{"id":"56f619e0-d816-4365-9ad1-30df0a27a7cc","arxiv_id":"1908.07372","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"high","formal_verification":"none","parameter_count":7,"one_line_summary":"Cricket chases are modeled with stochastic differential equations, producing a per-ball win probability and a three-parameter team description.","lead":"This paper models a cricket chase as a random process and derives a per-ball probability that the chasing team wins from the gap between required and actual run rates. It proposes that each team can be summarized by three parameters, drift, volatility, and mean reversion, instead of ranking tables.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Equation 15 is not the conditional win probability implied by the OU model in Eq. 11; the printed formula is inconsistent with the Gaussian conditional distribution from Eqs. 13–14.","rationale":"The reader's weakest_assumption concerns the validity of the OU process and constant parameters, but the more immediate and decisive problem is internal: Eq. 15 does not follow from Eq. 11 and the conditional normal distribution obtained from Eqs. 13–14. This is not a modeling assumption that could be fixed by better data; it is an algebraic inconsistency in the formula that the paper presents as the main result. I agree with the reader's overall REJECT verdict, but I locate the load-bearing failure one step earlier than the reader's stated weakest assumption. The negative fitted σ and the lack of out-of-sample validation are additional problems, yet they are secondary: even with perfect parameter fitting and validation, an incorrect Eq. 15 would invalidate the claimed win probabilities. The proposed concrete test settles the issue directly because it compares the published formula to the closed-form conditional probability for a parameter choice where the OU solution is unambiguous.","tokens_in":6396,"tokens_out":8776,"duration_ms":82237,"concrete_test":"Re-derive Eq. 15 from the Fokker-Planck solution of Eq. 11 without using Eq. 15, and evaluate both expressions at x0=1, x1=0, σ=1, α=1, t=0. The correct conditional probability is Φ(e^{-1}/√(0.5(1-e^{-2}))) ≈ 0.71; if Eq. 15 evaluates to ≈0.78 instead, Eq. 15 is not the OU win probability. If the printed formula is a typesetting error, the corrected formula should still be compared to this closed-form value at several (x0,x1,σ,α,t) points to confirm consistency with Eqs. 13 and 14.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that Eq. 15 gives P(X(1)>0 | X(t)=α) for dX_t = x0(x1-X_t)dt + σ dW_t. For this Markov process, the conditional distribution is normal with mean m = α e^{-x0(1-t)} + x1(1-e^{-x0(1-t)}) and variance v = σ^2/(2x0)(1-e^{-2x0(1-t)}), so the correct probability is P = 1/2[1+erf(m/√(2v))]. The printed Eq. 15 instead puts x1 inside an exponential multiplying α and uses α√2 e^{x1(1-e^{x0(t-1)})+(t-1)x0} / (σ√(...)); it does not reduce to the required form. Even in the simplified case x1=0, the argument of erf differs from α e^{-x0(1-t)}/√(2v) by an extra factor √2 and by how x0 appears under the square root. With x0=1, x1=0, σ=1, α=1, t=0, the correct OU value is about 0.71 while Eq. 15 interpreted literally gives about 0.78. Because this formula is the basis of the win-probability panels in Figs. 5, 7 and 9, the central quantitative claim is unsupported by the manuscript as written.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript introduces three stochastic differential equation (SDE) models for cricket, using the variable X(t) defined as the difference between net run rate and required run rate, and derives expressions for the win probability P(X(1)>0 | X(t)=alpha). Model 1 is Brownian motion with drift; Model 2 adds a wicket-dependent drift via a Poisson survival function; Model 3 is an Ornstein-Uhlenbeck process with parameters (x0, x1, sigma) intended to summarize team quality. The paper illustrates the models on a historic India-Sri Lanka ODI and proposes the parameter triplet as a quantitative team descriptor.","tokens_in":6743,"tokens_out":7336,"duration_ms":72157,"significance":"The conceptual idea of modelling cricket's score differential by an SDE is reasonable and