TennisEdge's AI model finds no betting value in N. Bartunkova vs B. Andreescu (Toronto on 2026-08-03) — the market is efficient, with N. Bartunkova favoured at 1.50.
Independent estimates of who wins, each from a different method. The blend is what the bot prices against the market.
| Method | N. Bartunkova | B. Andreescu |
|---|---|---|
| Elo ratings-based, surface-weighted |
47.7% | 52.3% |
| Machine learning 29-feature gradient-boosted model |
50.9% | 49.1% |
| Serve / return point-by-point simulation |
47.3% | 52.7% |
| Blend what the bot actually prices off |
52.4% | 47.6% |
| Market implied de-vigged from the opening odds |
61.0% | 39.0% |
We publish how well these score against the closing market, including where they lose: how accurate are tennis predictions?
| Player | Surface | Elo | Record | Matches |
|---|---|---|---|---|
| N. Bartunkova | clay | 1676 | 72–40 | 112 |
| B. Andreescu | hard | 1721 | 50–34 | 84 |
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 23, 2026.
N. Bartunkova leads 1–0 over 1 meeting. Full record.
| Date | Event | Surface | Winner | Score |
|---|---|---|---|---|
| Aug. 3, 2026 | Toronto | hard | N. Bartunkova | 2 - 0 |
| W | B. Andreescu | 2 - 0 | Aug. 3, 2026 |
| L | L. Tararudee | 1 - 2 | July 23, 2026 |
| W | Y. Yuan | 2 - 1 | July 21, 2026 |
| L | B. Krejcikova | 0 - 2 | July 3, 2026 |
| W | K. Siniakova | 0 - 2 | July 1, 2026 |
| L | N. Bartunkova | 2 - 0 | Aug. 3, 2026 |
| L | S. Zhang | 0 - 2 | June 29, 2026 |
| W | A. Sasnovich | 0 - 2 | June 25, 2026 |
| W | J. Teichmann | 2 - 1 | June 24, 2026 |
| W | P. Hercog | 2 - 0 | June 23, 2026 |
All Aug. 3, 2026 results & picks →
N. Bartunkova (A, age 20, #94) vs B. Andreescu (B, age 26, #137) - Toronto, hard, 1/64. MODELS vs MARKET ELO: A 47.7pct / B 52.3pct ML: A 50.9pct / B 49.1pct Sim: A 52.4pct / B 47.6pct All three cluster ~coin-flip. Betfair ML: A 1.63 (impl 62.6pct) / B 2.38 (impl 43.3pct). Market rates A a clear favourite; models split ~50/50 -> value screen flags B (+6.4pct ML edge, avg model B 49.7pct vs 43.3pct market). RECENT FORM (last 5 each) Bartunkova (A): 07-23 Prague clay L Tararudee(1502) 6-7 6-3 6-7 (GRIND loss to weak opp = UNDER-performing); 07-21 Prague clay W Yuan(1543); 07-03 Wimb grass L Krejcikova(1850) narrow; 07-01 W Siniakova(1766); 06-29 W Stearns(1470). Season pedigree strong: Birmingham Final + Rome Final, beat Keys(1908). Match-active (last match 10d ago), age 20 (rust-free, fast recovery). Transition clay->hard (risk to margins, not competitiveness). Andreescu (B): last match 2026-06-29 (L S.Zhang 1626 6-7 6-7). Prior window spotty: L Mertens 1-6 2-6; L French semi to Hruncakova(1493). Former GS champ (quality anchor keeps ELO high) but LONG LAYOFF ~34 days -> RUST RISK, first match at venue. SHAPE and PATH (Step 1b) Bartunkova: STABLE-to-mildly-FADING (ugly 3-set loss to #1502) but match-sharp with deep-run pedigree. Andreescu: rust-dominant. >21d idle = depressed level + wide margin variance in match one. Path read is the decisive qualitative factor and it runs AGAINST the side the models favour. FATIGUE (Step 2) A: matches_14d 2, last 10d ago, no extreme spot. B: fully rested (0 matches) but that rest IS the rust. Neither side carries a fatigue edge; 3-set hangover N/A. Fatigue plays no role. NARRATIVE (Step 3) No fresh Final let-down in window. Both transition to hard. Andreescu long layoff = rust, not a bettable let-down. ELO/RANK SANITY Andreescu #137 yet ELO-favourite (~52pct): high-ELO / modest-rank + 34d idle => STALE ELO overrating her. Discount that ELO edge, lean PASS unless live form independently backs it - it does not (pre-layoff form mixed). This is precisely why the model-vs-market gap exists: models are rust-blind, market prices inactivity correctly. ALT MARKETS Handicap B +2.5 @1.81 -> model covers 0.613 vs impl 0.565 = +4.8pct (BELOW 5pct bar; cover leans on same stale/rust-suspect sim inputs, and a rusty player can be blown out -> no clean margin thesis). Totals: Over @1.82 model 0.519 (edge -4.3pct), Under @1.86 model 0.481 (edge -6.9pct). No edge. AH shadow: B +2.5 @ 1.81 (handicap passed, sub-bar). alt_markets present at analysis time. GATES G1 edge-origin: FAIL - the +6.4pct B ML edge is a numeric artifact of a stale, rust-blind ELO overrating a former champ off a 34-day layoff; the path/shape read REDUCES B below market rather than confirming. No trustworthy mispricing to stake. G2 consensus: stake-tier only - moot (no qualifying edge); would have been lone_signal-capped at best (ELO-only for B; ML/sim ~even). G3 fatal-risk: n/a (no ML bet placed). VERDICT: PASS. Meaningful model-market disagreement exists but qualitative validation is negative: the side models favour (Andreescu) is the rusty side and the market is efficiently pricing inactivity the models miss. Backing B would be betting a stale-ELO artifact; backing A at 1.63 offers no value (models ~50pct vs market 63pct). No bettable edge in any market.