{"id":"916e02a6-0e7d-4242-b6cd-4cd3cca053c9","arxiv_id":"1908.06355","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":3.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"The paper re-derives Geometric Brownian Motion, the Fokker-Planck equation, and the Black-Scholes-Merton equation as consequences of maximum-entropy inference with scale invariance, continuity, and a drift constraint.","lead":"This paper derives the standard Geometric Brownian Motion stock model and the Black-Scholes option-pricing equation from the authors' entropic-inference framework, rather than assuming them. A smart generalist might read it to see whether a physics-style inference principle can replace the usual 'assume a stochastic process' starting point in quantitative finance.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Entropic derivation of GBM/Black-Scholes depends on externally imposed, uniform drift and volatility; without them the derived formulas fail.","rationale":"The reader's weakest assumption identifies exactly the same load-bearing concern: the derivation treats drift and volatility as externally given inputs and requires them to be constant and price-independent for the lognormal and Black-Scholes results to follow. The paper itself acknowledges this in Sections 2.2.3, 2.3, and 3.1, so the concern is not a hidden flaw but an explicit limitation of the central claim. The mathematical derivations up to Eq. (26), the Fokker-Planck equation, and the Black-Scholes formula are internally consistent given those inputs. The issue is that the claim of 'deriving' GBM and Black-Scholes is overstated: what is derived is the maximum-entropy form of the transition density under constraints whose parameters are assumed. Since the reader's verdict CONDITIONAL already reflects this (conditions (a) and (b)), our stress-test does not move the verdict. No independent reason to reject or accept the paper outright was found; the conditions are appropriate and should be satisfied by a clarification of the status of μ and σ and ideally an empirical demonstration of the framework's added value.","tokens_in":10735,"tokens_out":13617,"duration_ms":131868,"concrete_test":"Numerically solve the Fokker-Planck equation (33) with a non-uniform volatility, e.g., σ(S) = σ0 (S/S0)^γ, and a price-dependent drift, e.g., μ(S,t) = μ0 (S/S0)^β, using a finite-difference scheme; compute the resulting European call premium by discounted expected payoff and compare with the Black-Scholes formula (44). A material difference would confirm that the uniformity assumption is load-bearing for the central claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that GBM and Black-Scholes are derived from entropic inference is load-bearing on the assumption that the drift μ, volatility σ, and risk-free rate r_f are external inputs that are constant in time and independent of price. The transition density Eq. (26) is obtained by MaxEnt with the second-moment prior (13) and the mean constraint (16), but the values of these moments are set by the externally imposed α = 1/(σ²Δt) and by the assumed arithmetic drift ⟨ΔS/S⟩ = μΔt (Eq. 19). Section 2.2.3 explicitly states the drift 'is a piece of information that ought to be found' and that a separate model is needed. The finite-time lognormal distribution (Section 2.3) and the Black-Scholes call price (44) follow only 'assuming that the drift μ and volatility are uniform' (Section 2.3) and 'assuming that the risk free rate and the volatility are uniform, constant in time and independent of price' (Section 3.1). The paper itself notes that relaxing uniformity yields a Fokker-Planck solution that is not lognormal and that Black-Scholes no longer follows. Thus the framework does not derive the standard models from deeper information; it reformulates them after taking their key parameters and structural assumptions as inputs. This is the load-bearing condition: if the inputs were instead derived or the uniformity assumption relaxed, the stated results would change.