TennisEdge's AI model finds no betting value in O. Selekhmeteva vs A. Ito (Toronto on 2026-08-01) — the market is efficient, with O. Selekhmeteva favoured at 1.28.
Independent estimates of who wins, each from a different method. The blend is what the bot prices against the market.
| Method | O. Selekhmeteva | A. Ito |
|---|---|---|
| Elo ratings-based, surface-weighted |
59.8% | 40.2% |
| Machine learning 29-feature gradient-boosted model |
53.1% | 46.9% |
| Serve / return point-by-point simulation |
50.2% | 49.8% |
| Blend what the bot actually prices off |
59.3% | 40.7% |
| Market implied de-vigged from the opening odds |
71.4% | 28.6% |
We publish how well these score against the closing market, including where they lose: how accurate are tennis predictions?
| Player | Surface | Elo | Record | Matches |
|---|---|---|---|---|
| O. Selekhmeteva | clay | 1627 | 102–59 | 161 |
| A. Ito | hard | 1631 | 140–78 | 218 |
Our own ratings, computed from match results since 2018. Players with fewer than 5 rated matches are omitted rather than shown at a provisional starting value. Last updated July 13, 2026.
A. Ito leads 1–0 over 1 meeting. Full record.
| Date | Event | Surface | Winner | Score |
|---|---|---|---|---|
| Aug. 1, 2026 | Toronto | hard | A. Ito | 2 - 1 |
| L | A. Ito | 1 - 2 | Aug. 1, 2026 |
| L | E. Yaneva | 0 - 2 | July 13, 2026 |
| L | J. Vandromme | 2 - 0 | July 9, 2026 |
| W | Y. Kotliar | 0 - 2 | July 7, 2026 |
| L | E. Navarro | 2 - 1 | July 2, 2026 |
| L | K. Sebov | 0 - 2 | Aug. 2, 2026 |
| W | O. Selekhmeteva | 1 - 2 | Aug. 1, 2026 |
| L | Y. Ma | 2 - 1 | July 26, 2026 |
| W | A. Sharma | 1 - 2 | July 25, 2026 |
| L | G. Knutson | 0 - 2 | July 18, 2026 |
All Aug. 1, 2026 results & picks →
O. Selekhmeteva (A) vs A. Ito (B) — Toronto, hard, Final
MODELS vs MARKET
ELO 59.8% A | ML 53.1% A | blend 59.3% A. Betfair A 1.42 (~71.5% comm-adj), B 3.1.
Models make it near the ranking (84 vs 86 = coin-flippish, A modest fav); the MARKET
is the outlier making A a ~70% favourite. So the "edge" is: models like B more than
the market (blended B 40.7% vs comm-adj implied 33.4% => +7.3% on B). Question is
whether that model edge is real or an artifact.
RECENT FORM (last 5 each)
A (Selekhmeteva): Rome L 3-6 2-6 (Yaneva 1567, clay, blowout); Contrex L 6-0 6-1
(Vandromme 1458); Contrex W over Kotliar 1485; Wimb L in 3 to Navarro 1835; Wimb
W 6-1 7-5 (Kraus 1541). Deeper window: Eastbourne FINAL (grass, lost tight 3-set
2 TBs to Zakharova 1651), Strasbourg FINAL (clay), beat Kenin 1725, competitive vs
Kostyuk 1928. => genuine WTA-tour level; reaches tour finals; beats/tests 1650-1900.
B (Ito): Memphis QUALI final L to Y. Ma; Memphis quali SF W (Sharma); Prague SF L
4-6 4-6 to Knutson 1425; Athens QUALI L 2-6 1-6 (Korneeva); Athens quali final W.
Pattern: scrapping qualification/small-event draws, repeatedly losing to 1400-1500
level. Opponent ELOs mostly placeholder 1500 => B's rating built on thin/weak data.
PATH / DATA INTEGRITY
Neither player's history shows any current Toronto matches; a "Toronto Final" between
two rank-85 players is implausible for a WTA-1000. Match lists appear stale/incomplete
=> cannot read this week's path/shape. A idle since 07-13 (~19d, rust risk if real);
B last played 07-26 (~6d, rested). Treat path as unreadable, lean cautious.
FATIGUE
A: 0 matches 14d (rested / possible rust). B: 3 matches 14d, 307min, but last match
6 days ago => rested now, no extreme spot. B age 22 recovers fast regardless. No
fatigue edge either side.
READ
Despite equal official rankings, A competes at real WTA-tour level (multiple finals,
wins over 1650-1900) while B is grinding qualification draws and losing to sub-1500
players. A is clearly the sharper player right now; the market's ~70% on A is well
justified. The model's 59% (and the resulting B "value") is dragged by A's couple of
ugly losses plus B's stale/placeholder ELO. The qualitative form read CONTRADICTS the
model edge and VALIDATES the market. Fading the model, not the market.
GATES
G1 edge-origin: FAIL — numeric +7.3% on B exists, but it is a model artifact (stale
B ELO / A ELO dragged by blowout losses) that the form read actively refutes; no
trustworthy model-vs-market edge survives.
G2 consensus: n/a (stake-tier only) — would be lone-ish/contaminated (placeholder
opponent ELOs) => capped tier even if bet.
G3 fatal-risk: n/a — ML market, no margin bet.
VERDICT: PASS. Model likes B but A is the demonstrably better/sharper player and the
market price is fair; the B edge is a rating artifact I won't stake. No A value either
(market fair, model says A overpriced). Also material data-integrity doubt on the fixture.