{"id":"45d0e611-ce9c-4086-8bec-4d190153becb","arxiv_id":"2505.15296","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A calibrated agent-based market simulator computes liquidity risk surfaces for Hang-Seng futures, but its claimed emergent price impact is largely inherited from a fitted impact function.","lead":"This paper describes an agent-based computer simulation of a futures market, calibrated to one day of Hang-Seng Index tick data, that estimates the cost of buying or selling large orders. It is of interest because exchanges and clearing houses need such cost estimates to manage liquidation risk after a default.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Permanent impact is inherited from Eq. (8), not emergent: X_t accumulates the fitted square-root single-trade impact f_mi, and §3.3.2 reconstructs V_t by subtracting that same function. An ablation disabling the reflexive term is needed to support the claimed emergence.","rationale":"The paper is a serious attempt to build a calibrated ABM liquidity-risk calculator, and the LOB/auction mechanics plus the baseline/counterfactual cost decomposition are valuable. The central scientific claim, however, is the emergence of square-root transient and permanent impact. That claim is load-bearing because it motivates the model as a replacement for stylised impact functions. The weak point is Eqs. (7)-(8): fundamental traders' value belief shifts by the exact fitted single-trade impact function on every trade, so the permanent impact of a meta-order is essentially the sum of calibrated increments. Section 3.3.2 explicitly reconstructs V_t by subtracting this same cumulative impact from prices, which makes the permanent-impact part of the result a calibrated input rather than an emergent output. I agree with the reader's weakest_assumption. The proposed ablation is the minimal decisive test: disable X_t and see if permanent impact survives. If it does not, the headline should be reframed as 'calibrated single-trade impact accumulation with emergent transient dynamics,' and the model remains acceptable as a calibrated tool. This does not change the reader's CONDITIONAL verdict, so I keep UNCHANGED.","tokens_in":12784,"tokens_out":5804,"duration_ms":55803,"concrete_test":"Re-run the meta-order experiments of §5.1/§5.2 (e.g., Q=20,000 over the horizons in Fig. 5) with the reflexive fundamental term disabled, i.e., set X_t≡0 in Eq. (8) (equivalently f_mi(Q)≡0), keeping all other calibrated parameters, random seeds, and the baseline/counterfactual protocol identical. Measure the post-execution permanent impact as the residual mid-price difference between counterfactual and baseline. If permanent impact collapses to near zero or loses its concave square-root shape, the effect is inherited from Eq. (8) rather than emerged; if a substantial concave permanent impact survives, the emergence claim is supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing premise of the headline result—that transient and permanent impact 'emerge' from trader behaviours rather than being explicitly coded—is undercut by the construction in Eqs. (7)-(8) and §3.3.2. The reflexive fundamental value is \\tilde V_t = V_t + X_t, with X_t = X_{t-1} + f_mi(Q_t), and f_mi(Q)=0.561\\sqrt{Q} is fitted in §3.3.3 to the same day's second-level orderflow imbalance and mid-price changes. Because X_t is updated by this calibrated square-root function on every trade, the counterfactual-minus-baseline permanent impact is substantially the accumulated f_mi of the meta-order's own trades: the permanent square-root law is injected at the single-trade level, not a collective emergent outcome of heterogeneous agents and the LOB. The exogenous fundamental V_t is itself defined in §3.3.2 by subtracting \\sum f_mi(Q_i) from historical prices, so the same fitted function is used on both sides of the reconstruction. An ablation that removes Eq. (8) would reveal whether permanent impact is simply this term. The transient component may still emerge from the orderbook mechanics, but the paper's central claim bundles transient and permanent impact together; as written, the permanent component is best described as calibrated impact accumulation. The horizon dependence reported in Fig. 5 also warrants scrutiny, since summing f_mi over more, smaller slices does not automatically produce decay with horizon.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes an agent-based model (ABM) of a continuous double auction limit order book, calibrated to one day of Hang Seng Index Futures tick data, and uses it to estimate the cost of executing large meta-orders over different horizons and sizes. The model combines zero-intelligence order arrival calibrated to historical rates, Chiarella-style fundamental/momentum/noise trader demands, and a novel 'reflexive fundamental value' in which traders update their fair value by a fitted single-trade impact function. The authors claim that transient and permanent market impact 'emerge' from these behaviours, reproduce a square-root impact law, and produce a liquidity risk surface for practical use by exchanges and risk managers.","tokens_in":13142,"tokens_out":3475,"duration_ms":31906,"significance":"If the central claims were fully supported, the framework would be a practically useful tool for estimating liquidation costs and for optimizing execution strategies under liquidity risk, with a computationally explicit