{"id":"a016453f-3ee1-4f62-951f-619e49133842","arxiv_id":"2608.07690","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"An imbalanced dealer flow is exactly equivalent to a balanced flow with a shifted skew, a widened spread, and a multiplied holding cost.","lead":"A market maker facing more sellers than buyers (or the reverse) should shift her quoted midpoint in the direction of the flow and slightly widen her quotes, and both adjustments follow from a simple symmetry in the underlying math. The paper shows that any imbalanced customer flow is exactly equivalent to a balanced flow with three known corrections, which makes the result useful for pricing, backtesting, and machine learning in dealer markets.","discovery_kind":"unification","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1 needs an explicit interiority condition: when the max(0,·) in Eq. (1) binds (K < -1/h), Eq. (3) and the compression into the balanced problem fail, so the 'exactly' claim is unproven as stated.","rationale":"I read the paper's central claim as an exact symmetry theorem, and the algebraic identity (4) is correct; the consistency equation (3) is also the right Bellman equation in the interior, exactly as the reader's verification indicates. The reader's weakest assumption, constant hazard at the quotes made, is the right empirical caveat. My stress test found a sharper, more internal gap: even with a perfectly exponential win curve, the theorem's proof silently assumes that the unconstrained optimum m = K + 1/h is feasible at every state with positive stationary mass. The paper acknowledges the max(0,·) boundary only parenthetically in Section 2.2 and never restricts Eq. (3) or the theorem accordingly. Because the transformation shifts the boundary threshold, the exact compression cannot hold at boundary states, so the theorem as written is incomplete. This does not overturn the economic intuition or the algebraic core; it should be fixed by an explicit interiority condition or a proof that the boundary is never visited. The reader's CONDITIONAL verdict therefore stands, with the condition sharpened to include non-binding boundary constraints, so I report no change to the verdict.","tokens_in":7130,"tokens_out":20544,"duration_ms":227584,"concrete_test":"Re-derive the consistency equation from the average-reward DP with the constraint m≥0, replacing h^{-1}e^{-1-hK} by sup_{m≥0}(m-K)e^{-hm} = max(-K, h^{-1}e^{-1-hK}), and test whether the mapping K → K - δ, C → C, c → M c leaves this equation invariant at states with K < -1/h. A numerical version: solve the DP on a finite lattice for c(x)=a x^2 with exponential F for q=1/2 and q≠1/2, compute the stationary distribution, and compare value and quote functions at boundary states; if the predicted δ/γ mapping fails there beyond numerical tolerance, the theorem requires an explicit interiority/non-binding-boundary assumption.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 1 rests on Eq. (3), which replaces the value of an enquiry G(K) = sup_{m≥0}(m-K)e^{-hm} by the unconstrained exponential form h^{-1}e^{-1-hK}. But Eq. (1) explicitly allows the optimal markup to be zero, and Section 2.2 notes 'away from the region where the max(·,0) binds.' When K < -1/h, the constrained optimum is m=0 and G(K) = -K under the model's exponential win curve on m≥0, so Eq. (3) is not valid at those states. This is not a measure-zero technicality: with a convex carrying cost, K(x) = ε + C(x) - S(x) tends to -∞ as x→+∞, and such lattice states carry positive stationary mass in the positive-recurrent inventory chain. Worse, Theorem 1's translation S → S - δ changes the location of the boundary K = -1/h, so the max operator does not commute with the compression; the imbalanced solution and the transformed balanced solution will differ at boundary states. The theorem therefore needs an explicit assumption that the max never binds on the support of the ergodic distribution, or a proof that it does not, before the 'same solution' claim is exact as stated.