{"id":"0af03d29-7274-478c-a46f-e25514cfcd72","arxiv_id":"2501.14640","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Impartial chess movesets on Young diagrams yield new P-position descriptions, Sprague-Grundy formulas, and CGH classifications, including the first tame-but-not-miserable family and the first game in the tame-returnable-not-forced region.","lead":"The paper studies chess-piece games played on Young diagrams, which are block shapes that encode integer partitions, and works out which positions are winning, losing, or have particular Sprague-Grundy values. It also places these games into the Conway-Gurvich-Ho classification scheme, including new infinite families in regions that previously had no known examples.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 4.12's proof uses a subpartition equality that is false as written, leaving the headline (R,S1) classification unverified.","rationale":"The Reader's CONDITIONAL verdict is well supported: the paper contains no machine-checked proofs, and several classification claims are delegated to 'the others are similar' or omitted proofs. In reading the worked parts, I found no counterexample to Theorem 3.16 or to the Grundy formulas on generalized staircases; the P-position description for Rook appears internally consistent. The sharper issue is in Theorem 4.12, which the Reader lists among the strongest claims. Its proof displays (R,⟨ℓ+1,ℓ^k⟩)[k−ℓ,ℓ] as equal to the rectangle ⟨ℓ^{ℓ+1}⟩, but under the paper's own subpartition definition this displayed subpartition is empty for all k>ℓ≥3. The intended index is almost certainly [k−ℓ,0], since that does produce the rectangle. This is not a mere stylistic gap: the proof's central structural assertion about swap positions and (t,t)-positions is obtained from that rectangle, so as written the argument for 'forced and tame but not miserable' does not go through. I am not claiming the theorem is false; a small computational check on Young diagrams of bounded size would confirm whether the corrected index yields the claimed Conway pairs. This supports keeping the verdict CONDITIONAL rather than ACCEPT or REJECT. My agreement with the Reader is partial because the Reader emphasized omitted 'similar' cases, whereas I see an explicit indexing error in a fully written but incorrect-looking proof of one of the headline results.","tokens_in":22638,"tokens_out":17613,"duration_ms":151443,"concrete_test":"Implement the recursive definitions of SG and the misère Grundy value G− for Rook positions, and exhaustively compute Conway pairs on all subpositions of λ=⟨ℓ+1,ℓ^k⟩ for ℓ=3,k=5 and ℓ=4,k=6. Check: (i) whether [k−ℓ,ℓ] is empty; (ii) whether, with the corrected index [k−ℓ,0], every subposition has Conway pair in {(0,1),(1,0)}∪{(t,t):t≥0}, with the only nonterminal swap positions inside the 2×2 block at [k−1,ℓ−2]; (iii) whether from the root there is a move to a (0,1)-position but no move to a (1,0)-position. If all three hold, the theorem survives as a one-character index correction; if any (t,0)- or (0,t)-position with t≥2 appears, the family is not tame and the headline claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Under Definition 2.1 and the subpartition convention λ[i,j]=⟨λ_{i+1}−j,λ_{i+2}−j,…⟩ with trailing nonpositive entries removed, the subpartition displayed in Theorem 4.12, (R,⟨ℓ+1,ℓ^k⟩)[k−ℓ,ℓ], is empty for every k>ℓ≥3: after deleting k−ℓ rows and ℓ columns, each remaining entry is ℓ−ℓ=0. The proof then identifies it with the rectangle ⟨ℓ^{ℓ+1}⟩=ℓ+1,ℓ. Using [k−ℓ,0] would give that rectangle, but as written the equality is false. This matters because Theorem 4.12 is one of the two headline claims placing an infinite family in a previously empty CGH region; its proof's assertion that all non-swap positions are (t,t)-positions is anchored on this rectangle's internal structure, so the region claim is not established by the written argument. Theorem 4.13's proof is similarly sketchy about the local structure of [4^7], and Theorem 4.11 is omitted entirely, so the full classification table in Figure 7 and Table 1 remains conditional even if the indexing error is a simple typo.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript extends Berlekamp's impartial chess games from rectangular boards to Young diagrams of integer partitions. It defines chess-piece movesets on partitions, proves Sprague-Grundy results for King, gives a P-position characterization for Rook (Theorem 3.16), and provides a Conway-Gurvich-Ho (CGH) classification for restrictions to rectangles, staircases, and generalized staircases. The headline classification results are Theorem 4.12, placing (R,S1) as forced and tame but not miserable, and Theorem 4.13, placing (Q,S2) as tame but not miserable and returnable but not forced, together with Theorem 4.11 for Queen on rectangles and generalized staircases. The paper claims these occupy previously unknown or sparsely occupied CGH regions.","tokens_in":22863,"tokens_out":6267,"duration_ms":49299,"significance":"If the classification results are correct, the paper makes a solid contribution to combinatorial game theory: it connects impartial chess to partition theory, provides a complete P-position characterization