TennisEdge's AI model finds no betting value in T. Kostovic vs M. Kubka (Warsaw on 2026-08-03) — the market is efficient, with T. Kostovic favoured at 1.91.
Independent estimates of who wins, each from a different method. The blend is what the bot prices against the market.
| Method | T. Kostovic | M. Kubka |
|---|---|---|
| Elo ratings-based, surface-weighted |
38.1% | 61.9% |
| Machine learning 29-feature gradient-boosted model |
32.6% | 67.4% |
| Serve / return point-by-point simulation |
40.1% | 59.9% |
| Blend what the bot actually prices off |
40.8% | 59.2% |
| Market implied de-vigged from the opening odds |
50.0% | 50.0% |
We publish how well these score against the closing market, including where they lose: how accurate are tennis predictions?
| Player | Surface | Elo | Record | Matches |
|---|---|---|---|---|
| T. Kostovic | clay | 1583 | 52–32 | 84 |
| M. Kubka | hard | 1620 | 118–71 | 189 |
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 16, 2026.
M. Kubka leads 2–0 over 2 meetings. Full record.
| Date | Event | Surface | Winner | Score |
|---|---|---|---|---|
| Aug. 3, 2026 | Warsaw | hard | M. Kubka | 2 - 1 |
| Feb. 28, 2025 | W15 Leimen | hard | M. Kubka | 2 - 0 |
| L | M. Kubka | 1 - 2 | Aug. 3, 2026 |
| W | V. Wolff | 0 - 2 | July 26, 2026 |
| W | A. Siskova | 0 - 2 | July 25, 2026 |
| W | J. Pastikova | 0 - 2 | July 24, 2026 |
| W | K. Supova | 0 - 2 | July 23, 2026 |
| W | T. Kostovic | 1 - 2 | Aug. 3, 2026 |
| W | J. Mikulskyte | 2 - 1 | July 19, 2026 |
| W | K. Pavlova | 2 - 1 | July 18, 2026 |
| W | H. Shi | 2 - 1 | July 17, 2026 |
| W | C. Noel | 2 - 0 | July 16, 2026 |
All Aug. 3, 2026 results & picks →
═══ Kostovic (A) vs Kubka (B) — Warsaw, hard ═══ MODELS vs MARKET ELO: A 38.1% / B 61.9% ML: A 32.6% / B 67.4% Blend: A 40.8% / B 59.2% Book: 1.91/1.91 (B implied 52.4% post-margin, 50% de-margined). Models lean B; blended edge on B vs book price = 6.8% (ELO 9.5%, ML 15% — wide model disagreement). NO exchange price (betfair_ml null, all alt_markets null) → book-priced bar applies (edge >= 8%). ELO/RANKING SANITY A (Kostovic) is the HIGHER-ranked player (178) yet models favour B (Kubka, 270). Strong-rank/low-ELO pattern on A → ELO likely UNDERSTATES A; do NOT back B purely on the ELO/model gap. Edge discounted. RECENT FORM (last 5 each) A Kostovic (19): just WON W100 Amstetten title 07-26 (5-0 run, opp ELO all 1500/unrated). Prior: Rome 1/8 loss to Bronzetti (clay), Rome win vs Ruggeri 6-2 6-3. Title winner, match-sharp base but on unrated fields; last match 7d ago (rested, no acute fatigue). B Kubka (25): WON W75 Vitoria Gasteiz title 07-19 (5-0, opp ELO unrated). Prior hard: W35 Don Benito SF loss to Ryser, wins over weak fields (ELO 1049-1435). Last match 14d ago — mild rust. PATH / SHAPE Both are recent ITF title-winners stepping up to a WTA-level draw; opponent ELOs in both windows are largely 1500 placeholders → shape read is low-confidence. Neither played at Warsaw this week (both carry rust: A 7d, B 14d). No SHARPENING/FADING signal reliable off unrated fields. FATIGUE A: last match 07-26 (7d rest) — Amstetten back-to-back run is stale, no 3-set-hangover window. B: 14d idle. Both effectively rested; fatigue non-factor. GATES G1 edge-origin: FAIL — blended model edge on B = 6.8% vs book, below the 8% book-priced bar; ELO/ML disagree widely and the staked edge shrinks further once the strong-rank/low-ELO discount on A is applied. G2 consensus: stake-tier only (no block). G3 fatal-risk: n/a (ML market). MARKET CHOICE alt_markets: none at analysis time AH: not quoted VERDICT: PASS. Model edge sits below the no-exchange 8% bar, models disagree, higher-ranked player is on the fade side of the ELO read (a discount signal), and this is exactly the low-liquidity ITF-tier spot the record loses on. No stakeable edge.