TennisEdge's AI model finds no betting value in J. Munar vs A. Blockx (Montreal on 2026-08-06) — the market is efficient, with A. Blockx favoured at 1.68.
Independent estimates of who wins, each from a different method. The blend is what the bot prices against the market.
| Method | J. Munar | A. Blockx |
|---|---|---|
| Elo ratings-based, surface-weighted |
55.5% | 44.5% |
| Machine learning 29-feature gradient-boosted model |
54.7% | 45.3% |
| Serve / return point-by-point simulation |
27.8% | 72.2% |
| Blend what the bot actually prices off |
44.5% | 55.5% |
| Market implied de-vigged from the opening odds |
42.0% | 58.0% |
We publish how well these score against the closing market, including where they lose: how accurate are tennis predictions?
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 13, 2026.
| W | R. Hijikata | 2 - 0 | Aug. 2, 2026 |
| L | D. Stricker | 2 - 0 | July 13, 2026 |
| L | J. Lehecka | 3 - 1 | July 4, 2026 |
| W | J. Fearnley | 0 - 3 | July 2, 2026 |
| W | F. Cerundolo | 3 - 0 | June 30, 2026 |
| L | L. van Assche | 2 - 1 | July 26, 2026 |
| W | L. Darderi | 2 - 1 | July 25, 2026 |
| W | R. A. Burruchaga | 1 - 2 | July 24, 2026 |
| W | K. Jacquet | 1 - 2 | July 22, 2026 |
| L | T. Droguet | 2 - 1 | July 14, 2026 |
All Aug. 6, 2026 results & picks →
J. Munar (A) vs A. Blockx (B) — ATP Montreal (hard), 1/32-finals ═══ MODELS vs MARKET ═══ ELO : Munar 55.5% ML : Munar 54.7% Sim : Munar 27.8% (raw resimulate baseline; blended model_prob_a 44.5%) Betfair : A 2.36 / B 1.67 -> comm-adj implied Munar 43.6% Pinnacle : 2.33 / 1.654, no-vig Munar 41.5% (fetched 2026-08-04 05:55, ~fresh) The three models are not merely disagreeing, they are 27 percentage points apart (ELO/ML 55% vs sim 28%). The market sits almost exactly in the middle at 42-44%, and the two sharp sources (Betfair, Pinnacle) agree with each other to within 1.5pp. When my instruments are that internally incoherent, the market is the sharpest single estimate on the table, and the burden of proof is on me. Which model is wrong, and why: - ELO 55.5% is surface-contaminated. Munar is ranked 38 and Blockx 36 — the two are level on ranking, so ELO's 55% for Munar is not a rank/ELO divergence story, it is a SURFACE story: Munar's rating is built on clay, his best and most-played surface. This is hard court, his weakest. No ELO/ranking red flag, but the ELO edge does not transfer to this surface. Discount it. - ELO also lags a fast-improving 21-year-old. Blockx's rating has not yet absorbed beating Darderi (1937, Elite) last week. - Sim 27.8% is over-extended. It is driven by the serve gap (Blockx hold 85.5% vs Munar 78.7%, SPW .633 vs .593) and the point-model amplifies that differential into a lopsided match probability. Both sets of inputs are also off-surface: Blockx's 50-match window is entirely clay/grass, Munar's is clay/grass plus one hard match. Neither serve number is a clean hard-court read, so I will not lean on the sim's tails either. I could not justify a bounded input correction from the data (I cannot say WHICH serve number is off or by how much — both windows are off-surface in the same direction), so per the framework I leave the baseline alone rather than free-handing a probability. ═══ RECENT FORM (5 most recent each) ═══ J. MUNAR (29, rank 38) 08-02 Montreal hard W vs R. Hijikata (1734, Strong) 7-6 6-3 GRIND 07-13 Gstaad clay L vs D. Stricker (1537, Avg) 6-2 6-4 07-04 Wimbledon grass L vs J. Lehecka (1826, Strong) 6-4 6-4 4-6 6-4 07-02 Wimbledon grass