for Rook on all Young diagrams (Theorem 3.16), generalizes Nim and Wythoff equivalences, and supplies new infinite families in CGH regions that were previously empty or nearly empty. The self-contained proofs of Theorem 3.16, Lemma 3.23, and the King lemmas are coherent and appear correct. The main caveat is that the CGH table depends on several proofs that are omitted or compressed (Theorem 4.11) or contain an indexing error (Theorem 4.12), so the headline classifications are not yet fully verified.","major_comments":[{"comment":"In the proof of Theorem 4.12, the displayed subpartition identity (⟨ℓ+1,ℓ^k⟩)[k−ℓ,ℓ] = ⟨ℓ^{ℓ+1}⟩ is false: by Definition 2.1, subtracting ℓ from each of the remaining ℓ+1 rows gives all entries 0, so the subpartition is empty for every k>ℓ≥3. The subsequent structural assertions about this subposition (swap positions in the last two columns, and (0,0)- and (1,1)-positions elsewhere) depend on the subpartition being the rectangle ℓ+1,ℓ, which would correspond to the index [k−ℓ,0]. Because this identification is the anchor for showing that all non-swap positions are (t,t)-positions, Theorem 4.12 is not established by the written argument. The authors should correct the index or supply a valid subpartition identity and re-verify the tameness claim.","section":"Theorem 4.12"},{"comment":"Theorem 4.11, which classifies (Q, rectangles) and (Q, generalized staircases) as miserable but not pet and returnable but not forced, is stated with its proof omitted 'in the interest of brevity.' This result is load-bearing for the completeness of Figure 7 and Table 1, and the preceding paragraph indicates that the rectangle case requires separate treatment for r=c=2, r=2,c=3, and r,c≥3. Without a proof or a reference to a complete proof, the CGH classification is conditional. Please provide the full proof or move the statement to a conjecture section.","section":"Theorem 4.11"},{"comment":"The proof of Theorem 4.13 asserts, without derivation, that (Q,⟨4^7⟩) contains only swap and (t,t)-positions with a specific column structure, and that (Q,⟨5,4^k⟩[0,4]) is a (0,1)-position. These assertions are the basis for concluding that every position in the family is (t,t) for t≥2, and hence for the tameness claim. Because (Q,S2) is one of the two headline new-region claims, the local computation should be shown or replaced by a reproducible argument (e.g., a small table of Conway pairs for ⟨4^7⟩).","section":"Theorem 4.13"}],"minor_comments":[{"comment":"In the definition of Queen, the moveset is written as Q = (1,1)^+ ∪ (1,0)^+ ∪ (1,1)^+, with (1,1)^+ repeated and (0,1)^+ missing; the later text in Section 3.6 gives the correct set Q = (1,0)^+ ∪ (0,1)^+ ∪ (1,1)^+. Please correct the definition.","section":"Definition 3.1"},{"comment":"Theorem 4.9 states that p, (N, staircases), (K, staircases), (p, generalized staircases), and (D, generalized staircases) are pet and returnable but not forced, but the proof only treats (D, generalized staircases) and says 'the others are similar.' Since the movesets and the witness partitions differ for each family, a few more details or a general criterion would improve verifiability.","section":"Theorem 4.9"},{"comment":"The proof of Proposition 3.7 asserts that the relevant disjunctive sums are in P without spelling out the mirroring strategy; a sentence describing the pairing of moves would make the argument easier to check.","section":"Proposition 3.7"},{"comment":"In the proof of Theorem 4.10, the sentence 'in addition they may they include both swap positions' contains a typo, and the maximality argument for the generalized-staircase extension is compressed; please clarify.","section":"Theorem 4.10"}],"recommendation":"major_revision","confidential_remarks":"The paper contains several genuinely interesting results, and the core ideas are sound, but the CGH classification table is not yet fully supported: one proof has a false indexing equality, another is omitted, and a third is too sketchy on a key local computation. These are fixable within the manuscript's scope, but they are load-bearing for the paper's main claims. I would also encourage the authors to check whether the claimed 'first infinite families' in the tame-but-not-miserable region are consistent with the literature beyond [5], since the survey in that paper may have been updated."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First, the substance: the rook P-position theorem (Theorem 3.16) and the SG formulas for rook on generalized staircases (Proposition 3.19/Corollary 3.20) are genuinely new and, as far as I checked, correct. The King lemmas and the reductions of Pawn/Knight to Downright are also clean and self-contained. This part of the paper is worth having.\n\nThe CGH classification is the flashier contribution, and here the paper is not finished. Theorem 4.11, which puts Queen on rectangles and generalized staircases into the miserable-but-not-pet region, has its proof omitted 'in the interest of brevity.' For a classification table that advertises previously empty regions, that is a load-bearing omission, not a brevity matter. Theorems 4.8–4.10 repeatedly say 'the others are similar' after proving one representative; I can believe that, but a referee can't verify it from the preprint.