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript develops an entropic-inference framework for stock-price dynamics and European option pricing. It argues that scale invariance of the formalism selects the logarithm of price as the natural dynamical variable, and then, using a continuity prior and a drift constraint, maximum entropy yields a Gaussian transition density for log returns with mean (mu - sigma^2/2) Delta t and variance sigma^2 Delta t, i.e., the transition law of geometric Brownian motion. The corresponding Fokker-Planck equation is derived. Imposing mu = r_f as the risk-neutral information, and assuming r_f and sigma are constant and price-independent, the paper obtains the lognormal terminal distribution, the Black-Scholes call and put formulas, put-call parity, and the Black-Scholes-Merton partial differential equation.","tokens_in":10998,"tokens_out":10383,"duration_ms":107505,"significance":"The paper is a clear and largely self-contained attempt to reconstruct standard finance models from maximum-entropy updating. Its main value is conceptual: if the derivation were fully rigorous, it would show that GBM and Black-Scholes are the least-biased models consistent with scale invariance, continuity, and externally supplied mean and variance of returns, rather than ad hoc assumptions. The authors are commendably explicit about the conditional nature of the results: they state that the drift is an input to be supplied by a separate model, and that lognormality and Black-Scholes hold only under uniformity of mu, sigma, and r_f. These disclaimers define the scope of the claim. The paper does not provide empirical tests or new pricing formulas, and it does not derive the parameters from deeper information; it is best read as a reformulation of GBM and Black-Scholes within entropic inference. Within that scope, the formulas are checkable, the Fokker-Planck equation is obtained from the transition density, and the option pricing section reproduces the standard results.","major_comments":[{"comment":"The requirement that probability densities be scalars under the scale transformation is imposed rather than derived. For a genuine density, the correct transformation involves a Jacobian, so Eq. (3) does not follow from Eq. (2) unless one additionally assumes that the reference measure df is scale invariant. This is a substantive assumption that selects the logarithm as the dynamical variable and hence the lognormal family; it should be presented as an axiom of the model, not as a consequence of investor preferences.","section":"Section 2.1, Eqs. (2)-(3)"},{"comment":"The derivation of the transition density injects the GBM parameters through Eq. (19), which supplies the drift mu, and Eq. (24), which supplies the volatility sigma. Eq. (24) equates <(ln S'/S)^2> with <(Delta S/S)^2> after 'squaring the Taylor expansion,' but the squared expansion contains third- and fourth-order terms, and the equality is only valid to leading order in Delta t. More fundamentally, the constraint values in Eq. (25) are exactly the mean and variance of log-returns under GBM, so Eq. (26) follows by construction. The paper's own caveats after Eq. (31) and in Section 3.1 confirm this reading: relaxing uniformity yields a Fokker-Planck solution that is not lognormal and a Black-Scholes formula that no longer follows. The central claim in the abstract that GBM is 'derived' should therefore be qualified as a conditional MaxEnt reconstruction from externally supplied mu and sigma.","section":"Section 2.2.3, Eqs. (17)-(26)"},{"comment":"The paper states that the risk-neutral measure is derived by imposing the constraint mu = r_f. This is not a derivation from no-arbitrage; no-arbitrage alone does not determine the drift of a pricing measure, and a change of measure normally requires an equivalent-martingale-measure argument. The identification mu = r_f is an additional modeling assumption and should be labeled as such, since all subsequent option-pricing results depend on it.","section":"Section 3.1, Eq. (34)"},{"comment":"The quantity V defined by the unbounded integral of (S_T - K) is not the price of a European call or put; for a call the payoff is (S_T - K)^+ and for a put it is (K - S_T)^+. As written, Eq. (48) is the undiscounted forward price minus the strike. The backward-Kolmogorov argument should be applied to the actual payoff function g(S_T) with appropriate boundary conditions; otherwise Eq. (52) is derived for an affine function, not for an option payoff, and the BSM equation for options is not established by the argument given.","section":"Section 3.2, Eq. (48)"}],"minor_comments":[{"comment":"The integrand for the put payoff should be (K - S), not (S - K); as printed, V_p is negative.","section":"Section 3.1, Eq. (45)"},{"comment":"There is a stray comma in the integral expression 'integral dS~, P(...)'; the notation should be cleaned up.","section":"Section 3.1, Eq. (40)"},{"comment":"The text