W vs J. Fearnley (1747, Strong) 4-6 6.3-7.7 4-6 GRIND 06-30 Wimbledon grass W vs F. Cerundolo (1736, Strong) 6-1 6-4 6-3 DOM A. BLOCKX (21, rank 36) 07-26 Estoril clay L vs L. van Assche (1764, Strong) 6-4 4-6 7-5 (FINAL) 07-25 Estoril clay W vs L. Darderi (1937, Elite) 5-7 7-6 7-6 GRIND 07-24 Estoril clay W vs R. Burruchaga (1813, Strong) 4-6 7-5 4-6 GRIND 07-22 Estoril clay W vs K. Jacquet (1619, Avg) 6-3 4-6 6-7 GRIND 07-14 Umag clay L vs T. Droguet (1780, Strong) 3-6 6-2 6-0 ═══ SHAPE, PATH & SHARPNESS (Step 1b) ═══ MUNAR — STABLE, arguably FADING. The tell is Gstaad 07-13: a straight-sets loss to Stricker (1537, Avg) on CLAY, his own best surface. That is under-performing the bracket badly on the surface where he should be strongest. The Cerundolo DOM on grass (6-1 6-4 6-3 over a Strong opponent) is the one genuinely over-performing scoreline, but it is five weeks old. The Hijikata win two days ago needed a tiebreak to get going. PATH: 1 match at this venue on this surface, won 08-02. Munar is the match-sharp side, conditions-adjusted, 2 days rest. This is a real and correctly-identified positive. BLOCKX — SHARPENING on opposition quality, grindy on margins. Run to the Estoril FINAL: beat Darderi (1937, Elite — a bracket above him), Burruchaga (1813, Strong), Jacquet, then lost the final 7-5 in the third. Beating an Elite player is over-performing vs bracket. But every single one of those four matches went three sets, including against a 1619 — he wins ugly and he wins close. PATH: first match at Montreal (seeded, bye — Pinnacle league confirms "R2" while Munar played R64). Coming off clay. Last match 07-26, 9 days ago. TRANSITION: Blockx clay -> hard, and his entire 90d window is clay/grass with no hard-court match at all. That is a genuine transition flag. Munar also arrives off clay/grass but has already resolved it by winning a hard match here. Counterweight: 9 days is not the >21d rust bucket — it is a normal gap that gives ample hard-court practice time, and he is 21. So the honest scoreboard on the qualitative side: Munar owns the sharpness/path edge; Blockx owns the opposition-quality and shape edge. They largely cancel. Nothing here is strong enough to override two agreeing sharp prices. ═══ FATIGUE (Step 2) ═══ MUNAR (29): 1 match/14d, 3 sets, 99 min, last 08-02, back-to-back 0. Fresh. BLOCKX (21): 4 matches/14d, 12 sets, 761 min (avg 190!), back-to-back 2, last match 07-26. Blockx's raw court-time load looks alarming, but it ENDED NINE DAYS AGO. There is no extreme spot: no 150min match yesterday, no 3-in-3-days now, no same-day double, no best-of-5. Per the calibrated evidence, cumulative 14d minutes with a 9-day recovery window is context, not an edge — and he is 21, the fastest- recovering tier. The 3-set-hangover haircut (-4.5pp) explicitly does not apply: it is a day-after effect, not a nine-days-after effect. This is precisely the spot the framework warns about — a visible "he played 761 minutes" fatigue story that the market may shade and that the data says is already fully washed out. If anything that argues the value is ON Blockx, not against him. Fatigue: roughly equal. It plays no part in this decision. ═══ NARRATIVE (Step 3) ═══ Let-down: Blockx LOST a Final on 07-26 — 9 days ago, outside the 0-3d NOTE window. Not live. Venue momentum: Munar 1-0 at Montreal in window; Blockx no record. Title recency: none for either. Surface transition: Blockx clay -> hard (flagged above). Munar already played one hard match here. ═══ MARKET CHOICE (Step 4b) ═══ ML: my honest probability for Munar is ~0.44 — I discount the clay-built ELO/ML 55% and I refuse the sim's 27.8% tail, landing essentially where the blend and the market are. Edge vs comm-adj 0.4363 = +0.4%. To probe (3%) at p=0.44 I would need 2.52; to bet full (5%) I would need 2.65. Betfair offers 2.36. Nowhere close. And to reach even the probe bar I would have to believe the ELO/ML pair over Pinnacle + Betfair + the sim, on Munar's worst surface. ML A: 2.36 -> need 2.52 (probe) / 2.65 (full). PASS. ML B: 1.67 -> comm-adj implied 61.1%, above any probability I hold. PASS. HANDICAP: market line is A +1.5 @ 2.06 / B -1.5 @ 1.85 (note the model's quoted cover is for spread 2.5, not the traded 1.5 — they are not the same bet). A +1.5 @ 2.06 -> comm-adj implied 49.8%; sim has A +2.5 at only 39.0%, so A +1.5 is lower still (~35%). Heavily negative. PASS. B -1.5 @ 1.85 -> comm-adj implied 55.3%. The sim implies ~65% and that looks like a 10% edge — it is a mirage. It rests entirely on the sim's 72% Blockx win prob, the outlier both other models flatly contradict. Price the cover off a realistic ~58% Blockx win prob and -1.5 lands ~50-53%, BELOW the 55.3% you must pay. Margin-specific evidence (the only kind allowed in a cover thesis) actively contradicts the lay: (1) Blockx's own scoreline distribution — all four Estoril matches went three sets, including 7-6 7-6 vs Darderi and a third-set 7-5, and he needed three sets against a 1619; he is a grinder who wins close, not a blowout artist. (2) Munar's style/scorelines — a clay-bred returner who keeps games on the board: tiebreak off Hijikata, four sets off Lehecka, three off Fearnley. Two grinders on a neutral hard court is not a -1.5 profile. That is a self-negating cover thesis, and G3 is a hard veto on margin bets. AH quote for the record: A +1.5 @ 2.06 (pass), B -1.5 @ 1.85 (vetoed). TOTALS: 22.5, over 1.98 / under 1.96. Sim under 55.5% vs comm-adj 52.3% = +3.2%, under the 5% totals bar (totals get no probe tier). It also depends on a 3.0-game downward bias correction applied to a 24.6-game raw expectation to land just under the line — far too fine a distinction to stake. And the same grinder-vs- grinder read that killed the handicap argues OVER, against the sim. PASS. Best available: nothing clears its bar. No market staked. ═══ GATES (Step 4a) ═══ GATES G1 edge-origin: FAIL — best pre-narrative ML edge vs comm-adj Betfair 0.4363 is +0.4% at my honest p=0.44, below even the 3% probe floor; the apparent 11% ELO/ML edge is clay-contaminated rating that sim, Pinnacle and Betfair all contradict. G2 consensus: PASS (stake-tier only, never blocks) — would be lone_signal/capped anyway: models split 2-1 with a 27pp internal spread and no independent on-surface H2H or live-event evidence. G3 fatal-risk: FAIL (margin only) — B -1.5 cover thesis is self-negated by Blockx's own all-three-set win distribution incl. vs a 1619; hard veto on handicap/totals. ═══ VERDICT: PASS ═══ Two adjacent-ranked players, three models 27 points apart, and two sharp prices agreeing within 1.5pp in the middle of them. The one clean qualitative edge I have — Munar match-sharp on the surface vs a seed making his hard-court debut off a clay final — is real but is answered by Blockx's superior recent opposition quality and by Munar arriving on his weakest surface in FADING shape after losing straight sets to a 1537 on clay. Backing Munar needs 2.52 and the exchange pays 2.36. No bet at any price on offer.