limit order book and a clean decomposition of transaction cost into market risk and market impact. The paper is clearly written and provides a detailed calibration procedure, and the Monte-Carlo liquidity risk surfaces are a useful industrial artifact. However, the load-bearing claim that permanent impact 'emerges' rather than being explicitly coded is severely weakened by the construction of the reflexive fundamental value, which directly accumulates the fitted single-trade impact function. The manuscript needs either an ablation or a substantial reframing of its central claim before its conclusions can be accepted.","major_comments":[{"comment":"The headline claim that permanent impact emerges is contradicted by the model construction. In Eq. (8), the signal X_t is updated as X_t = X_{t-1} + f_mi(Q_t), and f_mi(Q) = 0.561 sqrt(Q) is fitted to the same day's orderflow imbalance and mid-price changes in §3.3.3. The permanent impact of a meta-order is therefore, to first order, the accumulated fitted single-trade impact of its own trades; the square-root law is injected at the single-trade level rather than emerging from collective agent behaviour. The paper should run an ablation that disables or replaces Eq. (8) and show whether permanent impact survives, or explicitly revise the 'emergence' claim in §7, which currently states that the authors do not explicitly model permanent impact.","section":"§3.2.1, Eq. (8); §7"},{"comment":"The calibration of the exogenous fundamental value V_t raises a circularity concern. In §3.3.2, V_t is constructed by subtracting the cumulative fitted impact sum_i f_mi(Q_i) from historical trade prices, and then the same f_mi is used in Eq. (8) to drive the reflexive fundamental value. The counterfactual-minus-baseline permanent impact measured in §5.1 may therefore largely reproduce the input calibration rather than independently validate it. To support the liquidity-risk surface, the authors should demonstrate calibration stability across multiple days or validate the model on a held-out day.","section":"§3.3.2, §3.3.3"},{"comment":"The empirical claim that impact decays with horizon is stated qualitatively, and no statistical quantification is provided (e.g., standard errors over the Monte-Carlo runs, or a fitted relation between impact and horizon). Given that X_t accumulates f_mi over successive slices, it is not self-evident that spreading the same total size over a longer horizon reduces the accumulated impact; the paper should report the simulated impact-vs-horizon curve with error bars and compare it directly with the fitted f_mi, rather than only with the external Bloomberg-style model.","section":"§5.1, Fig. 5"}],"minor_comments":[{"comment":"Typos: 'marker orders' should be 'market orders', and 'cummulative' should be 'cumulative'.","section":"§3.3.2"},{"comment":"The symbol delta_p is used for the exponentially distributed offset in the ZI model in Eq. (1) and for the sampled depth in Eq. (4); these are conceptually distinct and the notation should be disambiguated.","section":"Eqs. (1), (4)"},{"comment":"References [23] and [24] appear to be the same paper (McGroarty et al., 2019) with only the volume/page formatting differing; please merge the duplicate entry.","section":"References"},{"comment":"The text refers to 'Figure 7(e)', 'Figure 7(c) and (d)', and 'Figure 7(a) and (b)', but the figure caption lists panels in a single line; please ensure the panel labels are clearly visible and cited consistently.","section":"§5.2, Fig. 7"},{"comment":"The text says 'Strategy C executes at the end and is always sub-optimal under the no-drift assumption,' but the figure shows the efficient frontier; clarifying how 'sub-optimal' is defined (higher cost for the same risk?) would improve readability.","section":"§5.3, Fig. 8"}],"recommendation":"major_revision","confidential_remarks":"The reader's concern about circularity is well-founded and is the main reason for the major revision recommendation. The manuscript is an industry working paper, and its practical engineering contribution may still be valuable; however, the scientific claim of 'emergence' of permanent impact must be either supported by an ablation experiment or substantially reframed. I would suggest the editor request the ablation and a validation on a second day of data."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is an industry-grade agent-based model for transaction costs on HSI futures, with a genuinely useful decomposition of implementation shortfall into market risk and market impact via paired baseline/counterfactual simulations. The reflexive fundamental value (Eqs. 7–8), where traders mark the fundamental up/down by an accumulated single-trade impact function, is new to the ABM literature, and the resulting liquidity risk surface is a practical tool. I'd send it to a serious referee.\n\nThat said, the central claim that permanent impact \"emerges\" from trader behavior rather than being coded does not hold as written. Equation (8) sets X_t = X_{t-1} + f_mi(Q_t), with f_mi(Q) = 0.561√Q fitted to the same day's orderflow imbalance and price changes (Section 3.3.3). The permanent impact is therefore the accumulated calibrated impact of the meta-order's own trades; it is injected, not emergent. Worse, the exogenous fundamental V_t is itself reconstructed in Section 3.3.2 by subtracting that same cumulative f_mi from historical prices, so the same fitted function appears on both sides of the calibration. An ablation that disables Eq. (8) (or replaces it with a fixed V_t) is necessary to see what the orderbook mechanics alone produce; the transient component may still be emergent, but the paper bundles the two.