L. Darderi vs G. Diallo — Prediction & Analysis

TennisEdge's AI model finds no betting value in L. Darderi vs G. Diallo (Montreal on 2026-08-05) — the market is efficient, with L. Darderi favoured at 1.90.

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 25, 2026.

Recent form

All Aug. 5, 2026 results & picks →

Full analysis

L. Darderi (A, 24, ATP 25) vs G. Diallo (B, 24, ATP 40) — ATP Montreal R2 (1/32), HARD

═══ MODELS vs MARKET ═══
ELO      A 31.4%  |  ML  A 39.9%  |  blended model A 38.0%  |  sim baseline A 26.8%
Betfair ML  A 2.12 / B 1.86  → devig A 46.7% / B 53.3%   (eff A 2.064 BE 48.45%; eff B 1.817 BE 55.04%)
Pinnacle    A 2.03 / B 1.848 → no-vig A 47.65% / B 52.35%; drift_a 0.0pp (firm), fetched 14:55Z (fresh)
Book quote  A 1.90 / B 1.92 (context only — exchange is the baseline)

All three models are on Diallo (60–73%); BOTH sharp venues sit at a near-coinflip
(B 52–53%). The estimates line up monotonically — the more market-aware the number,
the higher Darderi (sim 26.8 < ELO 31.4 < ML 39.9 < Pinnacle 47.65). The sim is the
outlier, and its inputs explain why.

ELO/ranking sanity: Darderi ranked 25 but is the ELO dog by a distance vs a No.40.
This is NOT stale ELO — it is a genuine surface split (see form): his rank is built
on clay. So the direction of the model lean is legitimate; the MAGNITUDE is not.

═══ SIM INPUT AUDIT & CALIBRATION (Step 4b) ═══
Baseline: A spw .556 / rpw .377 (43 matches) — that 43-match window is almost entirely
CLAY (his whole 90d bar one hard match). Clay depresses SPW for the same player, so
.556 understates Darderi's hard-court serve. B spw .581 (50 matches) is built on
hard/grass. Result: sim reads A at 26.8%, ~21pp below the sharp price — inputs biased.
Shaded ONE defensible input: spw_a .556 → .591 (+0.035, clamp ±0.05, clamped=false)
→ A 44.15% / B 55.85%, expected total 22.5 games, expected margin −0.5.
That lands between ML (39.9) and Pinnacle (47.65) — the band I trust. All market edges
below are read off THIS single coherent distribution.

═══ RECENT FORM (5 most recent) ═══
DARDERI (A) — clay machine, hard-court unknown
  07-30 Los Cabos HARD  L  Svrcina (1738)   3-0  ← single partial set, looks retired/abandoned
  07-25 Estoril  clay   L  Blockx (1864)    5-7 7-6 7-6  (two TBs, elite-bracket, top shape)
  07-24 Estoril  clay   W  Faria (1806)     6-2 5-7 2-6  3-setter
  07-23 Estoril  clay   W  P. Martinez(1644) 6-0 6-1  DOM
  07-19 Bastad   clay   L  Rublev (1935)    6-4 6-3  FINAL, competitive
  Window: Bastad FINAL, Estoril SF, Rome SF — all on CLAY. Off clay: grass 1-4, hard 0-1.
  Wins are overwhelmingly GRIND (TBs vs Borges, Vallejo, Zverev, Jodar, Ofner, Burruchaga).
DIALLO (B) — hard/grass server, home crowd
  08-03 Montreal HARD   W  Jacquet (1769)   6-7 6-3 6-2  GRIND, ~2h, YESTERDAY
  07-28 Los Cabos HARD  L  T. Boyer (1622)  5-7 2-6  UNDER-performing vs weak bracket
  07-02 Wimbledon grass L  Sonego (1643)    5 sets
  06-30 Wimbledon grass W  Bonzi (1722)     GRIND
  06-25 Eastbourne grass L  Draper (1848)   6-1 6-4
  Window 4-6. Wins are all GRIND (four straight tiebreak-laden wins).

SHAPE (scoreline vs expectation):
  Darderi — STABLE/strong on clay (7-6 7-6 loss to a 1864 = over-performing), but
  UNPROVEN-to-poor on hard: his only hard match ended 3-0 in one partial set.
  Diallo — FADING. Yesterday's win was a GRIND over a 1769 after dropping a TB, and
  before it a 5-7 2-6 loss to a 1622 — under-performing vs bracket twice running.
  A FADING side is not the side to back into a 62-73% model number. This is the core
  reason I side with the sharp coinflip over the models.

