F. Tiafoe vs M. Cilic — Prediction & Analysis

TennisEdge's AI model finds no betting value in F. Tiafoe vs M. Cilic (Montreal on 2026-08-04) — the market is efficient, with F. Tiafoe favoured at 1.56.

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 July 28, 2026.

Recent form

All Aug. 4, 2026 results & picks →

Full analysis

F. TIAFOE vs M. CILIC — ATP Montreal, hard, 1/32-finals (R2)

═══ MODELS vs MARKET ═══
  ELO (=model):    Tiafoe 57.6%
  ML:              Tiafoe 79.4%
  Sim (baseline):  Tiafoe 62.2%  (spw_a .596 / rpw_a .424 vs spw_b .593 / rpw_b .403; 50/47 matches)
  Betfair ML:      A 1.56 / B 2.58  -> A 60.8% raw, 65.3% commission-adjusted
  Pinnacle:        1.575 / 2.51, no-vig A 61.4%, drift_a 0.0pp (fetched 10:55Z, ~1h old)

The sim (62.2%) lands almost exactly on both market anchors (60.8% / 61.4%) and between the
two models. ML at 79.4% is the outlier (classic over-confident ML on a name-favourite);
ELO at 57.6% is the only side-B-friendly number and it is the one I distrust here — see
ELO/ranking note. Sim-market convergence = no calibration needed, sim inputs accepted.

ELO/ranking sanity: Cilic is ranked 47 yet ELO implies he is a 42% dog to a rank-30 player
in decent form — that reads as an ELO still carrying residue from a much better era while
the live record shows losses to Kouame (1418) and Landaluce (1579). ELO OVERSTATES Cilic,
so the lone model that likes the dog is the contaminated one. Edge discounted, not enlarged.

═══ RECENT FORM (5 most recent each; all fetched matches used for counts) ═══
TIAFOE (A) — 28, rank 30, hard:
  07-28 Washington R16  L  T. Atmane (1738)     6-4 3-6 4-6   competitive 3-set loss
  07-04 Wimbledon R16   L  A. Bublik (1836)     5 sets, 2 TBs  narrow loss to Strong = neutral+
  07-02 Wimbledon R32   W  J. Choinski (1589)   4 sets         GRIND
  07-01 Wimbledon R64   W  T. Atmane (1480)     4 sets, TB     GRIND
  06-21 Halle FINAL     W  T. Fritz (2011)      6-4 6-4        DOM over ELITE — TITLE
  (also Halle SF 6-1 6-3 DOM, QF over FAA in 3)
CILIC (B) — 37, rank 47, hard:
  08-04 Montreal R64    W  S. Shimabukuro (1665) 7-6 6-1       GRIND (TB set), 90 min
  06-29 Wimbledon R64   L  D. Medvedev (1799)    1-6 2-6 4-6   blown out
  06-16 London R32      L  U. Humbert (1642)     straight sets
  06-13 Hertogenbosch QF L D. Medvedev (1833)    3 sets
  06-12 Hertogenbosch R16 W N. Borges (1578)     3 sets
  (window also: L Kouame 1418, L Landaluce 1579)

═══ STEP 1b — SHAPE / PATH / SHARPNESS ═══
  Tiafoe: STABLE. Halle title with a 6-4 6-4 DOM over Fritz (2011) is the best single data
    point on either side; since then two losses but both to live opposition, the Bublik
    5-setter with two tiebreaks reads neutral-to-positive. The Atmane loss (a man he beat at
    Wimbledon) is the one blemish — not enough to call FADING.
  Cilic: FADING — 1-8 across the fetched window with straight-set losses to a 1642 and a
    1579 and a 1-6 2-6 4-6 hiding from Medvedev. The Montreal R1 win needed a tiebreak
    against a 1665: winning ugly against a modest opponent = under-performing vs bracket.
  Path: Cilic is match-sharp at the venue (1 match played, won, conditions grooved).
    Tiafoe enters on a bye — first match here, 7 days since Washington: mild rust risk,
    not the >21d variety. Net: the path/sharpness read favours CILIC slightly, which is
    an argument against backing A at a 65% commission-adjusted price, not for backing B.
  Surface: no transition either side (both last played hard).