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a steady-state market-making model in which trade opportunities arrive as a Poisson process and each arriving customer is a seller with probability q and a buyer with probability 1−q. The dealer quotes a markdown or markup against a common fair price, and wins the trade if her quote beats the best competing response; the best competitor's displacement is assumed to be exponentially distributed with mean width w. The dealer's policy is characterized by an indifference liquidation cost ν(x), and the central result (Theorem 1) is an algebraic identity: the steady-state consistency equation for the imbalanced problem is equivalent to that of a balanced (q=1/2) problem after translating the skew by δ=(1/2h)log(q/(1−q)), widening the non-discretionary width Δ by γ=(1/h)log(1/(2√(q(1−q)))), and multiplying the carrying cost by M(q)=1/(2√(q(1−q))). Several corollaries follow: a flat book should still skew toward the flow, skew responds at first order in imbalance while width responds at second order, and the constant-width linear-skew heuristic is identified as the small-skew solution for a specific cost structure. The paper also sketches practical uses for RFQ markets, including filling-ratio diagnostics and transfer learning for reinforcement learning.","tokens_in":7446,"tokens_out":6231,"duration_ms":57208,"significance":"If the exactness claim holds, the paper provides a clean and useful symmetry result: one balanced solve supplies solutions for every imbalance level via explicit parameter adjustments that involve no fitted constants. The elementary identity (4) is verifiable, and the paper is honest about the historical provenance of the symmetrization and about the local nature of the exponential assumption. The result is potentially significant for OTC/RFQ market making, where one-sided flow is common, and the corollaries offer falsifiable cross-sectional predictions (first-order skew shift, second-order width response) that could be tested on RFQ archives. The main weaknesses are the compressed derivation of the consistency equation and an unaddressed boundary issue with the nonnegativity constraint on markups, both of which affect the exactness claim.","major_comments":[{"comment":"The proof of Theorem 1 uses the unconstrained exponential form of the enquiry value G(K) in Eq. (3), but Eq. (1) contains a max(0,·) operator. When K < -1/h, the optimal markup is m=0 and G(K) = -K, so Eq. (3) is not valid at those states. The paper does not state an interiority condition in Theorem 1 and does not prove that the constraint never binds on the support of the ergodic inventory distribution. For a convex carrying cost, K(x) can be negative for large |x|, and such states can carry positive stationary mass. Moreover, the skew translation S → S - δ shifts the location of the boundary K = -1/h, so the max operator does not commute with the compression. The theorem's 'same solution' claim is therefore unproven as stated; either add an explicit assumption that the max never binds on the visited inventory states, or extend the derivation to the constrained region.","section":"§2.2, Eq. (1); §2.3, Eq. (3); Theorem 1"},{"comment":"Eq. (3) is the load-bearing consistency equation, but the paper states it with only the sentence 'Comparing the two yields a functional equation for ν.' Given that Theorem 1 and all corollaries rest on this equation, the manuscript should provide a full derivation from the dealer's optimality (e.g., the average-reward Hamilton–Jacobi–Bellman equation or an indifference argument), or at least a precise statement of the normalization and boundary conditions. Without this, a reader cannot verify the exactness of the transformation claimed in Theorem 1.","section":"§2.3, Eq. (3)"},{"comment":"The paper claims in Remark 2 that the exponential assumption is needed only locally, with an error of the order of the hazard's relative variation, verified numerically in an external script. This is a substantive weakening of Theorem 1's hypothesis, but the error estimate is not derived and the numerical check is not described. If the theorem is to be advertised as exact, the conditions for exactness (global constant hazard, plus the interiority condition above) should be stated as assumptions; the local version should be clearly labeled as an approximation with a documented error analysis.","section":"Remark 2"}],"minor_comments":[{"comment":"'All trades are of sizes' appears to be a typo; it should read 'of size s'.","section":"§2, first paragraph"},{"comment":"The notation '1/h(m↑(x;s))' is easily misread; clarify that h is evaluated at the markup, e.g., write m = max(K + 1/h(m), 0).","section":"§2.2, Eq. (1)"},{"comment":"'Keywords:market making' is missing a space after the colon; the abstract would also benefit from defining q and