 58, 65, 63, 63, 82, 82, 78, 84, 81, 81, 105, 103, 103, 126]private def in339_p : Fin 56 → (ℕ) := fun i => in339_data_p[i.val]!private def in339_score : Fin 56 → ℕ := fun i => if ∀ b ∈ in339_s i, in339_f i < b then 1 else 0private theorem in339_local : ∀ i, 3 * in339_p i + 70 ≤ (∑ k : Fin 3, in339_p (in339_next i k)) + 139 * in339_score i := by decide +kernelprivate theorem in339_family : ∀ i, (in339_s i).card = 3 ∧ in339_f i ∉ in339_s i ∧ Function.Injective (in339_d i) ∧ in339_s i = univ.image (in339_d i) := by decide +kernelprivate theorem in339_complete : ∀ T : Finset (Fin 8), T.card = 3 → ∃ i, in339_s i = T := by decide +kernelprivate theorem in339_transition : ∀ i j, in339_s (in339_next i j) = insert (in339_f i) ((in339_s i).erase (in339_d i j)) := by decide +kernelprivate theorem in339_certificate : ∃ (w : Fin 56 → ℕ) (H : Finset (Fin 8) → ℕ), (∀ A : Finset (Fin 8), ∀ a ∈ A, (∑ i, if in339_f i = a ∧ in339_s i ⊆ A.erase a then w i else 0) + H (A.erase a) ≤ H A) ∧ 2 * H univ < ∑ i, w i := by let w : Fin 56 → ℕ := fun i => if i = 18 ∨ i = 19 ∨ i = 20 then 1 else 0 let H : Finset (Fin 8) → ℕ := fun A => if ({5, 6, 7} : Finset (Fin 8)) ⊆ A then 1 else 0 refine ⟨w, H, ?_, ?_⟩ · decide +kernel · decide +kernelprivate theorem in339_bound (next : Fin 56 → Fin 3 → Fin 56) (p s : Fin 56 → ℕ) (hlocal : ∀ i, 3 * p i + 70 ≤ (∑ k : Fin 3, p (next i k)) + 139 * s i) (w : Fin 56 → ℕ) (hstat : ∀ j, 3 * w j = ∑ i, ∑ k : Fin 3, if next i k = j then w i else 0) : (∑ i, w i) ≤ 2 * (∑ i, w i * s i) := by have hbalance : 3 * (∑ i, w i * p i) = ∑ i, w i * (∑ k : Fin 3, p (next i k)) := by calc 3 * (∑ i, w i * p i) = ∑ j, (3 * w j) * p j := by simp only [Finset.mul_sum, Nat.mul_assoc] _ = ∑ j, (∑ i, ∑ k : Fin 3, if next i k = j then w i else 0) * p j := by apply sum_congr rfl intro j _ rw [hstat j] _ = ∑ i, w i * (∑ k : Fin 3, p (next i k)) := by simp only [Finset.sum_mul] rw [Finset.sum_comm] apply sum_congr rfl intro i _ rw [Finset.sum_comm] simp only [ite_mul, zero_mul] simp only [Finset.sum_ite_eq, Finset.mem_univ, if_true] rw [Finset.mul_sum] have hsum := Finset.sum_le_sum (fun i (_ : i ∈ (univ : Finset (Fin 56))) => Nat.mul_le_mul_left (w i) (hlocal i)) have hleft : (∑ i, w i * (3 * p i + 70)) = 3 * (∑ i, w i * p i) + 70 * (∑ i, w i) := by calc _ = ∑ i, (3 * (w i * p i) + 70 * w i) := by apply sum_congr rfl intro i _ ring _ = _ := by rw [sum_add_distrib, ← mul_sum, ← mul_sum] have hright : (∑ i, w i * ((∑ k : Fin 3, p (next i k)) + 139 * s i)) = (∑ i, w i * (∑ k : Fin 3, p (next i k))) + 139 * (∑ i, w i * s i) := by calc _ = ∑ i, (w i * (∑ k : Fin 3, p (next i k)) + 139 * (w i * s i)) := by apply sum_congr rfl intro i _ ring _ = _ := by rw [sum_add_distrib, ← mul_sum] rw [hleft, hright] at hsum have hbound : 70 * (∑ i, w i) ≤ 139 * (∑ i, w i * s i) := by omega omegatheorem condorcet_stationary_weight_restriction_obstruction : ∃ (S : Fin 56 → Finset (Fin 8)) (f : Fin 56 → Fin 8) (d : Fin 56 → Fin 3 → Fin 8) (next : Fin 56 → Fin 3 → Fin 56), (∀ i, (S i).card = 3 ∧ f i ∉ S i ∧ Function.Injective (d i) ∧ S i = univ.image (d i)) ∧ (∀ T : Finset (Fin 8), T.card = 3 → ∃ i, S i = T) ∧ (∀ i j, S (next i j) = insert (f i) ((S i).erase (d i j))) ∧ (∀ w : Fin 56 → ℕ, (∀ j, 3 * w j = ∑ i, ∑ k : Fin 3, if next i k = j then w i else 0) → (∑ i, w i) ≤ 2 * ∑ i, if ∀ b ∈ S i, f i < b then w i else 0) ∧ (∃ (w : Fin 56 → ℕ) (H : Finset (Fin 8) → ℕ), (∀ A : Finset (Fin 8), ∀ a ∈ A, (∑ i, if f i = a ∧ S i ⊆ A.erase a then w i else 0) + H (A.erase a) ≤ H A) ∧ 2 * H univ < ∑ i, w i) := by refine ⟨in339_s, in339_f, in339_d, in339_next, in339_family, in339_complete, in339_transition, ?_, in339_certificate⟩ intro w hstat have hb := in339_bound in339_next in339_p in339_score in339_local w hstat simpa only [in339_score, mul_ite, mul_one, mul_zero] using hb