J. Duckworth vs A. De Minaur — Prediction & Analysis

TennisEdge's AI model finds no betting value in J. Duckworth vs A. De Minaur (Montreal on 2026-08-04) — the market is efficient, with A. De Minaur favoured at 1.10.

Model probabilities

Independent estimates of who wins, each from a different method. The blend is what the bot prices against the market.

We publish how well these score against the closing market, including where they lose: how accurate are tennis predictions?

Elo ratings

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.

Head to head

A. De Minaur leads 2–0 over 2 meetings. Full record.

Recent form

All Aug. 4, 2026 results & picks →

Full analysis

J. Duckworth (A) vs A. De Minaur (B) — ATP Montreal, hard, 1/32-finals

MODELS vs MARKET (player A)
  Book odds: A 6.85 (14.6% raw) / B 1.10 — no-vig implied A ~13.2%; brief implied_prob_a 13.8%
  ELO      : A 16.1%  -> +1.5% vs book price
  ML       : A 14.2%  -> -0.4% vs book price
  Blend    : A 21.3%  -> +6.7% vs book price (this is the whole "value_edge 7.5%")
  Raw sim  : A 42.2%  -> not credible (see SIM AUDIT)
  Pinnacle : null. Betfair exchange: null.
  ELO/ranking sanity: A ELO-dog and ranked 86; B ranked 6 with strong ELO — models and
  ranking agree, no farmed-rating divergence on either side.

SIM AUDIT (Step 4b input validation)
  GET /resimulate/ baseline: Duckworth spw .615 / rpw .410 / hold .866 over 50 matches;
  De Minaur spw .607 / rpw .433 / hold .763 over 50 matches. Sim win prob A = 42.2%.
  These inputs are opponent-unadjusted and here they are badly contaminated: Duckworth's
  serve numbers are accumulated at Challenger/ATP-250 level (Mayo 1529, Damas 1553,
  Dedura 1500, Hussey 1427 in the window), while De Minaur's are built against Tsitsipas
  1819, Nakashima 1839, Cobolli, Paul 1962, Cerundolo 1894, Darderi 1946. A hold_pct of
  .866 for the world 86 versus .763 for the world 6 is a field-quality artifact, not a
  matchup fact. The sim diverges by 26pp from ELO and 28pp from ML — divergence, not
  convergence, so per Step 4b the inputs are suspect and the sim is NOT the anchor. No
  justified numeric override exists in the dog's favour (any honest correction would push
  Duckworth DOWN, not up), so I take ELO/ML (14–16%) as the real read and treat the 21.3%
  blend as sim-inflated.

RECENT FORM (max 5 shown; all fetched matches used for classification)
  J. Duckworth (34, rank 86)
   08-03 Montreal hard  W  C. O'Connell (1715, Strong)  6-3 6-1   DOM
   07-30 Los Cabos hard L  C. Norrie (1838, Strong)     6-4 6-4   close straight-sets loss
   07-29 Los Cabos hard W  A. Mayo (1529, Avg)          3 sets    GRIND
   07-02 Wimbledon grass L F. Cobolli (1656)            2 TBs     GRIND loss
   06-30 Wimbledon grass W T. Griekspoor (1647, Strong) 4 sets    GRIND
  A. De Minaur (27, rank 6)
   07-31 Washington hard L B. Nakashima (1839, Strong)  6-7 4-6   tight loss to Strong
   07-30 Washington hard W C. Hewitt (1557, Avg)        6-2 6-3   DOM
   07-29 Washington hard W S. Tsitsipas (1819, Strong)  3 sets    GRIND, quality scalp
   07-06 Wimbledon grass L F. Cobolli (1724)            3 sets    tight
   07-04 Wimbledon grass W Z. Svajda (1799, Strong)     4 sets    GRIND