says 'We attain the Weiner process'; this should be 'Wiener process.'","section":"Section 2.2.3"},{"comment":"Equation (53) contains a misplaced comma in '1/2 sigma^2 S^2, partial^2 E / partial S^2'; this is a typographical error.","section":"Section 3.2, Eq. (53)"},{"comment":"Reference [12] has corrupted text ('Ann. Sci. c. Norm. Supr. 1990' should be 'Ann. Sci. Ecole Norm. Sup. 1900'), and several references are missing volume or page information.","section":"References"},{"comment":"The backward-Kolmogorov equation in Eq. (50) is stated without derivation; it would strengthen the presentation to note that it follows as the adjoint of the Fokker-Planck equation (33) for the homogeneous case.","section":"Section 3.2, Eq. (50)"}],"recommendation":"major_revision","confidential_remarks":"Reference [35] appears to be the authors' own published version of this same paper in Entropy; the manuscript should clarify this self-reference rather than present it as an independent citation. I also recommend that the editor view the paper as a conditional MaxEnt reformulation of GBM and Black-Scholes rather than a derivation of those models from deeper first principles; with that framing and the technical fixes above, the contribution is publishable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper does exactly what its title promises: it shows that a Gaussian transition density in log-price, with the usual mean and variance, is the maximum-entropy distribution under a continuity constraint and a drift constraint, and that the risk-neutral version with μ = r_f gives the Black-Scholes formula. That is a legitimate result, but it is a reformulation, not a derivation from deeper principles. The scale-invariance argument that selects log S as the variable is a postulate about investor preferences, and the drift μ and volatility σ are external inputs. The paper is transparent about this—Section 2.2.3 says the drift must be found elsewhere, and Section 2.3 notes that relaxing uniformity breaks the lognormal. So the stress-test note is correct: the derivation is load-bearing on assumptions the framework does not itself derive.\n\nWhat is genuinely good: the exposition is clear, the ED machinery is applied carefully, and the Fokker-Planck and BSM equation derivations are clean. The paper also cites the relevant literature honestly, including its own FX work. For a reader unfamiliar with entropic inference, this is a nice worked example of how constraints and MaxEnt produce a familiar stochastic model.\n\nThe soft spots are the standard ones for this literature. The 'derivation' of GBM relies on specifying the mean and variance, which is exactly what the model is supposed to explain. The scale-invariance argument is plausible but not forced; one could imagine other functions f(S) that are not logarithmic if one weakened the scalar-density requirement. And the risk-neutral measure is imposed by setting μ = r_f, which is the usual assumption but not derived. None of these are fatal if the paper is read as a reformulation, but the abstract overstates things by saying the dynamics are derived without assuming a process—there are still assumptions hidden in the constraints.\n\nWho should read this: people interested in the entropic inference program and its applications to finance. It is not going to change how options are priced, and it does not offer new empirical content. I would not cite it in my own work unless I were specifically writing about ED applications. But it is a competent, honest paper that would benefit from a referee suggesting a clearer statement of what is assumed versus derived. I would send it to peer review; it is not a desk reject.\n\nMy verdict: worthy of a serious referee, with the expectation that revision clarifies the status of μ, σ, and the scale-invariance postulate.","headline":"A clean MaxEnt reformulation of GBM and Black-Scholes that is honest about its assumptions; the derivation is a repackaging rather than a first-principles result, but it is a useful pedagogical piece for the entropic inference community.","tokens_in":11553,"tokens_out":2388,"would_cite":false,"duration_ms":24892,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["91G20","91G80"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper derives geometric Brownian motion and Black–Scholes-Merton pricing from entropic inference alone.","keywords":["maximum entropy method","entropic dynamics","geometric Brownian motion","European options","Black-Scholes