\n\nThere is also a concrete error in the proof of Theorem 4.12. The subpartition (R, ⟨ℓ+1, ℓ^k⟩)[k−ℓ, ℓ] is empty for every k > ℓ ≥ 3 under the paper's own Definition 2.1: each entry is ℓ−ℓ = 0 after removing trailing nonpositive parts. The intended index is [k−ℓ, 0], which does give the rectangle ⟨ℓ^{ℓ+1}⟩. So this is likely a typo, but as written the proof asserts a false equality, and the argument that the region (R,S1) is forced-and-tame rests on that rectangle's internal structure. Fixing the index may repair the proof, but the submitted text doesn't establish the claim.\n\nNone of this kills the central rook result. The weakness is concentrated in the misère/CGH section, where the authors ask the reader to take a lot on faith. The significance of the new CGH occupants is real if the proofs are supplied. I'd send it to a serious referee — the main theorems are important enough — but the referee instructions should ask for a complete proof of Theorem 4.11, full casework for the 'similar' theorems, and a correction to the subscript. As it stands, the paper is a strong preprint with an unfinished final act.","headline":"Rook P-positions and SG formulas are real progress; the CGH classification table still needs fuller proofs — and one headline proof has a subscript typo.","tokens_in":23428,"tokens_out":3919,"would_cite":true,"duration_ms":38527,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["91A46","05A17"],"pacs":[],"model":"deepseek-v4-flash","headline":"Playing impartial chess on Young diagrams yields a complete losing-position rule for Rook and places Rook and Queen games in two previously empty CGH classification regions.","keywords":["impartial combinatorial game","Young diagram","integer partition","Sprague-Grundy value","misère play","Conway-Gurvich-Ho classification","Rook game","Queen game"],"falsifier":"Compute the Conway pairs of (Q, k_{r,c}) for a small generalized staircase not covered by the omitted proof of Theorem 4.11, such as (Q, 2_{3,3}) or (Q, 3_{2,2}), by direct recursive mex calculation; if any subposition is non-swap and has a move to exactly one of the two swap types, the claim that (Q, generalized staircases) is miserable fails.","tokens_in":22431,"feed_emoji":"♟️","tokens_out":8417,"duration_ms":70710,"temperature":0.7,"pith_summary":"This paper extends Berlekamp's impartial chess games from rectangular boards to Young diagrams of integer partitions. Its central result is a complete description of the losing positions of Rook: a partition λ is a P-position if and only if λ has Dyson rank 0 and the subpartition λ[1,1] has at least r−1 cells in each part, where r is the number of parts. The authors also compute Sprague-Grundy values for King and Rook on special partition families, show that Pawn and Knight reduce to the known game Downright, and prove that misère play is equivalent to normal play on a truncated diagram. Using the CGH classification of normal/misère interplay, they classify all these games and exhibit the first infinite families in two previously unoccupied regions, generalizing Nim and Wythoff, which arise as Rook and Queen on rectangles.","feed_headline":"Rook and Queen games fill two empty regions of impartial-game taxonomy","feed_subtitle":"A complete rule for Rook's losing positions and two new classes of misère-play games on integer partitions.","key_machinery":"The central machinery is the moveset formulation of impartial chess on Young diagrams, where a position is a partition λ and a move deletes a subpartition λ[i,j]. The proof structure relies on three pillars: (1) the directed acyclic graph DAG_M(λ) of a position, which supports partition-equivalence and game-equivalence relations; (2) Sprague-Grundy values and misère Grundy values, combined into Conway pairs (SG, G^−) that drive the CGH classification; and (3) the truncation operation, whose key property is that misère P-positions of (M, λ) coincide with normal-play P-positions of the truncated position. For Rook, the P-position proof introduces 'ample' partitions—partitions whose rows and columns each contain a P-position—and shows that a position is losing exactly when it has rank 0, equal top two parts, and an ample interior.","core_discovery":"Theorem 3.16 gives the complete normal-play P-position classification for Rook: (R, λ) is a losing position if and only if λ has rank 0 and λ[1,1] ≥ r−1, where r is the number of parts of λ. The CGH theorems place the restrictions (R, S1) and (Q, S2) into the regions 'forced and tame but not miserable' and 'tame but not miserable and returnable but not forced', respectively, with S1 = {[ℓ+1, ℓ^k][i,j] : ℓ ≥ 3, k > ℓ} and S2 = {[5, 4^k][i,j] : k ≥ 7}; the latter region was previously occupied only by a single small artificial game. These are the paper's most consequential assertions; together with Theorems 4.8–4.11 and the Rook characterization, they yield the full CGH table for these families.","pith_inferences":["If the omitted 'similar-case' proofs in Theorems 4.8–4.11 hold, the CGH regions containing Nim, Wythoff, and Downright now have natural infinite families from chess, suggesting the taxonomy is not artificially sparse.","The