PATH: Diallo 1 match at Montreal (played-in, home crowd) — positive. Darderi is
first-match-at-venue with only one hard court match since a long clay block →
transition + rust risk, and that widens his margin variance in match one.

═══ FATIGUE (brief-supplied) ═══
A Darderi (24): 4 matches/14d, 12 sets, 459 min, last match 07-30 → 5 days rest.
  High cumulative load but well rested; per 07-27 calibration cumulative minutes alone
  is noise. No extreme spot.
B Diallo (24): 2 matches/14d, 6 sets, 235 min, last match 08-03 = YESTERDAY, a WON
  3-SET match → the measured 3-set-hangover applies: −4.5pp, surface-agnostic, applied
  ONCE, not stacked with any minutes penalty. Age 24 = baseline, no modulation.
  Encoded as a number, not a veto (G3 2026-08-04 rule).
  → B 55.85% pre-narrative − 4.5 = 51.35% net; edge vs BE 55.04% = −3.7%.
  Mirror on A = 48.65% vs BE 48.45% = +0.2% — i.e. an edge that exists ONLY because of
  the hangover. Pre-narrative A is −4.3%. Backing A here would be a textbook G1 failure.

═══ NARRATIVE ═══
Let-down: none in window (first Final = Bastad 07-19, LOST, 16d ago → outside all bands).
Title recency: no title in window. Venue momentum: Darderi 0 matches at Montreal;
Diallo 1-0. Surface transition: Darderi clay→hard with a single, aborted hard match
(TRANSITION risk, widens his margin variance); Diallo none (hard→hard).
player_notes: null — no press facts either way (absence means nothing).

═══ MARKET CHOICE ═══
All probabilities from the ONE calibrated distribution (spw_a .591):
  B ML   1.86  p .5585  BE .5504  edge +0.81%   (min raw 1.93)
  A ML   2.12  p .4415  BE .4845  edge −4.30%
  B −0.5 1.91  p .5352  BE .5363  edge −0.11%
  A +0.5 2.00  p .4648  BE .5128  edge −4.80%
  U 23.5 1.91  p .5652  BE .5363  edge +2.89%   (needs raw 1.99 for the 5pp bar)
  O 23.5 2.04  p .4348  BE .5030  edge −6.82%
AH: A +0.5 @ 2.00 / B -0.5 @ 1.91 — both passed, no edge (not a dog-ML stake, shadow line logged).
Best market = Under 23.5 at +2.9pp — short of the 5pp bar, and margin markets get NO
probe tier. On the RAW (uncalibrated) sim the under showed +9.8% and B ML +7.0%; both
of those edges are pure artefacts of the sim's contaminated 73%-for-Diallo read.
Sharp cross-check (margin): Pinnacle total 23.0 with the OVER favoured (devig O 51.8%)
→ sharps expect ~23.2 games vs my 22.5, so sharp-implied under-23.5 ≈ 54% against my
56.5% — the sharp line materially contradicts the under, which kills it outright.
Pinnacle AH +0.5 devigs to A 49.1% / B 50.9%, likewise contradicting any B −0.5 lean.
Independent confirmation of the sharp view: BOTH players' wins in this window are
tiebreak-heavy grinds — the raw sim's 21.3-game expectation is inconsistent with the
actual scoreline distributions of both men.
refutation_check: not run — PASS decision, no web access permitted.

GATES
  G1 edge-origin: FAIL — best pre-narrative numeric edge is Under 23.5 at +2.9pp (5pp bar; no probe tier on margin markets); B ML +0.8pp. The 7-10% edges exist only off contaminated clay-derived SPW inputs.
  G2 consensus:   [tier: capped] — 3 models on B but all share the same off-surface stat contamination, and live this-event shape (Diallo's grind win, Boyer loss) runs against them; moot, no bet.
  G3 fatal-risk:  [none] — Diallo's 3-set-win-yesterday encoded numerically (−4.5pp), not vetoed; sharp Pinnacle total 23.0 contradicts the raw sim's 21.3 and kills the under.

═══ VERDICT: PASS ═══
The apparent value here is manufactured by one broken input. Fix the clay-derived SPW
and the whole slate of edges collapses to 0-3pp, with both sharp venues firm at a
coinflip and no drift. The two live signals that are genuinely mine — Diallo FADING
(grind over a 1769, loss to a 1622) and his measured 3-set hangover — both point AWAY
from the side the models like, while the only edge on Darderi (+0.2%) is 100%
narrative-derived and pre-narrative deeply negative. Nothing to stake on either side,
in any market. Efficient price; pass.