model","Black-Scholes-Merton equation","put-call parity","risk-neutral valuation"],"falsifier":"Estimate the conditional distribution of high-frequency log-returns for individual stocks as a function of price level. If the mean or variance of $\\Delta\\ln S$ depends on $S$, or if the distribution is systematically non-Gaussian, then the uniformity assumption that produces the lognormal at maturity and the Black–Scholes formula is violated in that market.","tokens_in":10451,"feed_emoji":"📈","tokens_out":7368,"duration_ms":65097,"temperature":0.7,"pith_summary":"This paper claims that the two standard workhorses of quantitative finance—the geometric Brownian motion model of stock prices and the Black–Scholes–Merton option-pricing equation—do not have to be assumed as hypotheses about the market. They can be derived from entropic inference, a maximum-entropy procedure for updating probabilities when information is incomplete. The derivation chooses the logarithm of price as the dynamical variable because of scale invariance, then extracts a Gaussian transition density from constraints of continuity and directionality. Imposing a risk-neutral drift reproduces the Black–Scholes call price and the Black–Scholes–Merton differential equation. If the derivation is correct, these models are consequences of a single inferential principle, and modifications of the constraints offer a systematic route to new pricing models.","feed_headline":"Maximum entropy alone yields Black–Scholes option prices","feed_subtitle":"Stock dynamics and option prices are derived from information constraints, not assumed.","key_machinery":"The machinery is maximum-relative-entropy inference applied in three steps. Scale invariance forces the choice $x=\\ln S$ as the dynamical variable, since only the log turns a multiplicative rescaling of price into an additive shift. Continuity is encoded as a constraint $\\langle(\\Delta\\ln S)^2\\rangle=k$ that makes the prior a sharp Gaussian with precision $\\alpha=1/(\\sigma^2\\Delta t)$; this parameter defines the entropic clock that converts inference steps into time steps. A directionality constraint $\\langle\\ln(S'/S)\\rangle=k'$ introduces the drift, and a second-order Taylor expansion relates it to $\\mu\\Delta t-\\tfrac{1}{2}\\sigma^2\\Delta t$. The same constraints with $\\mu=r_f$ produce the risk-neutral measure, and the Fokker–Planck equation for the density follows from the entropic definition of an instant.","core_discovery":"The central claim is that the log-price transition density is fully determined by information constraints, not by postulating a stochastic process. The paper obtains\n$$P(\\ln S'|\\ln S)=\\frac{1}{Z}\\exp\\left[-\\frac{1}{2\\$sigma^{{2}}$\\$\\Delta$ t}\\left(\\ln S'-\\left(\\ln S+\\mu\\$\\Delta$ t-\\frac{1}{2}\\$sigma^{{2}}$\\$\\Delta$ t\\right)\\right)^{2}\\right],$$\nso the log-price performs a Wiener process with drift $\\mu\\Delta t-\\frac{1}{2}\\sigma^{2}\\Delta t$ and variance $\\sigma^{2}\\Delta t$, equivalent to geometric Brownian motion for the price. Setting the drift equal to the risk-free rate $r_f$ gives the risk-neutral measure, and integrating the discounted payoff yields the Black–Scholes call price $C=S_0 N(d_1)-e^{-r_f T}K N(d_2)$; differentiating the expected payoff with respect to time yields the Black–Scholes–Merton partial differential equation. The paper presents this as a derivation, not a new model: the formulas are the familiar ones, but they now follow from maximum entropy.","pith_inferences":["One testable extension: if realized log-returns are non-Gaussian or their variance depends on price level, the derived distribution should fail; the framework would then predict option prices from the Fokker–Planck solution rather than from Black–Scholes.","The same scale-invariance-plus-constraints recipe could in principle be applied to any positively priced asset, suggesting a unified entropic derivation of other market models such as foreign-exchange option pricing.","Because the derivation makes every input an explicit constraint, it offers a diagnostic tool: estimating the constraints from option prices could show exactly which assumed symmetries are violated in real markets."],"forward_implications":["The familiar GBM transition density is recovered: log returns are Gaussian with variance $\\sigma^2\\Delta t$, and prices at finite maturity are lognormal.","If drift or volatility are not constant in time or price, the finite-time distribution solves a