w explicitly (the body defines them, but the abstract uses these symbols without introduction).","section":"Abstract and keywords"},{"comment":"The proof of Corollary 1 uses Sδ(0)=0 'by symmetry'; this should be justified (e.g., by uniqueness of the balanced solution or by an explicit parity argument) rather than asserted.","section":"Corollary 1"}],"recommendation":"major_revision","confidential_remarks":"The core idea is appealing and the algebra checks, but the missing derivation of Eq. (3) and the unhandled max constraint make the exactness claim incomplete. If the author can supply the derivation and either prove or explicitly assume interiority, the paper would be a solid contribution. The external script 'verify local exponentiality.py' should be included as supplementary material for the numerical claim in Remark 2."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper finds a real and useful symmetry in a standard request-for-quote market-making model: the steady state with imbalanced arrival flow compresses onto the balanced problem via a skew shift, a width widening, and a cost multiplier. That packaging is new, and it turns a practitioner regularity (flow-shading at flat inventory) into an explicit formula, delta = (w/2) log(q/(1-q)). The corollaries are clean: skew is first-order in imbalance, width second-order, and the constant-width/linear-skew heuristic reappears as a small-skew special case. The paper is honest about the algebraic lineage (birth-death symmetrization, refs [19-21]) and the literature review is careful about what is and is not new. I found no self-citation or curve-fitting issues; the predictions are parameter-free given an observable width, which is exactly the right kind of claim for a theory note.\n\nThe main soft spot is real and needs fixing. Eq. (3) replaces the enquiry value sup_m (m-K)e^{-hm} with the unconstrained exponential form, which is only valid when the optimal markup is nonnegative. The paper explicitly says \"away from the region where the max(.,0) binds\" in Section 2.2 but never carries that qualification into Theorem 1. The stress-test is right that this is not a measure-zero nuisance: for a convex carrying cost, the strike K drifts below -1/h at large inventory, those states carry positive stationary mass, and the translation S to S - delta shifts the boundary so the compression does not commute with the max operator. The theorem needs an explicit interiority assumption (or a proof that the max never binds on the support) before the \"exactly\" claim is true as written. This is fixable, but it is a genuine gap in the central result, not a stylistic nit.\n\nTwo smaller issues: Eq. (3) is stated without derivation (the reader says it follows from a standard average-reward HJB setup, and I agree, but a sketch would help), and the numerical verification script referenced in Remark 2 is not included, so that part is not independently checkable. Neither is serious.\n\nWho is this for? Anyone modeling OTC dealer markets or designing RFQ algorithms. The symmetry is a useful benchmark even if the boundary patch changes the theorem's fine print. The paper deserves a serious referee; my own verdict would be conditional revision, not reject. I would bring it to a reading group and cite it if I wrote on this topic.","headline":"A genuinely useful symmetry for RFQ market making, with an exactness claim that needs a boundary condition patch before it is true as stated.","tokens_in":787,"tokens_out":1133,"would_cite":true,"duration_ms":46468,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["91G15","93E20"],"pacs":[],"model":"deepseek-v4-flash","headline":"One symmetry absorbs all order imbalance in OTC quoting","keywords":["market making","order imbalance","skew","bid-ask spread","request for quote","inventory cost","over-the-counter trading","steady-state symmetry"],"falsifier":"Run the steady-state analysis with a non-exponential win curve, for instance a Weibull survival function with shape parameter different from 1, and compare the imbalanced solution to the balanced solution adjusted by $(\\delta,\\gamma,M(q))$; if they differ, exactness fails. A simpler empirical version is to estimate from a request-for-quote archive the local hazard of the best competing quote over the strikes a dealer actually visits; if it is not close to constant, the predicted zero-inventory skew $\\delta=(w/2)\\log(q/(1-q))$ should be off by roughly