could lead to a compact, interpretable team descriptor and a ball-by-ball win-probability update. The manuscript provides explicit formulas and a clear framework, which is a useful starting point. However, the central probability formulas contain algebraic errors (Eqs. 4, 6, and 15), the variance expression in Eq. 7 is not derived and appears malformed, the reported fit includes a negative volatility, and the illustrative match is inside the fitting period, so the quantitative claims are not supported as written.","major_comments":[{"comment":"For X(1) ~ N(mu, sigma^2), the correct win probability is P(X(1)>0) = 1/2[1 + erf(mu/(sigma sqrt(2)))]. Equation (4) instead has (1 - mu) in the numerator, which is dimensionally wrong and numerically incorrect (e.g., at mu=0 it gives a value depending on 1/sigma rather than 1/2). Similarly, Eq. (6), the conditional probability given X(t1)=alpha, should be 1/2[1 + erf((alpha + mu(1-t1))/(sigma sqrt(2(1-t1))))], but the printed formula contains '1 - mu(1-t1) - alpha', which has the wrong sign structure and is not monotone increasing in alpha as it should be. These errors invalidate the Model 1 probability curves shown in Figures 2 and 5.","section":"2.1, Eqs. (4) and (6)"},{"comment":"The claimed 'closed form solution for the variance of the above probability' is introduced with 'After few steps of calculations' and is not derived. The expression mixes a probability-like term with Owen's T function without defining the random variable over which the variance is taken, and it states independence of alpha even though the probability in Eq. (6) depends on alpha. As written, Eq. (7) cannot be checked, reproduced, or used; it should either be derived rigorously with all assumptions stated, or removed.","section":"2.1, Eq. (7)"},{"comment":"Equation (15) is the central formula of the paper, giving P(X(1)>0 | X(t)=alpha) for the Ornstein-Uhlenbeck process dX_t = x0(x1-X_t)dt + sigma dW_t. For this process, the conditional distribution is normal with mean m = alpha e^{-x0(1-t)} + x1(1-e^{-x0(1-t)}) and variance v = sigma^2/(2x0)(1-e^{-2x0(1-t)}), so the correct probability is 1/2[1 + erf(m/sqrt(2v))]. The printed Eq. (15) has an extra factor sqrt(2), places x1 inside an exponential multiplying alpha, and omits the x1(1-e^{-x0(1-t)}) term; it does not reduce to the correct expression even in the case x1=0. For example, with x0=1, x1=0, sigma=1, alpha=1, and t=0, the correct OU value is about 0.71, while Eq. (15) interpreted literally gives about 0.80. Because this formula underlies the win-probability panels in Figures 5, 7, and 9, the central quantitative claim is unsupported by the manuscript as written.","section":"2.3, Eq. (15)"},{"comment":"The reported fit for team India gives sigma = -2.76. Since Eq. (14) depends on sigma^2 and Eq. (15) uses sigma in the denominator of an erf argument, a negative value is not a harmless sign convention: it would flip the sign of the probability correction and contradicts the stated interpretation of sigma as a volatility. The paper must report a nonnegative sigma (or explicitly define a signed convention and justify it), and should provide parameter uncertainties; without this, the fitted triplet (x0, x1, sigma) is not physically meaningful.","section":"2.3, Fig. 8"},{"comment":"The illustrative India-Sri Lanka match (April 2, 2011) lies inside the 2005-2017 period used to fit x0, x1, and sigma for India (Fig. 8). Therefore the win-probability curves in Figures 5, 7, and 9 are in-sample evaluations, not predictions. The abstract claims a method to calculate winning probability as a progression of balls, but no out-of-sample test is provided; a demonstration such as fitting on pre-2011 data and evaluating on later matches is needed to support the predictive claim.","section":"2.3, Figs. 5 and 8"}],"minor_comments":[{"comment":"The definition of X(t) is inconsistent: Eq. (1) defines X(t) = NR(t) - RR(t), while the Conclusion states 'Xt := RR - NR'; this sign discrepancy must be resolved throughout the paper.","section":"Section 1, Eq. (1) and Conclusion"},{"comment":"There are numerous typographical errors, including 'statical cumulants' (Abstract), 'winner process' (Section 