Fokker–Planck equation and is no longer lognormal, so the Black–Scholes formula should be replaced by the corresponding solution.","The risk-neutral valuation principle is obtained by one constraint, $\\mu=r_f$, not added as a separate assumption.","The call and put formulas satisfy put-call parity, which the paper interprets as evidence that the expected-payoff pricing is arbitrage-free.","The Black–Scholes–Merton PDE follows by differentiating the expected payoff with respect to time, with the backward Kolmogorov equation supplying the evolution of the transition density."],"supporting_citations":[{"why":"Defines the entropic inference framework and relative-entropy updating used throughout the derivation.","marker":"[1]"},{"why":"Introduces entropic time, which the paper uses to turn inference steps into dynamics and to define the entropic clock.","marker":"[4]"},{"why":"Bachelier's original stochastic model of stock prices, the geometric Brownian motion that the paper derives rather than assumes.","marker":"[12]"},{"why":"Samuelson's rediscovery of geometric Brownian motion for stock prices, serving as the baseline process the derivation reproduces.","marker":"[14]"},{"why":"Black and Scholes' early option valuation work, one of the targets the paper rederives entropically.","marker":"[16]"},{"why":"Black and Scholes' option pricing formula and corporate-liabilities analysis, the standard result the paper derives from risk-neutral constraints.","marker":"[17]"},{"why":"Merton's rational option pricing theory and the Black–Scholes–Merton differential equation, which the paper derives from the expected payoff.","marker":"[18]"},{"why":"Axiomatic derivation of the principle of maximum entropy and minimum cross-entropy, justifying the updating rule the whole argument relies on.","marker":"[29]"},{"why":"Source for risk-neutral valuation and no-arbitrage discounting, supplying the constraint used to obtain the risk-neutral measure.","marker":"[34]"}],"fun_headline_variants":["Entropic inference derives Black-Scholes from information","No assumptions: Black-Scholes emerges from entropy alone","Scale invariance and maximum entropy yield Black-Scholes","From information constraints to Black-Scholes pricing","Black-Scholes from entropy, not economic assumptions"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the drift and volatility are externally supplied constants and that investors are indifferent to absolute price levels; if either fails, the derived lognormal maturity distribution and the Black–Scholes formulas no longer follow.","fun_headline_variants_meta":{"raw":{"variants":["Entropic inference derives Black-Scholes from information","No assumptions: Black-Scholes emerges from entropy alone","Scale invariance and maximum entropy yield Black-Scholes","From information constraints to Black-Scholes pricing","Black-Scholes from entropy, not economic assumptions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000983,"raw_usage":{"total_tokens":4194,"prompt_tokens":987,"completion_tokens":3207,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":603,"completion_tokens_details":{"reasoning_tokens":3136}},"tokens_in":603,"tokens_out":3207,"duration_ms":20744,"temperature":1.0,"reasoning_tokens":3136,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:46:40.835195+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Estimate the conditional distribution of high-frequency log-returns for individual stocks as a function of price level. If the mean or variance of $\\Delta\\ln S$ depends on $S$, or if the distribution is systematically non-Gaussian, then the uniformity assumption that produces the lognormal at maturity and the Black–Scholes formula is violated in that market.","supporting_citations":[{"cited_title":"Th´ eorie de la sp´ eculation","cited_arxiv_id":null,"evidence_quote":"Bachelier's original stochastic model of stock prices, the geometric Brownian motion that the paper derives rather than assumes."},{"cited_title":"Axiomatic derivation of the prin ciple of maximum entropy and the principle of minimum cross-entropy","cited_arxiv_id":null,"evidence_quote":"Axiomatic derivation of the principle of maximum entropy and minimum cross-entropy, justifying the updating rule the whole argument relies on."},{"cited_title":"C.; Basu, S","cited_arxiv_id":null,"evidence_quote":"Source for risk-neutral valuation and no-arbitrage discounting, supplying the constraint used to obtain the risk-neutral measure."}],"review_version":1}