TennisEdge's AI model finds no betting value in A. Shelbayh vs B. Harris (Washington on 2026-07-25) — the market is efficient, with B. Harris favoured at 1.36.
A. Shelbayh vs B. Harris — ATP Washington, Semi-finals (hard) ═══ MODELS vs MARKET ═══ ELO: 54.2% A | ML: 49.6% A | Ensemble: 49.9% A Book: 3.00 / 1.36 (implied A 31.2%) — context only Betfair exchange: 2.58 / 1.55 → commission-adj A implied ~40.0% Nominal model-market gap on Shelbayh: ~10 pts. ═══ ELO / RANKING SANITY ═══ Shelbayh is the ELO-favourite yet ranked 279 vs Harris 148 — farmed-rating flag. His ELO was built on clay challenger runs (Heilbronn title vs 1350-1458 field, Bratislava title vs 1350-1546 field). Live hard-court evidence: one match, a straight-set 6-4 6-1 loss to M. Zheng (1770) five days ago. ELO edge discounted; ML (49.6%) likely inherits the same inflated rating as a feature. ═══ RECENT FORM (last 5 each) ═══ Shelbayh (22, #279): 07-20 Bloomfield Hills hard L vs Zheng (1770) 6-4 6-1 — straight sets, one blowout set 06-24 Mallorca grass L vs Dimitrov (1779) 2-6 4-6 06-22 Mallorca grass W vs Moutet (1704) — DOM vs Strong 06-21 Mallorca "Final" L vs Polmans (1341) — lost to a Weak opponent 06-20 Mallorca grass W vs Shevchenko (1687) — DOM vs Strong Grass/clay form has flashes (two DOM wins vs Strong), but the hard-court sample is a single poor loss and the Polmans defeat is ugly. Harris (31, #148): 07-21 Bloomfield Hills hard L vs Suresh (1603) 6-2 6-4 — straight sets vs Avg 06-30 Wimbledon grass L vs Khachanov (1750) in 4 competitive sets 06-25 Wimbledon W vs Zhou (1646) — GRIND (tiebreak) 06-24 Wimbledon W vs Piros (1608) — DOM vs Avg 06-22 Wimbledon W vs Napolitano (1518) — DOM vs Avg Also arrives off a flat straight-set loss to a weaker opponent. ═══ FATIGUE ═══ A: 1 match / 14d, 76 min, last 07-20. B: 1 match / 14d, 81 min, last 07-21. Roughly equal (diff <30 min, matches equal). Ages 22 vs 31 — irrelevant here, neither side carries load. No surface transition: both played hard most recently. ═══ NARRATIVE ═══ Let-down: none live (Shelbayh's Mallorca/Bratislava "finals" are stale, >30d). Venue momentum: neither has Washington history in window. No H2H in fetched data. ═══ ALT MARKETS ═══ Handicap/totals not priced (null) — ML only. GATES G1 edge-origin: PASS — 10-pt model-vs-exchange gap exists before any fatigue/narrative adjustment G2 consensus: PASS — both models sit above market on A; market is the lone dissenter G3 fatal-risk: FAIL — edge rests on a farmed clay-challenger ELO (rank 279 vs 148 divergence); only hard-court evidence is a 6-4 6-1 loss incl. a blowout set, which directly negates the dog thesis ═══ VERDICT: PASS ═══ The numeric edge is real on paper but its source is an inflated ELO with zero live hard-court validation — exactly the discount-and-pass case. Harris side has no numeric edge (market 66% vs model ~50%). No bet.