2.1, should be 'Wiener process'), and repeated 'loosing' for 'losing'. These should be corrected in a thorough revision.","section":"Abstract and throughout"},{"comment":"The notation is inconsistent: Eq. (2) writes N(mu t, sigma^2 t), using sigma^2 as a variance, while later the text and Eq. (4) write N(mu, sigma), with sigma as a standard deviation. Standardize the notation to avoid confusion.","section":"2.1, Eq. (2)"},{"comment":"Equation (11) is missing parentheses: it should read dX_t = x0(x1 - X_t)dt + sigma dW_t. As printed, 'dXt = x0x1 - x0Xtdt + sigma dWt' is ambiguous.","section":"2.3, Eq. (11)"},{"comment":"Equations (12) and (13) use X_{t-1} as the initial value, but the solution of the OU process from time 0 should involve X_0; the notation should be corrected and the time indices made consistent.","section":"2.3, Eqs. (12) and (13)"}],"recommendation":"reject","confidential_remarks":"The manuscript reads as an early draft with multiple unresolved derivations and no valid empirical validation. The errors in Eqs. (4), (6), and (15) are not superficial typographical issues; they change the numerical outputs of the figures. Even if corrected, the in-sample validation would need to be redone with out-of-sample data. I recommend rejection; a substantially revised version with corrected formulas and a proper validation study could be reconsidered by the journal or a more specialized venue."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper introduces a genuinely cricket-specific stochastic state variable: X(t) = NR − RR, the gap between net run rate and required run rate. It then models this as an Ornstein–Uhlenbeck process and adds a wicket-dependent drift via a Poisson survival function. That choice of state variable is real and the wicket modification is a thoughtful touch. The writing is clear and the data analysis, though simple, uses actual ODI matches.\n\nHowever, the mathematics as printed does not support the claims. The stress-test note is right: Equation 15 is not the conditional win probability implied by Equation 11. For an OU process, the conditional distribution of X(1) given X(t)=α is Gaussian with mean α e^{x0(t−1)} + x1(1−e^{x0(t−1)}) and variance σ²/(2x0)(1−e^{2x0(t−1)}). Equation 15 instead puts x1 inside an exponential multiplying α and drops the factor 2 in the variance term. In a simple test (x0=1, x1=0, σ=1, α=1, t=0), Equation 15 gives about 0.78 while the correct value is about 0.71. That is not a rounding error—it is an internal contradiction with the paper's own Equations 13 and 14.\n\nEarlier problems compound this. Equation 4 writes P(X(1)>0) with (1−μ) instead of μ, and Equation 6 repeats the same sign error. Equation 7 for the variance of the win probability is malformed and appears without a derivation. The fitted volatility for India is reported as σ=−2.76, which is not a valid parameter from an SDE fit. The illustrative India–Sri Lanka match is inside the fitting window, so the displayed “prediction” is in-sample, not an out-of-sample test. None of these flaws destroy the underlying idea, but they are not minor typos: they break the central quantitative result.\n\nWhat the paper does well is simpler. It takes a real sports quantity, defines a plausible dynamics, and shows how per-ball win probabilities would follow. The figures are legible, the model progression from Brownian motion to mean reversion is natural, and the wicket perturbation via Poisson survival is a creative bridge between cricket structure and SDE machinery.\n\nWho this is for: someone in sports analytics who wants a compact parameterized model of a run chase, and a referee who enjoys checking algebra. It deserves a serious referee, but the referee should be prepared to send it back for major revision, or to reject and invite resubmission with corrected equations and out-of-sample validation. I would not cite it in its current form, but I would bring it to a reading group as a useful case study in how a promising modeling idea can be undermined by unchecked formulas.","headline":"A cricket-specific SDE state variable with real analytic potential, but the published formulas—especially Equation 15—contradict the model they are