Tennis match prediction has been studied extensively, yet the literature offers no controlled comparison of Elo ratings, classical machine learning, and deep neural networks under identical experimental conditions, leaving practitioners without clear guidance on model selection. We address this gap with a unified empirical study on 133,138 professional men’s tennis matches from the Association of Tennis Professionals tour (1968–2024). Four approaches are evaluated on the same temporally split data with a common 16-feature set and an aligned evaluation protocol: an enhanced Elo rating system, ten classical machine learning algorithms, seventeen deep neural network configurations spanning 207,000 to 21,000,000 parameters, and a hybrid Elo–machine learning (ELO-ML) approach that augments classical learners with three Elo-derived features. A tuned Elo baseline alone reaches 65.87% accuracy, the best of ten classical machine learning algorithms reaches 66.30%, seventeen deep neural network configurations cluster at 66.15–66.22%, and the hybrid ELO-ML approach reaches 67.52% (McNemar’s test, $p<0.001$ for all ELO-ML pairwise comparisons). All four approaches sit within a 1.65 pp band whose upper edge lies below the 70–72% accuracy commonly cited for bookmaker odds, indicating that pre-match prediction under universally available features is a difficult task in which Elo alone already captures most of the predictable signal and algorithmic sophistication adds only marginal headroom. Deep neural networks deliver substantially better probability calibration than the other approaches (Expected Calibration Error 0.0077 vs. 0.0142). Model capacity exhibits sharply diminishing returns: all seventeen network configurations, spanning a 100-fold range in parameter count (207,000 to 21,000,000), fall within a 0.07 pp accuracy band. The study establishes a controlled benchmark for tour-level tennis prediction, quantifies how narrow the headroom above Elo actually is, provides modest but consistent empirical support for the statistically enhanced learning framework, and supplies deployment-ready operating points for sports analytics practitioners.

Tennis match prediction has been studied extensively, yet the literature offers no controlled comparison of Elo ratings, classical machine learning, and deep neural networks under identical experimental conditions, leaving practitioners without clear guidance on model selection. We address this gap with a unified empirical study on 133,138 professional men’s tennis matches from the Association of Tennis Professionals tour (1968–2024). Four approaches are evaluated on the same temporally split data with a common 16-feature set and an aligned evaluation protocol: an enhanced Elo rating system, ten classical machine learning algorithms, seventeen deep neural network configurations spanning 207,000 to 21,000,000 parameters, and a hybrid Elo–machine learning (ELO-ML) approach that augments classical learners with three Elo-derived features. A tuned Elo baseline alone reaches 65.87% accuracy, the best of ten classical machine learning algorithms reaches 66.30%, seventeen deep neural network configurations cluster at 66.15–66.22%, and the hybrid ELO-ML approach reaches 67.52% (McNemar’s test, p < 0.001 for all ELO-ML pairwise comparisons). All four approaches sit within a 1.65 pp band whose upper edge lies below the 70–72% accuracy commonly cited for bookmaker odds, indicating that pre-match prediction under universally available features is a difficult task in which Elo alone already captures most of the predictable signal and algorithmic sophistication adds only marginal headroom. Deep neural networks deliver substantially better probability calibration than the other approaches (Expected Calibration Error 0.0077 vs. 0.0142). Model capacity exhibits sharply diminishing returns: all seventeen network configurations, spanning a 100-fold range in parameter count (207,000 to 21,000,000), fall within a 0.07 pp accuracy band. The study establishes a controlled benchmark for tour-level tennis prediction, quantifies how narrow the headroom above Elo actually is, provides modest but consistent empirical support for the Statistically Enhanced Learning framework, and supplies deployment-ready operating points for sports analytics practitioners.

Large language models are increasingly used to interpret free-text clinical case narratives, yet small, clinically plausible documentation changes can shift predicted diagnoses. In this paper, we propose a model-agnostic evaluation framework that generates constrained counterfactual rewrites of deidentified case reports, filters candidates with complementary lexical-overlap and sentence-embedding similarity gates. Free-text predictions are adjudicated with alias-aware fuzzy matching and graded clinical criteria, including a conservative unknown category that avoids over-crediting generic answers. The framework supports three evaluation modes: single-turn baseline inference, adversarial single-turn stress testing on filtered rewrites, and multi-turn simulations that reveal evidence incrementally while tracking instability signals. Across a corpus of published case-report narratives evaluated on both proprietary and open endpoints, exact or alias-matched accuracy is modest on clean inputs and robustness worsens under clinically plausible reframing. Errors are driven mainly by non-adjudicable or off-granularity outputs rather than direct selection of distractor diagnoses. This work provides a reproducible procedure for testing diagnostic language models for stress before deployment.

Exponentially weighted moving average (EWMA) charts are frequently employed in statistical process control, leveraging exponential decreasing weights on historical observations to detect shifts in the process mean. The triple exponentially weighted moving average (TEWMA) chart extends this approach by applying three sequential weightings to increase responsiveness. In this paper, a theorem (theorem1) is presented, offering a universal method for constructing chart statistics in the EWMA (k) X chart framework. The theorem simplifies the multi-layer EWMA calculations and clarifies how these statistics evolve. Building on this theoretical basis, a parallel-processing simulation algorithm was developed to accommodate any EWMA (k) X chart configuration efficiently across a wide range of parameters and shift magnitudes. A subsequent comparative study evaluated how effectively optimized EWMA and TEWMA charts minimize the out-of-control average run length (OOC ARL) under a fixed in-control average run length (IC ARL). Applied first to synthetic data and then to real-world annual flow volume measurements of the Nile River (1871-1930), the optimized EWMA chart consistently detected shifts more rapidly than its TEWMA counterpart, yielding shorter OOC ARLs and demonstrating greater robustness. These findings highlight the practical applicability of theorem in designing multi-layer EWMA charts and the advantages of the optimized EWMA approach for detecting process mean shifts.