TennisEdge's AI model finds no betting value in M. Kupres vs V. Savinykh (Vancouver on 2026-07-25) — the market is efficient, with V. Savinykh favoured at 1.42.
MATCH: M. Kupres (22, rank 860) vs V. Savinykh (35, rank 1373) - Vancouver, hard, Semi-finals
MODELS vs MARKET
ELO prob A: 41.7% | ML prob A: 53.1% | Combined model: 44.0%
Book: A 2.62 / B 1.42 (implied A 35.1%) - STALE vs exchange
Betfair (executable): A 2.04 / B 1.60 -> A ~49.0%, B ~62.5% implied
Edge A: 0.440 - 0.503 = -6.3% (commission-adjusted 2.04 -> eff 1.988)
Edge B: 0.560 - 0.637 = -7.7% (commission-adjusted 1.60 -> eff 1.570)
The listed value_pick A (+8.9%) exists only against the stale book 2.62; the exchange
prices A at 2.04, which kills the bet rather than confirming it (Pricing Discipline #1).
Lone ML model (53.1% A) vs exchange 49% is ~+2.8% after commission - below bar, ELO disagrees.
RECENT FORM
Kupres (A): only 1 match in window -
07-22 W35 Saskatoon 1/16: Lost vs M. Kuramochi (ELO 1449, Weak) 5-7 1-6
Just lost to a sub-1500 opponent three days ago; thin, negative sample.
Savinykh (B): last 5 of 6 -
07-18 W75 Vitoria Gasteiz: Lost vs Mikulskyte (1500, Avg) 4-6 4-6
07-17 W75 Vitoria Gasteiz: Won vs Martinez Vaquero (1500, Avg) 6-3 6-7 1-6 (GRIND, tiebreak)
07-16 W75 Vitoria Gasteiz: Won vs Inoue (1500, Avg) 6-4 1-6 6-4 (GRIND, 3 sets)
07-15 W75 Vitoria Gasteiz: Won vs Anson Sanchez (1500, Avg) 4-6 2-6
07-08 W50 Corroios-Seixal 1/16: Lost vs Dilek (1528, Avg) 6-4 6-3
All opponents Avg-bracket; grindy scorelines, no DOM wins vs Strong/Elite.
DATA GAP: this is a Vancouver SEMI-FINAL but neither player's history contains any
Vancouver matches - this week's results (both players presumably 2-0+ here) are missing.
Form read is materially incomplete for both sides.
FATIGUE
Kupres: 1 match/14d, 72 court-min, last match 07-22 (3 days rest). Fresh (age 22, recovers fast).
Savinykh: 4 matches/14d, 356 court-min, back-to-back x3 - but last match 07-18, 7 days rest.
At 35 recovery is slower, but a full week of rest neutralises the Vitoria load.
Verdict: roughly equal - no actionable fatigue edge (and none would be stakeable anyway per G1).
NARRATIVE
Let-down: none (no recent Finals for either). Venue momentum: none on record.
Titles: none in window. Surface transition: Kupres last on hard (no transition);
Savinykh's Vitoria surfaces unknown (null) - adds uncertainty, not an edge.
ELO/ranking sanity: ELO favours Savinykh (35, rank 1373) over Kupres (rank 860) - a
veteran's ELO vs a young player's climbing rank; models split on the winner, so neither
ELO nor ML is independently trustworthy here.
ALT MARKETS: handicap/totals null - ML only.
GATES
G1 edge-origin: [FAIL] - no numeric model-vs-market edge at the executable Betfair price (A -6.3%, B -7.7%); the +8.9% 'value' is an artifact of a stale book quote
G2 consensus: [FAIL] - models split (ELO 42% A vs ML 53% A) with no hard independent evidence either way; this week's Vancouver results are missing from the data
G3 fatal-risk: [FAIL] - semifinal form for both players is invisible (no Vancouver matches in history); any thesis would rest on stale, incomplete form
VERDICT: PASS. The exchange contradicts the book's soft price on Kupres, both sides are
negative-edge at executable odds, and the form data omits the very event being played.
No stakeable market.