derived from, so the central claim is unsupported as written.","tokens_in":7286,"tokens_out":4289,"would_cite":false,"duration_ms":45217,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60H10","60J70"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper argues that a cricket match's progression is captured by one mean-reverting stochastic variable, so that every team is characterized by three parameters and the win probability can be updated ball by ball.","keywords":["stochastic differential equation","Ornstein-Uhlenbeck process","cricket","win probability","sports analytics","net run rate","mean reversion"],"falsifier":"Take a large sample of real matches and, for many time points $t$ and observed leads $\\alpha$, compare the empirical frequency of wins with Eq. (15) evaluated at the fitted $(x_0, x_1, \\sigma)$. If the residuals $X(1) - E[X(1) \\mid X(t)=\\alpha]$ are skewed or heavy-tailed, or if the fitted parameters drift systematically across opponents, the central claim is falsified.","tokens_in":6166,"feed_emoji":"🏏","tokens_out":8565,"duration_ms":85750,"temperature":0.7,"pith_summary":"This paper introduces a stochastic differential equation formalism for cricket, centered on the lead variable $X_t = NR(t) - RR(t)$, the difference between net run rate per ball and required run rate per ball. It argues that $X_t$ evolves as an Ornstein-Uhlenbeck process $dX_t = x_0(x_1 - X_t)dt + \\sigma dW_t$, which makes the final lead conditionally Gaussian and gives a closed-form win probability at any ball. If correct, every team reduces to three physically meaningful numbers—reversion speed, long-run average lead, and volatility—and live win probabilities can be computed from the current match state. The paper demonstrates the approach on a real India–Sri Lanka match, where the probability stays near 0.5 even when the raw lead looks negative.","feed_headline":"Cricket win odds follow one mean-reverting equation","feed_subtitle":"Three numbers per team—reversion speed, long-run edge, volatility—update the live win chance every ball.","key_machinery":"The load-bearing object is the Ornstein-Uhlenbeck process in Eq. (11), a mean-reverting random walk with constant volatility. Its explicit solution is Gaussian, so the conditional distribution of the final lead given the current lead is normal with mean $\\alpha e^{-x_0(1-t)} + x_1(1 - e^{-x_0(1-t)})$ and the corresponding variance; Eq. (15) is the tail probability of that normal above zero. The same equations yield the stationary mean $x_1$ and stationary variance $\\sigma^2/(2x_0)$, which is how the physical meanings of the three parameters are extracted.","core_discovery":"The central claim is that the net-run-rate lead is the fundamental state variable of a cricket match, not a heuristic scoreboard statistic. Its evolution is $dX_t = x_0(x_1 - X_t)dt + \\sigma dW_t$, an Ornstein-Uhlenbeck process whose solution yields closed-form expressions for the mean, $E[X_t]$, and variance, $Var[X_t]$. The probability that the second-innings team wins at the end, $P(X(1)>0 \\mid X(t)=\\alpha)$, is then the explicit error-function expression in Eq. (15). The three fitted constants receive physical interpretations: $x_1$ is the long-run lead the team tends to settle at, $x_0$ controls how quickly it returns there, and $\\sigma$ measures unpredictability.","pith_inferences":["Editorial inference: The same mean-reverting Gaussian machinery could be applied to any timed sport with a score differential by redefining $X_t$ as the score lead and fitting the same three parameters.","Editorial inference: The paper fits parameters to past matches but does not report a full out-of-sample test; a natural extension is to fit on one set of seasons and evaluate Eq. (15) on held-out matches, which would quantify how much predictive power the three parameters actually carry.","Editorial inference: Because opponents enter only through the fitted parameters, a natural refinement left implicit is to let $x_1$ depend on the opposing team's strength, turning the model into a generative rating system that explains head-to-head results instead of merely fitting them."],"forward_implications":["A team's quality is summarized