TennisEdge's AI model finds no betting value in J. Tjen vs L. Samsonova (Toronto on 2026-08-03) — the market is efficient, with L. Samsonova 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 | J. Tjen | L. Samsonova |
|---|---|---|
| Elo ratings-based, surface-weighted |
58.0% | 42.0% |
| Machine learning 29-feature gradient-boosted model |
42.4% | 57.6% |
| Serve / return point-by-point simulation |
79.2% | 20.8% |
| Blend what the bot actually prices off |
52.9% | 47.1% |
| Market implied de-vigged from the opening odds |
36.3% | 63.7% |
We publish how well these score against the closing market, including where they lose: how accurate are tennis predictions?
| Player | Surface | Elo | Record | Matches |
|---|---|---|---|---|
| J. Tjen | hard | 1759 | 106–31 | 137 |
| L. Samsonova | hard | 1750 | 30–32 | 62 |
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 Aug. 3, 2026.
L. Samsonova leads 3–0 over 3 meetings. Full record.
| Date | Event | Surface | Winner | Score |
|---|---|---|---|---|
| Aug. 3, 2026 | Toronto | hard | L. Samsonova | 2 - 0 |
| April 24, 2026 | Madrid | clay | L. Samsonova | 2 - 0 |
| Feb. 3, 2026 | WTA Abu Dhabi | hard | L. Samsonova | 2 - 0 |
| L | L. Samsonova | 0 - 2 | Aug. 3, 2026 |
| L | A. Kalinskaya | 1 - 2 | July 30, 2026 |
| W | R. Sramkova | 2 - 0 | July 27, 2026 |
| L | D. Kasatkina | 1 - 2 | July 1, 2026 |
| W | L. Fernandez | 0 - 2 | June 29, 2026 |
| W | J. Tjen | 0 - 2 | Aug. 3, 2026 |
| L | D. Shnaider | 2 - 0 | July 31, 2026 |
| W | Xin. Wang | 0 - 2 | July 29, 2026 |
| W | M. Keys | 2 - 1 | July 27, 2026 |
| L | M. Bouzkova | 1 - 2 | July 4, 2026 |
All Aug. 3, 2026 results & picks →
J. Tjen (A, #40, age 24) vs L. Samsonova (B, #21, age 27) — WTA Toronto 1000, hard, R1 MODELS vs MARKET (Tjen prob_a): ELO 58.0% | ML 42.4% | sim 79.2% | model 52.9% | Betfair 38.8% (2.58) / Samsonova 62.5% (1.6) The four estimates disagree wildly and split direction: ELO/sim lean Tjen, ML/market lean Samsonova. That spread is itself a warning — no coherent edge. ELO vs RANKING: Tjen is ELO-favourite (58%) yet ranked #40, BELOW Samsonova's #21. Classic high-ELO/modest-rank divergence = rating likely farmed on weaker fields. Discount the ELO edge; back it only if live form independently confirms — it does not. SIM INPUT CHECK (resimulate baseline): Tjen spw .608/rpw .427 vs Samsonova .562/.411 → sim outputs Tjen 79.2% to win. This diverges from EVERY other signal including the market by 40pts. Inputs are opponent-unadjusted 50-match averages; Tjen's are inflated (softer fields), Samsonova's diluted (top-25 schedule vs elite opposition). The sim has the winner directionally wrong, so its totals distribution (under 21.5 = 58.9%) is built on a false blowout premise and is NOT trustworthy. Correction needed far exceeds the ±0.05 clamp — cannot rehabilitate. RECENT FORM (hard court, most recent first): Samsonova (SHARPENING, match-sharp — 3 matches this week at Washington): - 07-31 Washington QF, Lost Shnaider (1836) 3-6 4-6 (2 sets, competitive vs Strong) - 07-29 Won Xin. Wang (1735) 6-2 6-2 — DOM over Strong - 07-27 Won M. Keys (1925 ELITE) 6-3 6-4 (won as 3-6 6-4 6-4) — big scalp In-form, on best surface, played into rhythm. Made Washington QF. Tjen (thin hard sample, poor season shape): - 07-30 Washington 1/8, Lost Kalinskaya (1876) 6-1 3-6 6-7(tb) — competitive vs Elite - 07-27 Won Sramkova (1681) 6-4 6-2 — solid - Prior: grass/clay run mostly LOSSES (Kasatkina, McNally, Maria, Krueger, Navarro, Osorio, Stearns). Weak overall trajectory. FATIGUE: Samsonova last played 07-31 (QF, 2-set LOSS — no 3-set hangover), 2 days rest. Tjen last 07-30 (3-set loss), 3 days rest. Neither hits an extreme spot; roughly equal, no edge either way. MARKET CHOICE: - ML: no validated edge. Tjen's model edge (ELO/sim) is a farmed-rating/biased-sim artifact contradicted by live form; qualitative read backs Samsonova = the market side. Backing Samsonova at 1.6 needs >62.5%, and ML/ELO/sim all sit below market — no value. - Totals under 21.5 @1.98: nominal 7.1% edge, BUT sourced from a sim that has the winner 40pts wrong. Two players who both go 3 sets often = real over risk the biased under-lean ignores. Untrustworthy — rejected. - Handicap: model null; resim spreads are from Tjen's (wrong-way favourite) perspective — unusable. - AH: shadow not applicable (no dog ML staked). GATES (candidate = totals under / Tjen ML): G1 edge-origin: FAIL — Tjen ML edge is farmed-ELO/biased-sim, refuted by live form, not a clean model-vs-market gap. Totals edge rests on a directionally-broken sim. G2 consensus: n/a (stake tier only) G3 fatal-risk: FAIL (totals = margin bet) — under thesis self-negates: two frequent-3-set players make 21.5 easily beatable, and the under-lean comes from the same sim that misprices the winner. VERDICT: PASS. Models diverge wildly and the two credible ones (ML + market) align with the qualitative form read (Samsonova sharper, higher-ranked, on best surface). ELO/sim edges are artifacts. No coherent, validated edge in any market.