\n\nSecond soft spot: validation is a single day (2022-12-23), with no out-of-sample test, no code or data released, and the Chiarella parameters are calibrated to stylized facts on that same day. That makes the quantitative surface a calibrated output, not a prediction. The horizon dependence in Fig. 5 also deserves scrutiny: accumulating f_mi over more, smaller slices does not by itself produce decay with horizon, and the paper doesn't explain the mechanism behind that decay.\n\nOn balance, the framework and decomposition are worth engaging. The fix is straightforward: reframe the permanent-impact mechanism as calibrated impact accumulation, run the ablation, add a second day and maybe a second contract, and release artifacts. With that, it becomes a credible practitioner tool. Without it, the \"emergence\" claim overstates the evidence.\n\nI'd accept it for peer review but expect major revision.","headline":"Useful industrial ABM for liquidity risk, but the 'emergent' permanent impact is really injected through a calibrated single-trade impact function in the reflexive fundamental value; the paper needs an ablation and out-of-sample checks.","tokens_in":13689,"tokens_out":1996,"would_cite":false,"duration_ms":18194,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that transient and permanent market impact, including the square-root law, emerge from simple trader behaviours in an agent-based market simulator rather than being explicitly coded.","keywords":["agent-based modelling","liquidity risk","market impact","square-root law","continuous double auction","reflexive fundamental value","Hang-Seng Index Futures","optimal execution"],"falsifier":"Run the simulator twice with identical orderflow and trader behaviour, once with the reflexive update active and once with $X_t \\equiv 0$: if the square-root-shaped permanent impact survives the frozen case, the emergence claim is refuted, and if it disappears, the impact is an artifact of the fitted $f_{\\rm mi}$ rather than an independent outcome. A complementary check is to compare the simulated impact curve with measured meta-order impact on a different day than the calibration day.","tokens_in":12565,"feed_emoji":"📉","tokens_out":13279,"duration_ms":100619,"temperature":0.7,"pith_summary":"The paper tries to establish that an agent-based simulation of a real futures market can price the cost of executing a large order without assuming a market-impact formula. In the model, fundamental, momentum, and noise traders submit orders to an exchange engine with a limit order book, and every trade updates traders' estimate of the asset's fair value because traders behave as if orderflow carries information. The claim is that this reflexive update is enough: transient impact follows the square-root law and relaxes to a permanent level, matching empirical meta-order behaviour. If correct, the simulator can compute transaction costs for any execution strategy, split them into market impact versus market risk, and build a liquidity risk surface for a contract such as the Hang-Seng Index Futures.","feed_headline":"Simulated markets reproduce the square-root cost of trading","feed_subtitle":"Agent-based Hang-Seng futures model shows impact emerging from trader rules, not a coded formula.","key_machinery":"The load-bearing mechanism is the reflexive fundamental value $\\tilde{V}_t = V_t + X_t$, where $V_t$ is an exogenous fair-value random walk and $X_t$ accumulates the fitted single-trade impact $f_{\\rm mi}(Q_t)$ after each trade. It is the channel through which traders treat orderflow as informed, so every executed trade changes perceived fair value and shifts subsequent quotes. Around it sit the exchange's continuous double auction (price-time priority matching on a limit order book) and the three trader behaviours the simulator uses: fundamental traders acting on $\\tilde{V}_t$, momentum traders following mid-price trends, and noise traders. The reflexive update is what converts the fitted one-trade impact function into aggregate transient and permanent impact, which is why the paper can claim the impact is emergent.","core_discovery":"The central claim is that price impact is an emergent outcome of the agent rules, not a parameterised component of the simulator. The model contains no meta-order impact formula; it contains only a single-trade impact function $f_{\\rm mi}(Q)=0.561\\sqrt{Q}$ calibrated from historical orderflow imbalance and mid-price changes on 2022-12-23, and feeds it into a reflexive fundamental value $\\tilde{V}_t = V_t + X_t$, with $X_t = X_{t-1} + f_{\\rm mi}(Q_t)$. Fundamental traders act on $\\tilde{V}_t$, so each executed trade shifts perceived fair value and the next round of quoting and trading. The paper's demonstration is that simulated liquidation of a meta-order produces a concave, square-root-shaped transient impact that decays to a permanent impact, and that the resulting cost surface for the Hang-Seng Index Futures has the expected shape: higher cost for larger size, lower cost for longer horizon.","pith_inferences":["A stricter test of the emergence claim would freeze the reflexive update ($X_t \\equiv 