by the vector $(x_0, x_1, \\sigma)$, so head-to-head comparisons reduce to comparing these three numbers instead of combining arbitrary rating statistics.","Given any ball index $t$ and current lead $\\alpha$, Eq. (15) gives an explicit update of the chasing team's win probability, allowing live odds to be refreshed ball by ball.","The long-run mean $x_1$ and stationary variance $\\sigma^2/(2x_0)$ separate 'how good a team is' from 'how unpredictable a team is'.","Fitting Eqs. (13)–(14) to historical ball-by-ball data recovers the observed mean and variance trajectories, and Model 2 shows how lost wickets perturb the win probability and then allow it to relax back."],"supporting_citations":[{"why":"Supplies the Ornstein-Uhlenbeck process whose solution and conditional moments give Equations 13–15.","marker":"[11]"},{"why":"Provides the Brownian-motion model of sports-score progression that motivates treating the lead as a diffusion with drift and volatility.","marker":"[3]"},{"why":"Introduces the idea of a sports game having an implied volatility, supporting the interpretation of sigma as unpredictability.","marker":"[2]"},{"why":"Supplies the premise that sports outcomes are inherently limited in predictability, motivating a probabilistic rather than deterministic model.","marker":"[1]"}],"fun_headline_variants":["Math models cricket: one equation predicts win chance","Ornstein-Uhlenbeck cricket: three numbers decide matches","Cricket's winning probability is an error function","Stochastic differential theory cracks cricket's outcome","One SDE for cricket: from runs to win probability"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole calculation rests on assuming that $X_t$ follows the Ornstein-Uhlenbeck process $dX_t = x_0(x_1 - X_t)dt + \\sigma dW_t$ with constant, historically fitted parameters, so that the final lead is conditionally normal with the mean and variance from Eqs. 13–14; if real match dynamics are not mean-reverting with constant volatility, Eq. 15 is not the true win probability.","fun_headline_variants_meta":{"raw":{"variants":["Math models cricket: one equation predicts win chance","Ornstein-Uhlenbeck cricket: three numbers decide matches","Cricket's winning probability is an error function","Stochastic differential theory cracks cricket's outcome","One SDE for cricket: from runs to win probability"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000189,"raw_usage":{"total_tokens":1245,"prompt_tokens":765,"completion_tokens":480,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":381,"completion_tokens_details":{"reasoning_tokens":405}},"tokens_in":381,"tokens_out":480,"duration_ms":4977,"temperature":1.0,"reasoning_tokens":405,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:48:49.227664+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a large sample of real matches and, for many time points $t$ and observed leads $\\alpha$, compare the empirical frequency of wins with Eq. (15) evaluated at the fitted $(x_0, x_1, \\sigma)$. If the residuals $X(1) - E[X(1) \\mid X(t)=\\alpha]$ are skewed or heavy-tailed, or if the fitted parameters drift systematically across opponents, the central claim is falsified.","supporting_citations":[{"cited_title":"Stochastic volatility with an OrnsteinUhlenbeck process: an extension","cited_arxiv_id":null,"evidence_quote":"Supplies the Ornstein-Uhlenbeck process whose solution and conditional moments give Equations 13–15."},{"cited_title":"A Brownian motion model for the progress of sports scores","cited_arxiv_id":null,"evidence_quote":"Provides the Brownian-motion model of sports-score progression that motivates treating the lead as a diffusion with drift and volatility."},{"cited_title":"The implied volatility of a sports game","cited_arxiv_id":null,"evidence_quote":"Introduces the idea of a sports game having an implied volatility, supporting the interpretation of sigma as unpredictability."},{"cited_title":"Exploring limits to prediction in complex social systems","cited_arxiv_id":null,"evidence_quote":"Supplies the premise that sports outcomes are inherently limited in predictability, motivating a probabilistic rather than deterministic model."}],"review_version":1}