TennisEdge's AI model finds no betting value in K. Majchrzak vs T. Paul (Washington on 2026-07-27) — the market is efficient, with T. Paul favoured at 1.25.
K. Majchrzak (A, #59, age 30) vs T. Paul (B, #20, age 29) — Washington, HARD, R32. MODELS (prob A / prob B) ELO: 26.75% / 73.25% (matches market AND ranking → credible) ML: 42.2% / 57.8% (OUTLIER, inflated) Blend: 29.6% / 70.4% Betfair ML: 26.7% / 74.1% (3.75 / 1.35) Resim sim: 35.5% / 64.45% (fresher 50-match/730d engine; margin -1.6g, total 22.2) Two camps: "Majchrzak live" = ML 42% + sim 35.5%; "Majchrzak clear dog" = ELO 27% + market 27% + rank (#59 vs #20). ELO agrees with rank and market → ELO is the trustworthy model; ML/sim are the outliers. WHY THE OUTLIERS ARE INFLATED (surface): Majchrzak's recent body of work is a GRASS title run (won Hertogenbosch beating De Minaur 1910, Medvedev 1858, FAA 1745) — genuinely elite grass form. But this match is HARD, which is Paul's BEST surface and NOT Majchrzak's. The sim's serve/return inputs are 2yr opponent/surface-unadjusted blends: Majchrzak's carry grass uplift, Paul's carry clay drag (mixed Roland Garros/Hamburg). On hard the true gap is wider than the sim's near-even matchup (0.535 vs 0.563) implies. So the sim/ML over-credit Majchrzak here; market/ELO are closer to right. RECENT FORM (last 5, all GRASS — no recent hard for either) Majchrzak: L Svajda(1769) 5-set; W Tabilo(1460) 6-3 7-5 7-5 DOM/Weak; L Lehecka(1783) straights; W De Minaur(1910) 6-3 2-6 7-6 GRIND/Elite [TITLE]; W Medvedev(1858) 7-6 6-1 GRIND/Elite. Shape SHARPENING — but on grass only. Paul: L Hurkacz(1773) 4-set; W Kwon(1610) 6-3 7-6 6-2; W Muller(1498) 6-1 6-2 6-1 DOM/Weak; (2 grass exos). Made London grass Final + Wimbledon R32. Shape STABLE, and hard is his best surface. PATH & FATIGUE: Both first match at Washington (no venue matches yet). Both ~24-25d off since Wimbledon (07-02 / 07-03) AND both grass->hard TRANSITION → rust + looser margins cancel between the two. Fatigue objects both well_rested/zeroed — no edge either side. Ages 30/29, neutral. MARKETS (edges vs commission-adjusted executable price, fresher resim distribution) A ML @3.75: sim 35.5% vs 27.7% -> +7.9% nominal (min 2.95) — but rests on disputed sim/ML; PASS on validation B ML @1.35: 64.45% vs 75% -> -10.6%, negative A +3.5 @2.0: cover 52.6% vs 51.3% -> +1.3%, below bar B -3.5 @1.9: cover 47.4% vs 53.9% -> -6.5% (high-variance lay, needs 10%) Over 22.5 @1.97: 48.1% vs 52.0% -> -4.0% Under 22.5 @1.94: 51.9% vs 52.8% -> -0.9% (Note: brief alt_markets.model was staler — B-3.5 0.578 / Under 0.561 — resim engine supersedes it; every market still fails its bar.) GATES G1 edge-origin: FAIL — the only numeric edge (A ML +7.9%) comes from the sim/ML, the two models the surface read discredits; no fatigue/narrative involved but the source itself is the disputed input, not a robust model-vs-market gap. G2 consensus: FAIL — market AND ELO (the credible model, matching rank) both sit against backing A; only sim+ML split for A, and the qualitative surface read points the OTHER way (hard = Paul's best, Majchrzak's grass form doesn't transfer). Lone outlier-model split = no bet. G3 fatal-risk: FAIL — surface mismatch is a direct negation: backing Majchrzak's competitiveness relies on grass numbers that overstate his hard-court level vs a hard-court specialist. VERDICT: PASS. Textbook efficient spot. ELO+ranking+market agree Paul is a ~74% favourite; the sim/ML that make Majchrzak look live are inflated by a grass title run that doesn't carry to Paul's best surface. No market clears its bar on the fresher distribution; the one nominal edge (A ML) fails all three gates. No bet.