TennisEdge's AI model finds no betting value in D. Svrcina vs T. Droguet (Montreal on 2026-08-01) — the market is efficient, with T. Droguet favoured at 1.61.
Independent estimates of who wins, each from a different method. The blend is what the bot prices against the market.
| Method | D. Svrcina | T. Droguet |
|---|---|---|
| Elo ratings-based, surface-weighted |
54.0% | 46.0% |
| Machine learning 29-feature gradient-boosted model |
58.6% | 41.4% |
| Serve / return point-by-point simulation |
57.2% | 42.8% |
| Blend what the bot actually prices off |
52.4% | 47.6% |
| Market implied de-vigged from the opening odds |
42.3% | 57.7% |
We publish how well these score against the closing market, including where they lose: how accurate are tennis predictions?
| Player | Surface | Elo | Record | Matches |
|---|---|---|---|---|
| D. Svrcina | clay | 1643 | 100–83 | 183 |
| T. Droguet | clay | 1787 | 133–65 | 198 |
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 14, 2026.
T. Droguet leads 1–0 over 1 meeting. Full record.
| Date | Event | Surface | Winner | Score |
|---|---|---|---|---|
| Aug. 1, 2026 | Montreal | hard | T. Droguet | 2 - 1 |
| L | T. Droguet | 1 - 2 | Aug. 1, 2026 |
| L | D. Shapovalov | 2 - 0 | July 31, 2026 |
| W | L. Darderi | 0 - 0 | July 30, 2026 |
| W | A. Walton | 2 - 1 | July 29, 2026 |
| L | G. Dimitrov | 1 - 2 | July 14, 2026 |
| W | L. van Assche | 1 - 2 | Aug. 3, 2026 |
| W | D. Svrcina | 1 - 2 | Aug. 1, 2026 |
| L | H. Gaston | 2 - 0 | July 22, 2026 |
| W | C. Ugo Carabelli | 2 - 1 | July 20, 2026 |
| L | D. Merida Aguilar | 2 - 0 | July 16, 2026 |
All Aug. 1, 2026 results & picks →
RE-CHECK of earlier LEAN — Svrcina (A) vs Droguet (B), Montreal, hard, Final.
MODELS vs MARKET
ELO A 54.0% | ML A 58.6% | Sim A 52.4% (blend ~55%). All three modestly favour A.
Book: A 2.20 / B 1.61 (B favourite). Betfair exchange: A 1.92 / B 1.94 — the exchange
has already corrected the soft book and prices the match near even (A comm-adj impl 53.4%).
The earlier LEAN rested on value_edge 0.10 computed vs the stale book 2.20; against the
executable exchange price that edge is gone.
EDGE (executable Betfair, 5% commission)
A @1.92 -> eff 1.874, impl 53.4%. Edge: ELO +0.7%, sim -1.0%, blend +1.6%, only the
ML outlier +5.2%. min_odds at 55% = 1.91. B @1.94 impl 52.8% -> models give B ~45%,
edge negative every way. No side clears the 5% bar off the coherent read.
RECENT FORM (last 5)
A Svrcina (hard, Los Cabos, match-sharp on surface): L Shapovalov(1817) 6-4 6-2 (competitive
loss to Strong); W Darderi(1610) 3-0 (ret/walkover); W Walton(1756) 3-set GRIND (2 TBs);
prior clay block weak (early exits Bastad/Braunschweig). Shape STABLE on hard.
B Droguet (ALL clay recently -> TRANSITION to hard + 10-day layoff = rust risk):
L Gaston 6-0 6-3 (heavy); W Ugo Carabelli(1809,Elite) 3-set grind; L Merida(1757) 6-1 6-3.
Shape STABLE/mixed on clay, unproven on this surface.
FATIGUE (Step 2)
A: 3 matches in 3 consecutive days (29-31 Jul), 261 court-min, back_to_back 2, played
YESTERDAY — an EXTREME 3-in-3 spot -> genuine disadvantage. But lost yesterday in 2 sets,
so NO 3-set-win hangover haircut applies. Age 23 = baseline. Net: A carries real fatigue.
B: 10 days rest (last 22 Jul), matches_7d 0 — fresh but rusty. Fatigue points AGAINST A,
which cuts against enlarging the already-thin A edge, and the exchange is NOT over-fading
A (already near even), so no fatigue-value angle survives.
NARRATIVE
B TRANSITION clay->hard = level-depress/rust for B (supports A ML directionally) but does
NOT create numeric edge. No fresh Final let-down signal. ELO~ranking roughly aligned (both
ranked ~110), ELO credible. alt_markets: handicap/totals NOT quoted; only betfair_ml present.
AH: not quoted.
GATES
G1 edge-origin: FAIL — before any fatigue/narrative adjustment the model-vs-exchange edge is
~+0.7% (ELO) to +1.6% (blend); only the lone ML outlier clears 5% and fatigue argues to
shade A DOWN, not up. No pre-adjustment edge >= 5%.
G2 consensus: n/a (stake-tier only) — all models + exchange broadly agree A ~53-55%.
G3 fatal-risk: n/a (ML, not a margin bet).
VERDICT: G1 fails. Exchange has converged with the models; the book-based LEAN edge is stale.
No executable edge >= 5% either side. PASS.