TennisEdge's AI model finds no betting value in A. Vukic vs Z. Svajda (Washington on 2026-07-27) — the market is efficient, with Z. Svajda favoured at 1.57.
═══ A. Vukic (A) vs Z. Svajda (B) — Washington, hard ═══
MODELS / MARKET (all opponent-adjusted views converge on Vukic ~46%)
ELO A 46.8% | ML A 47.0% | blend A 46.5% | book A 2.40 (impl 41.7%) | Betfair A 2.22 / B 1.78
Betfair comm-adj: A 46.3% / B 57.4%. Point-sim (biased, see below) A 54.2%.
value_pick A edge 6.95% is vs the SOFT BOOK (2.40); the executable exchange (2.22)
erases it — model A 46.5% vs comm-adj 46.3% = +0.2% edge. Stale quote, not value.
SIM DISCOUNT (why the 54% is an outlier — field-quality bias)
resimulate inputs are unweighted, NOT opponent-adjusted:
Vukic spw 0.612 / Svajda spw 0.613 — near-identical raw numbers.
But Vukic's were FARMED at Granby Challenger (opp 1372–1813) + Washington lead-in
(1481–1657); Svajda's were EARNED at Roland Garros (beat Cerundolo 1889, Popyrin
1797) & Wimbledon (faced De Minaur 1890, Majchrzak 1816). Same numbers, far tougher
field for Svajda → sim overrates Vukic, underrates Svajda. ELO IS opponent-adjusted,
agrees with rank (Vukic 99 vs Svajda 90) and the exchange. I trust the ~46% consensus
and discount the sim. Not correcting a single input (whole serve profile is inflated).
RECENT FORM (last 5 each; DOM/GRIND & counts use full window)
Vukic: SHARPENING on hard — W Washington F 6-0 6-1 vs Bolt(1641,DOM); W Wash SF 6-3 6-3
vs Dougaz(1606,DOM); W Granby F 6-4 1-6 6-1 vs Gea(1813); W Granby SF 6-4 3-6 6-3 vs
Holmgren(1657); W Granby QF 7-6 7-6 vs Galarneau(1637,GRIND). Two hard titles in 2wk,
match-sharp, margins widening. BUT this level was built vs Challenger fields.
Svajda: L Washington F 6-1 6-7 0-2 vs Hewitt(1481) — score suggests a retirement/stop
(unconfirmed); W Wash SF 7-5 6-1 vs Kozlov(1525,DOM). Thin hard sample. Real class
shows earlier: French Open R16 (beat Cerundolo 1889, Popyrin 1797), Wimb R16 vs
De Minaur. ELO/rank favour him for good reason.
FATIGUE (Step 2 calibration — cumulative minutes are CONTEXT, not an edge)
Vukic: 7 matches/14d, 758 court-min, back_to_back 2, last 07-26 — potential 3rd
consecutive match-day; age 30 (≥30 → recovers slowly, hardens one tier). A genuine
disadvantage IF a 3rd straight day, but his last two were quick blowouts (~55/75min).
Svajda: 2 matches/14d, 194 min — much fresher, but possibly banged up (final stopped 0-2).
The exchange has NOT drifted Vukic on the fatigue story (Betfair 2.22 is SHORTER than
book 2.40) — it is efficiently priced, so no "market over-prices fatigue" value to take.
NARRATIVE
Let-down: Vukic WON a Final 07-26 (1d ago) → LETDOWN spot; Svajda LOST a Final 07-26 → NOTE.
Venue: both deep at Washington-named events. Surface: both on hard, no transition.
None of this creates a numerical edge; models already reflect form.
GATES
G1 edge-origin: FAIL — vs executable Betfair, model A 46.5% vs comm-adj 46.3% = +0.2%.
Any "edge" only exists vs the stale book quote; strip narrative and there is no bar-clearing gap.
G2 consensus: FAIL — ELO, ML and the exchange all agree Vukic ~46%. The lone dissenter
(point-sim 54%) is a field-bias artifact, not independent evidence.
G3 fatal-risk: FAIL — a Vukic cover/back thesis is negated by his own 3rd-day/age-30
fatigue spot AND Svajda's unconfirmed final retirement; no clean side.
MARKET SCAN (all fail the bar)
ML A: +0.2% edge — pass. ML B: -3.9% — negative.
AH A +1.5 @1.92 (comm-adj 53.4%): needs cover >=58.4%; even the A-inflated sim only ~57%,
corrected lower → no edge. AH B -1.5 @1.99 (comm-adj 51.5%): corrected P(B by >=2)~0.50 → no edge.
Total Under 23.0 @1.99 (comm-adj 51.5%): sim under ~52.7% → +1.2%, below 5% bar.
VERDICT: PASS. Fully efficient — every opponent-adjusted model and the exchange converge
on Vukic ~46%; the only "value" is a stale book price the exchange contradicts, and the
only model disagreement is a serve-stat farming artifact. No market clears its bar.