0$) or fit $f_{\\rm mi}$ on a different day; the paper's calibration uses the same day's orderflow and price data for both the impact function and the exogenous value, so part of the permanent impact is inherited from the fitted curve.","The baseline/counterfactual cost decomposition could be applied to adaptive execution algorithms and to other instruments, turning the method into a general transaction-cost analytics pipeline.","Because the calibration uses a single trading day, an out-of-sample liquidity surface across several days would show whether the parameters and impact function are stable enough for production use."],"forward_implications":["Any execution strategy can be priced by paired baseline/counterfactual simulations, so a trader can compare schedules, horizons, and order types without a closed-form impact model.","Transaction cost decomposes into market impact and market risk, giving both expected cost and the uncertainty around a forced liquidation.","For the Hang-Seng Index Futures contract, the simulated liquidity risk surface rises concavely with size and falls with horizon, consistent with the square-root law.","The calibrated simulator reproduces the efficient frontier for execution schedules predicted by the optimal-execution framework: front-loaded strategies cost more but have lower variance, and balanced strategies are near-optimal."],"supporting_citations":[{"why":"Supplies the empirical single-trade impact concavity that $f_{\\rm mi}(Q)=0.561\\sqrt{Q}$ is fitted to match.","marker":"[6]"},{"why":"Documents the square-root law for meta-order impact that the emergent impact is claimed to reproduce.","marker":"[2]"},{"why":"Supplies the optimal-execution framework used to define strategy costs and the efficient frontier.","marker":"[1]"},{"why":"Provides the zero-intelligence order-arrival and order-pricing mechanics that the simulator extends.","marker":"[15]"},{"why":"Defines the fundamental, momentum, and noise trader demands used by the agents.","marker":"[22]"},{"why":"Provides the stylised-facts distance measure used to calibrate hidden trader parameters.","marker":"[16]"},{"why":"Supplies the surrogate-modelling optimisation used to fit the stochastic simulation to data.","marker":"[20]"},{"why":"Shows a prior agent-based simulator that does not clearly reproduce the square-root law, motivating the reflexive-value mechanism.","marker":"[13]"}],"fun_headline_variants":["Agent-based sim shows price impact emerges from rules","Liquidity risk model: impact emerges, not pre-programmed","Emergent square-root impact from agent-based futures model","Hang-Seng futures model: no impact formula, still realistic","Monte Carlo agent model reproduces trading cost curve"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole construction rests on the assumption that the fitted one-trade impact function, used to update traders' fair-value belief after every trade, faithfully and stably represents how information enters the price, and that the historical price path used to calibrate it contains no other systematic drivers.","fun_headline_variants_meta":{"raw":{"variants":["Agent-based sim shows price impact emerges from rules","Liquidity risk model: impact emerges, not pre-programmed","Emergent square-root impact from agent-based futures model","Hang-Seng futures model: no impact formula, still realistic","Monte Carlo agent model reproduces trading cost curve"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000221,"raw_usage":{"total_tokens":1486,"prompt_tokens":1017,"completion_tokens":469,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":633,"completion_tokens_details":{"reasoning_tokens":388}},"tokens_in":633,"tokens_out":469,"duration_ms":4721,"temperature":1.0,"reasoning_tokens":388,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T15:20:55.184207+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the simulator twice with identical orderflow and trader behaviour, once with the reflexive update active and once with $X_t \\equiv 0$: if the square-root-shaped permanent impact survives the frozen case, the emergence claim is refuted, and if it disappears, the impact is an artifact of the fitted $f_{\\rm mi}$ rather than an independent outcome. A complementary check is to compare the simulated impact curve with measured meta-order impact on a different day than the calibration day.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the empirical single-trade impact concavity that $f_{\\rm mi}(Q)=0.561\\sqrt{Q}$ is fitted to match."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documents the square-root law for meta-order impact that the emergent impact is claimed to reproduce."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the optimal-execution framework used to define strategy costs and the efficient frontier."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the zero-intelligence order-arrival and order-pricing mechanics that the simulator extends."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the fundamental, momentum, and noise trader demands used by the agents."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the surrogate-modelling optimisation used to fit the stochastic simulation to data."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows a prior agent-based simulator that does not clearly reproduce the square-root law, motivating the reflexive-value mechanism."}],"review_version":1}