═══ STEP 2 — FATIGUE ═══
  Tiafoe: 1 match/14d, 140 min, last 07-28. Fresh-to-rusty, not tired.
  Cilic: 1 match/14d, 90 min, last match dated 08-04 — back-to-back day, but a TWO-set win.
  Calibration: the measured 3-set-hangover haircut (-4.5pp) applies only to a 3-set win the
  day before; a 2-set win is ~neutral and is NOT haircut. Age 37 (>=30, slow recovery)
  hardens it one notch to "slight negative" for Cilic — qualitative only, no number applied.
  Tiafoe's 7 days off partially offsets it in the other direction (rust). Net: roughly equal,
  no fatigue edge either way, and no fatigue input to any margin thesis.

═══ STEP 3 — NARRATIVE ═══
  TITLE: Halle, 44d ago (grass) for Tiafoe — well outside any let-down window (5-7d max).
  Let-down: None for either player.
  Venue momentum: Cilic 1-0 at Montreal this week; Tiafoe no Montreal history in window.
  player_notes: null. No injury/withdrawal context available in the brief.

═══ MARKET CHOICE ═══
  ML A @1.56 (Betfair): eff 1.532 -> need 65.3%. Sim 62.2%, market 61.4% -> edge -3.1%. NO.
  ML B @2.58 (Betfair): eff 2.501 -> 40.0% implied. Sim gives B 37.8% (edge -2.2%);
     ELO gives B 42.4% (edge +2.45%) — BELOW the 3% probe floor even on the most
     generous model, and that model is the ELO I just discounted. NO.
  AH: B +2.5 @ 1.85 — sim 46.5% vs 54.1% implied -> edge -7.5%. NO.
      A -2.5 @ 2.04 — sim 53.5% vs 49.0% implied -> nominal +4.4%, but laying the
      favourite at -2.5 is the HIGH-VARIANCE tier (bar 10%, min 1.72 -> fails outright).
      Sharp cross-check kills it independently: Pinnacle's A -2.0 prices the cover at
      51.1% no-vig, so the sharp implied cover at the WIDER -2.5 is materially below our
      53.5%. Contradicted by the sharp line = stale quote, not value.
  TOTALS: Betfair 23.0, Under @2.00 -> eff 1.95 = 51.3% implied. Sim interpolates
      Under 23.0 at ~56.7% (22.5: 53.8%, 23.5: 59.5%) -> nominal +5.4%, min 1.85 (met).
      KILLED by the sharp cross-check: Pinnacle prices the IDENTICAL 23.0 line at
      Under 49.1% no-vig (1.893/1.962) — a 7.6pp contradiction on the same number, with
      the sharps if anything leaning OVER. Reinforcing doubt: the sim reaches 21.9
      expected games only after a -3.0 game bias_correction off a 24.9 raw, i.e. its
      totals output is heuristically shaded exactly in the direction of this bet.
  One coherent distribution: sim win prob 62.2% -> no market on this match clears its bar.
  refutation_check: n/a (PASS — no web access on passes)

GATES
  G1 edge-origin: [FAIL] — best pre-narrative numeric edge is 2.45% (ELO on B vs Betfair 2.58), under the 3% probe floor and 5% full bar; sim/Pinnacle/Betfair all agree the price is right.
  G2 consensus:   [tier: capped] — only ELO dissents and it looks stale for a 37yo losing to sub-1600s; sim and Pinnacle sit on the market. Moot given G1.
  G3 fatal-risk:  [none | capped 1.0u risk_flag] — no unencodable risk; the margin markets die on the Pinnacle sharp cross-check, not on a story.

═══ VERDICT: PASS ═══
An efficiently priced match. The sharpest number available (sim 62.2%) sits within 1pp of
both Betfair and Pinnacle, so there is nothing to sell. The tempting angles both fail on
their own terms: the Cilic ML "value" rests solely on an ELO that flatters a 37-year-old
who has lost to 1418- and 1579-rated opponents inside the window, and the Under 23.0 —
the only market with a nominal 5%+ edge — is contradicted on the identical line by
Pinnacle, whose price is the reference precisely for this check. Tiafoe is the likelier
winner (bye, Halle title, better bracket-adjusted results) but at 1.56 he is priced fully;
Cilic's venue sharpness vs Tiafoe's week off is the reason not to pay up, not a reason to
back the 37-year-old. Predicted winner Tiafoe, no stake.