the hazard's relative variation.","tokens_in":1672,"feed_emoji":"⚖️","tokens_out":1762,"duration_ms":93934,"temperature":0.7,"pith_summary":"Market makers in over-the-counter markets face one-sided customer flow: sellers arrive more often than buyers, or the reverse. This paper claims that imbalance does not add a new problem. The steady-state optimal quoting policy with imbalance $q$ is exactly the balanced-flow policy after three mechanical adjustments: the skew is translated by a computed amount, the quote width is widened, and the cost of carrying inventory is multiplied by $M(q)=1/(2\\sqrt{q(1-q)})$. No free parameter enters beyond the observable market width $w$, so one numerical solution of the balanced problem serves every imbalance. This explains why a dealer with a flat book should still lean quotes into the flow, and why practitioners skew before they widen.","feed_headline":"One symmetry absorbs all order imbalance in OTC quoting","feed_subtitle":"Skew shifts with the log-odds of flow, quotes widen, carry cost multiplies—no free parameter beyond market width.","key_machinery":"The engine is the elementary identity $$q $e^{{-hS}}$+(1-q)$e^{{hS}}$=2\\sqrt{q(1-q)}\\$\\cosh$\\bigl(h(S-\\delta)\\bigr),\\qquad \\delta=\\frac{1}{2h}\\log\\frac{q}{1-q},$$ which rewrites an imbalanced mixture of two exponentials as a balanced $\\cosh$, translated by $\\delta$ and rescaled by $2\\sqrt{q(1-q)}$. Substituted into the steady-state consistency equation that equates the cost of carrying inventory with the option value of the next trading opportunity, it shows the imbalanced problem is the balanced problem with carrying cost multiplied by $M(q)=e^{h\\gamma}$. The economics comes from identifying $S$ as the slope of the inventory cost (the skew), $C$ as its convexity (the discretionary width), and $w=1/h$ as the reciprocal hazard of the best competing quote.","core_discovery":"The paper's central discovery is Theorem 1: the imbalanced market making problem compresses onto the perfectly balanced one. With arrival imbalance $q$ and market width $w$, the imbalanced steady state is the balanced state with carrying cost $M(q)c(\\cdot)$, skew shifted by $\\delta=(w/2)\\log(q/(1-q))$, and non-discretionary width $\\Delta$ widened to $\\Delta+\\gamma$, where $M(q)=1/(2\\sqrt{q(1-q)})$ and $\\gamma=w\\log(1/(2\\sqrt{q(1-q)}))$. Three corollaries follow: a zero-inventory dealer should quote mid shifted by $\\delta$; skew responds to imbalance at first order while width responds only at second order; and one-sided flow acts like an additional cost of carrying inventory. The popular constant-width, linear-skew policy is exactly optimal only in one corner, balanced flow with a $\\cosh$-shaped carrying cost, and the symmetry supplies the flow term that this benchmark omits.","pith_inferences":["If the local-exponential assumption holds only approximately in a real request-for-quote market, the same $(\\delta,\\gamma,M(q))$ correction should still be the leading-order adjustment, with the error controlled by the hazard's relative variation across the strikes a dealer visits; quote-archive data could measure that directly.","The identity (4) is a general two-state symmetrization, so the same compression may carry over to other asymmetric two-sided models, such as limit order book systems with unequal buy and sell arrival rates, wherever an analogous consistency equation appears.","A cross-sectional prediction follows from the theorem: across OTC instruments, zero-inventory skew should scale with $w\\log(q/(1-q))$ and the width correction with $w\\log\\cosh(\\delta/w)$, giving a no-free-parameter empirical check.","The decomposition reframes constant-width, linear-skew heuristics as missing not merely inventory-cost curvature but the entire flow term, so flow-aware heuristics should add a flow intercept before adding any nonlinearity."],"forward_implications":["A dealer with zero inventory should quote a midpoint shifted by $\\delta=(w/2)\\log(q/(1-q))$ away from fair value, with the sign set by the net buying or selling pressure.","Skew responds to imbalance at first order while quote width responds only at second order, so mild one-sidedness moves the mid almost linearly and barely changes the spread.","One-sided flow acts like an additional cost of carrying inventory, so even a product with zero holding cost effectively taxes a dealer's book when arrivals are imbalanced.","A single balanced