STEP 1b — SHAPE / PATH
  Duckworth: SHARPENING (6-3 6-1 DOM over O'Connell 1715 yesterday; before that a
  competitive 6-4 6-4 loss to Norrie 1838). This is his best scoreline-vs-bracket of the
  window and it is real — it is exactly why this is not a zero-interest match.
  De Minaur: STABLE (beat Tsitsipas 1819, DOM'd Hewitt, lost 6-7 4-6 to Nakashima 1839 —
  no set of that Washington run was a sub-bracket result).
  Path: A is 1 match deep at Montreal and continuous on hard (Los Cabos -> Montreal), no
  surface transition, fully match-sharp. B last played 07-31 (4 days) — recent, no rust;
  he also arrives off three hard-court matches in Washington, so conditions-adjusted too.
  Neither side has a rust or transition flag. No motivation signal in the data for either.

STEP 2 — FATIGUE
  A: 3 matches/14d, 302 court-min, back_to_back 1, played YESTERDAY — but it was a
  2-set win, so the calibrated 3-set-hangover haircut (-4.5pp) does NOT apply (2-set win
  ~neutral). Age 34 (>=30, recovers slowly) hardens the residual by one tier -> slight
  disadvantage to Duckworth. Note the direction: fatigue here works AGAINST the dog, so
  it cannot manufacture dog value.
  B: 3 matches/14d, 315 court-min, last match 07-31, 4 days rest, age 27 baseline. No
  extreme spot on either side; cumulative-minutes gap (302 vs 315) is noise.

STEP 3 — NARRATIVE
  Let-down: none for either (A's last Final is Geneva 05-17, 79d, a loss; B's is
  Hertogenbosch 06-14, 51d, a loss). Title recency: none in window for either.
  Venue momentum: A 1-0 at Montreal this week; B first match at Montreal (direct entry,
  but only 4 days off tour, so not a rust case).
  Surface transition: neither (both continuous on hard since late July).
  player_notes: only a generic Washington preview piece for De Minaur, undated — no
  injury or fitness signal on either side. Nothing to add.

MARKET CHOICE
  alt_markets: none at analysis time
  AH: not quoted
  No Betfair exchange ML, no handicap, no totals, no Pinnacle reference. The only price
  available is the book 6.85 / 1.10, which cannot be the edge baseline under Pricing
  Discipline — the no-exchange path requires edge >= 8% vs the book quote AND a
  "book-priced" flag, or PASS. Best available edge is 6.7% and it rests entirely on the
  sim-inflated blend. No exchange price also means the 3–5% probe tier is unavailable
  (probes require an executable exchange price). ML is the only market; it fails its bar.

GATES
  G1 edge-origin: FAIL — pre-narrative edge vs the only available (book) price is +6.7%
    and only from the sim-contaminated blend; ELO +1.5%, ML -0.4%. No-exchange path needs
    >=8%, probe tier needs an executable exchange price (absent). Bar not met.
  G2 consensus:   stake-tier only — would be lone_signal (blend/sim alone, ELO and ML both
    sit on the market), i.e. 0.5–1.0u cap had G1 passed. Never blocks.
  G3 fatal-risk:  n/a for ML (no veto power); the flag would be that the sole "edge"
    source is the market-contradicting sim and Duckworth is the tired, older side.

VERDICT — PASS
  A genuinely interesting dog spot on the qualitative read: Duckworth is match-sharp,
  sharpening, on his best surface and just handed a 1715 opponent 6-3 6-1, while De
  Minaur arrives off a loss. That earns a look, not a stake. Both math models that are
  opponent-adjusted (ELO 16.1%, ML 14.2%) sit right on the market at 6.85; the only
  number arguing for the dog is a simulation whose serve/return inputs are demonstrably
  inflated by the weak fields Duckworth has been beating, which is precisely the failure
  mode Step 4b warns about. With no Betfair price to execute against, no Pinnacle line to
  corroborate, and a fatigue/age tilt pointing at the dog rather than away from him, this
  is a market that looks efficient with the one dissenting model being the least
  trustworthy. Pass.