golden-ratio coefficient that governs Queen's P-position density on rectangles might extend to all partitions; the paper's Lemma 3.22 gives a factor-2 bound for general partitions, and finding an infinite family with density approaching the golden-ratio bound would settle whether the constant can be improved.","The Rook characterization invites an analogous search for Queen's P-positions on Young diagrams; the paper argues this is likely hard because even the rectangular case is complicated.","A natural testable extension is to compute CGH classifications for thick-hook partitions, which the authors identify as a bridge between rectangles and generalized staircases."],"forward_implications":["Rook on a rectangle is game-equivalent to 2-pile Nim and Queen on a rectangle to Wythoff, so all results on Young diagrams strictly generalize these classical games.","The P-position rule for Rook gives a direct test for whether any Young-diagram position is a first-player loss, computable from the partition's rank and a single subpartition.","The game-tree equivalences mean Pawn and Knight positions can be solved by mapping to Downright via the explicit transformations φ_p and φ_N.","Misère versions of Downright, King, Rook, and Queen are normal play on the corner-removed partition λ^−, so misère play is exactly as tractable as normal play for those pieces.","The classification table shows Bishop, Pawn on rectangles, Rook on rectangles and generalized staircases, Queen on generalized staircases, and Knight on staircases are pet and forced, while other combinations occupy distinct known or new CGH regions."],"supporting_citations":[{"why":"Supplies the Downright game, the truncation operation, and the misère reduction that Section 4.1 builds on.","marker":"[3]"},{"why":"Introduces misère Grundy values, Conway pairs, and tame games, forming the basis of the CGH classification.","marker":"[4]"},{"why":"Extends the classification with pet, domestic, and returnable games, and classifies Wythoff and Subtraction, the baseline for the new regions.","marker":"[5]"},{"why":"Source of Berlekamp's impartial chess games on rectangular boards that this paper generalizes to Young diagrams.","marker":"[11]"},{"why":"Provides the King-on-rectangle Sprague-Grundy formula restated as Theorem 3.8.","marker":"[12]"},{"why":"Defines the Wythoff game, used to equate Queen on rectangles with Wythoff and to cite the golden-ratio bound.","marker":"[13]"}],"fun_headline_variants":["Rook and Queen games fill gaps in impartial-game taxonomy","Chess on Young diagrams: Rook's losing positions fully classified","Generalizing Nim and Wythoff to chess on integer partitions","New CGH regions from chess on partitions: Rook and Queen"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The completeness of the classification table depends on several proofs that are stated but not written: Theorems 4.8 and 4.9 claim 'the others are similar', and Theorem 4.11 omits its proof 'in the interest of brevity'; if any of those unshown cases contains a Conway pair outside the listed sets, the table in Figure 7 and Table 1 is wrong.","fun_headline_variants_meta":{"raw":{"variants":["Rook and Queen games fill gaps in impartial-game taxonomy","Chess on Young diagrams: Rook's losing positions fully classified","Generalizing Nim and Wythoff to chess on integer partitions","New CGH regions from chess on partitions: Rook and Queen"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000264,"raw_usage":{"total_tokens":1564,"prompt_tokens":869,"completion_tokens":695,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":485,"completion_tokens_details":{"reasoning_tokens":623}},"tokens_in":485,"tokens_out":695,"duration_ms":7547,"temperature":1.0,"reasoning_tokens":623,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T14:56:49.590332+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the Conway pairs of (Q, k_{r,c}) for a small generalized staircase not covered by the omitted proof of Theorem 4.11, such as (Q, 2_{3,3}) or (Q, 3_{2,2}), by direct recursive mex calculation; if any subposition is non-swap and has a move to exactly one of the two swap types, the claim that (Q, generalized staircases) is miserable fails.","supporting_citations":[{"cited_title":"Gottlieb, M","cited_arxiv_id":null,"evidence_quote":"Supplies the Downright game, the truncation operation, and the misère reduction that Section 4.1 builds on."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces misère Grundy values, Conway pairs, and tame games, forming the basis of the CGH classification."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Extends the classification with pet, domestic, and returnable games, and classifies Wythoff and Subtraction, the baseline for the new regions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Source of Berlekamp's impartial chess games on rectangular boards that this paper generalizes to Young diagrams."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the King-on-rectangle Sprague-Grundy formula restated as Theorem 3.8."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the Wythoff game, used to equate Queen on rectangles with Wythoff and to cite the golden-ratio bound."}],"review_version":1}