TennisEdge's AI model finds no betting value in K. Kawa vs L. Fruhvirtova (Warsaw on 2026-08-03) — the market is efficient, with K. Kawa favoured at 1.73.
Independent estimates of who wins, each from a different method. The blend is what the bot prices against the market.
| Method | K. Kawa | L. Fruhvirtova |
|---|---|---|
| Elo ratings-based, surface-weighted |
21.1% | 78.9% |
| Machine learning 29-feature gradient-boosted model |
44.5% | 55.5% |
| Serve / return point-by-point simulation |
39.2% | 60.8% |
| Blend what the bot actually prices off |
41.4% | 58.6% |
| Market implied de-vigged from the opening odds |
54.8% | 45.2% |
We publish how well these score against the closing market, including where they lose: how accurate are tennis predictions?
| Player | Surface | Elo | Record | Matches |
|---|---|---|---|---|
| K. Kawa | clay | 1713 | 106–66 | 172 |
| L. Fruhvirtova | hard | 1624 | 94–82 | 176 |
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 22, 2026.
Level at 1–1 over 2 meetings. Full record.
| Date | Event | Surface | Winner | Score |
|---|---|---|---|---|
| Aug. 3, 2026 | Warsaw | hard | K. Kawa | 2 - 1 |
| May 16, 2022 | French Open | clay | L. Fruhvirtova | 2 - 0 |
| W | L. Fruhvirtova | 2 - 1 | Aug. 3, 2026 |
| L | P. Badosa | 0 - 2 | July 22, 2026 |
| L | P. Udvardy | 2 - 1 | July 16, 2026 |
| W | S. Waltert | 2 - 0 | July 14, 2026 |
| L | S. Waltert | 2 - 0 | July 6, 2026 |
| L | K. Kawa | 2 - 1 | Aug. 3, 2026 |
| L | A. Zakharova | 2 - 0 | July 27, 2026 |
| L | B. Krejcikova | 0 - 2 | July 21, 2026 |
| L | X. Gao | 2 - 1 | July 19, 2026 |
| W | E. Micic | 0 - 2 | July 18, 2026 |
All Aug. 3, 2026 results & picks →
Kawa (A) vs Fruhvirtova (B) — Warsaw, hard.
MODELS vs MARKET
Market: A 1.73 (54.8%) / B 2.10 (47.6% impl). Value pick B, edge 13.5%.
ELO_a 0.211 (ELO favors B ~79%), ML_a 0.445, blend model_a 0.414 (B 58.6%).
Huge ELO/ranking inversion: A ranked 134, B ranked 155, yet ELO gives A only 21%.
ELO/RANKING SANITY (the crux)
B's high ELO is FARMED + STALE: built on deep runs at weak-field events
(a "French Open" final win over Tan ELO 1376; opponent ELOs 1376-1661), and
contradicted by live form. A is strong-rank/low-ELO -> ELO understates her.
Discount B's ELO edge heavily. The 13.5% value edge is largely a math artifact
of a discredited rating.
RECENT FORM / SHAPE
A (Kawa, 33, rk134): last 5 clay/grass. DOM vs Waltert 1759 (6-1 6-4, over-perf),
competitive loss to Badosa 1830 elite (5-7 4-6). STABLE, in clay form.
TRANSITION to hard this week.
B (Fruhvirtova, 21, rk155): FADING. Lost 5 of last ~7. Both recent HARD matches
(this surface) LOST to sub-1500 players: Zakharova 1500 (07-27, 7-6 6-1),
Gao 1500 (07-14). Poor hard-court live form on the exact surface.
FATIGUE (Step 2)
A: last match 07-22 (~11d), rested. B: last 07-27 (~6d), 4 matches/14d, 357min
cumulative but no extreme spot; both rested now. Cumulative load = context only,
no fatigue edge either side.
NARRATIVE
No qualifying let-down (B lost recent finals, not won). Both surface-adjacent;
A transitions off clay, B played hard most recently but lost. No signal supports B.
PRICING
alt_markets null -> NO Betfair exchange price. Book-priced ML = the segment the
backtest loses on (Pricing Discipline #4: need >=8% AND flag, or prefer PASS).
GATES
G1 edge-origin: PASS (numeric model-vs-market gap, not fatigue) BUT the edge is
driven by a farmed/stale ELO that live hard-court form contradicts.
G2 consensus: stake-tier only; would be lone_signal-tier at best (contaminated ELO).
G3 fatal-risk: n/a (ML market).
VERDICT: PASS. The only edge (value pick B) rests on a discredited ELO; the
qualitative read (B FADING, two recent hard-court losses to weak players) OPPOSES
rather than validates backing B. Market's 47.6% on B is already fair-to-generous.
Backing A has no model support (ML/ELO both below market). No exchange price.
No bettable edge either direction. PASS.