TennisEdge's AI model finds no betting value in A. Potapova vs V. Williams (Washington on 2026-07-27) — the market is efficient, with A. Potapova favoured at 1.20.
═══ A. Potapova (25, #38) vs V. Williams (46, #582) — Washington, hard, 1/16-finals ═══
MODELS vs MARKET
ELO: Potapova 90.8% | ML: 78.3% | sim (baseline): 72.2% | Betfair: 86.2% (1.16) / Venus 15.6% (6.4)
Book 1.20/4.50 (soft, shades Venus's name shorter); sharp exchange lengthens Venus to 6.4.
Big model spread. Sim is the LOW outlier on Potapova (72%) i.e. it rates Venus (28%) far above
everyone (market 15.6%, ELO 9%, ML 22%). That divergence flags the sim's Venus inputs, not an edge.
ELO/RANKING SANITY
Potapova #38 + ELO-fav → consistent, credible. Venus ranked 582 but models give her 9-28% on
pedigree/high historical ELO — ranking understates her, ELO/sim over-state her. Huge two-sided
uncertainty on a 46yo comeback; do NOT lean into either direction.
RECENT FORM (last 5 shown; all fetched used for classification)
Potapova — grass swing poor: L Bouzas Maneiro 6-2 6-3 (Wimb, 1674), L Alexandrova (ret, 1764),
L Sonmez 1-6 0-2 (ret, 1673); GRIND W vs Lamens 7-6 6-4 6-4 (1523); L Kalinskaya 4-6 6-2 6-7
(RG, 1785). Clay swing was elite though (Rome title 05-05, beat Gauff 2055 at RG). Now grass→
HARD TRANSITION, first hard match, ~4wk rust. Shape mixed/uncertain.
Venus — one match in window: L Begu 2-6 6-4 6-7 (Bad Homburg grass, 1627). Competitive 3-set loss,
took a set. But 46yo, sporadic play, ~5wk rust, grass→HARD transition. Sample of one.
FATIGUE
Both well-rested (0 recent matches). No extreme spot → non-factor. Both actually carry RUST RISK
(>21d idle; Venus especially at 46). Path: neither has played Washington — first match for both.
NARRATIVE
No let-down (Rome final 05-05 far outside window). Surface TRANSITION for BOTH (grass→hard) —
widens margin variance both ways, which cuts against any tight-cover thesis.
SIM VALIDATION & RE-RUN (handicap candidate)
Baseline sim: Venus +5.5 covers 67.3% → apparent +20% edge at Betfair 2.1. But baseline win prob
(Venus 28%) diverges hard from sharp market (15.6%) and ELO (9%). Venus's serve inputs (SPW .564,
hold .675) are built on a THIN 10-match sample at 46, likely inflated by grass/weaker-field play —
classic sim over-rating. Corrected spw_b .564→.52 (−0.044, market-validated, within ±0.05):
win probs CONVERGE to market/ELO (Potapova 72%→88%, Venus 28%→12%), expected margin 2.9→4.8,
and Venus +5.5 cover collapses 67.3%→49.8%. The handicap "edge" was a pure sim artifact.
MARKET-BY-MARKET (executable prices, 5% commission)
ML Potapova @1.16: eff 1.152, edge ~1% (corr 88% vs 86.8%); min_odds 1.193 > 1.16 → below bar. PASS.
ML Venus @6.4: eff 6.13 (16.3%) vs corr 12% → negative. PASS.
AH B +5.5 @2.1: corr cover 49.8%, eff 2.045 (48.9%) → edge ~1%. PASS.
AH A -5.5 @1.83: corr cover 50.2%, eff 1.789 (55.9%) → negative. PASS.
Totals: line 19 pick'em (1.98/1.96); corrected sim ~18 games leans under ~6% BUT that is an
artifact of the exact correction magnitude (1-game precision I can't defend); market line is
~right. PASS.
GATES (evaluated on best candidate: Venus +5.5)
G1 edge-origin: FAIL — the 20% edge exists ONLY under the baseline sim; a modest market-validated
input correction erases it to ~1%. Not a robust model-market disagreement.
G2 consensus: FAIL — market (15.6%), ELO (9%), ML (22%) all rate Venus far below the baseline
sim (28%); only support is one 5-week-old grass match. No hard independent evidence.
G3 fatal-risk: FAIL — rusty 46yo vs a sharp top-40 hitter can easily go 6-2 6-1 (margin 9),
directly negating a +5.5 cover. Units cannot buy this back.
VERDICT: PASS. Every market's apparent edge dissolves once the sim's over-rating of Venus is
corrected to market/ELO consensus. No reliable edge at any executable price. Predicted winner:
A. Potapova (~86%).