steady-state solve, computed once, yields the imbalanced solution for every $q$ through the same three-part correction.","The constant-width, linear-skew heuristic is exact only for balanced flow with a cosh-shaped carrying cost; outside that corner, a quadratic inventory-cost ansatz is internally inconsistent."],"supporting_citations":[{"why":"Introduced the dealer model with genuinely asymmetric buy and sell rates, the oldest imbalanced-flow setup.","marker":"[1]"},{"why":"Established quoting from value-function differences, the slope half of the slope-and-convexity characterization used here.","marker":"[3]"},{"why":"Defined the constant-width, linear-skew benchmark that Section 4 locates and corrects.","marker":"[4]"},{"why":"Provided the exact symmetric-arrival market making problem that Theorem 1 extends to imbalance.","marker":"[5]"},{"why":"Supplied econometric support for the win-curve assumption the model makes.","marker":"[11]"},{"why":"Documented the sealed-bid request-for-quote mechanism the model formalizes.","marker":"[16]"},{"why":"Documented electronic FX dealers skewing on flow at flat inventory, the practice Corollary 1 derives.","marker":"[17]"},{"why":"Observed numerically that flow-aware market makers skew even without inventory, which Corollary 1 captures in closed form.","marker":"[18]"}],"fun_headline_variants":["Imbalance maps to balanced OTC quoting via one symmetry","One symmetry turns imbalanced OTC skew into balanced case","OTC imbalance: skew shifts, quotes widen, carry cost scales","Zero-inventory dealers should still skew in OTC markets","First-order skew, second-order width: OTC imbalance effect"],"cache_read_input_tokens":10112,"weakest_assumption_plain":"The exact equivalence assumes the best competing quote has a locally constant hazard rate, an exponential distribution, at the prices the dealer actually quotes; if that hazard varies materially over the strikes she visits, the translation, widening, and cost multiplier are only approximate.","fun_headline_variants_meta":{"raw":{"variants":["Imbalance maps to balanced OTC quoting via one symmetry","One symmetry turns imbalanced OTC skew into balanced case","OTC imbalance: skew shifts, quotes widen, carry cost scales","Zero-inventory dealers should still skew in OTC markets","First-order skew, second-order width: OTC imbalance effect"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000327,"raw_usage":{"total_tokens":1832,"prompt_tokens":955,"completion_tokens":877,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":571,"completion_tokens_details":{"reasoning_tokens":794}},"tokens_in":571,"tokens_out":877,"duration_ms":6923,"temperature":1.0,"reasoning_tokens":794,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T00:24:50.240287+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the steady-state analysis with a non-exponential win curve, for instance a Weibull survival function with shape parameter different from 1, and compare the imbalanced solution to the balanced solution adjusted by $(\\delta,\\gamma,M(q))$; if they differ, exactness fails. A simpler empirical version is to estimate from a request-for-quote archive the local hazard of the best competing quote over the strikes a dealer actually visits; if it is not close to constant, the predicted zero-inventory skew $\\delta=(w/2)\\log(q/(1-q))$ should be off by roughly the hazard's relative variation.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduced the dealer model with genuinely asymmetric buy and sell rates, the oldest imbalanced-flow setup."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Established quoting from value-function differences, the slope half of the slope-and-convexity characterization used here."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defined the constant-width, linear-skew benchmark that Section 4 locates and corrects."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provided the exact symmetric-arrival market making problem that Theorem 1 extends to imbalance."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplied econometric support for the win-curve assumption the model makes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documented the sealed-bid request-for-quote mechanism the model formalizes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documented electronic FX dealers skewing on flow at flat inventory, the practice Corollary 1 derives."}],"review_version":1}