TennisEdge's AI model finds no betting value in A. Walton vs J. Brooksby (Montreal on 2026-08-03) — the market is efficient, with J. Brooksby favoured at 1.80.
Independent estimates of who wins, each from a different method. The blend is what the bot prices against the market.
| Method | A. Walton | J. Brooksby |
|---|---|---|
| Elo ratings-based, surface-weighted |
52.7% | 47.3% |
| Machine learning 29-feature gradient-boosted model |
42.1% | 57.9% |
| Serve / return point-by-point simulation |
60.1% | 39.9% |
| Blend what the bot actually prices off |
52.7% | 47.3% |
| Market implied de-vigged from the opening odds |
47.4% | 52.6% |
We publish how well these score against the closing market, including where they lose: how accurate are tennis predictions?
| Player | Surface | Elo | Record | Matches |
|---|---|---|---|---|
| A. Walton | hard | 1748 | 206–117 | 323 |
| J. Brooksby | hard | 1729 | 86–55 | 141 |
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. 1, 2026.
| W | J. Boulais | 2 - 0 | Aug. 1, 2026 |
| L | D. Svrcina | 2 - 1 | July 29, 2026 |
| L | M. Zheng | 2 - 1 | July 24, 2026 |
| W | T. Kokkinakis | 1 - 2 | July 22, 2026 |
| W | B. Bicknell | 1 - 1 | July 20, 2026 |
All Aug. 3, 2026 results & picks →
═══ A. Walton (A) vs J. Brooksby (B) — ATP Montreal R1 (hard) ═══
MODELS vs MARKET
ELO: Walton 52.7% (favors A)
ML: Walton 42.1% (favors B, 57.9%)
Market: Betfair 2.26/1.75 -> Walton 44.2% raw; Pinnacle no-vig 44.98%.
Models SPLIT. The lone model beating the market is ELO (on Walton); ML and
both market sources agree AGAINST Walton.
ELO/RANKING SANITY (decisive)
Walton ranked 103 yet ELO-favored = "high ELO, modest rank" = rating farmed
at weak events. His finals/wins in the window are vs weak fields: Boulais
1512, Ghibaudo 1454, Bicknell 1544, Ilagan/Legout ~1500s. Discount the ELO
edge; lean PASS unless live form independently backs it. It does not — see
below.
RECENT FORM (last 5 each)
Walton (27): 08-01 WON Montreal Final v Boulais(1512) 7-5 7.7-6.4 GRIND/Avg;
07-29 LOST v Svrcina(1720) 3s; 07-24 LOST v Zheng(1785) 3s; 07-22 WON v
Kokkinakis(1833) 3s (real scalp); 07-20 WON v Bicknell(1544). Match-sharp
on hard, but quality wins thin vs a lot of weak-field volume.
Brooksby (25): 07-31 LOST QF v Wong(1721) 2s; 07-30 WON v Winter(1512);
07-29 WON v Moutet(1764) 6-4 6-4 DOM/Strong; 07-03 LOST v Sinner(2020) 2s;
07-01 WON v Buse(1585) DOM. Legit ATP-level hard form; Los Cabos QF run.
SHAPE/PATH
Both match-sharp on hard (played-in). Brooksby's straight-sets win over Moutet
(1764) is the strongest recent bracket win by either; Walton's best is
Kokkinakis in 3. Edge on quality-of-competition read: Brooksby.
FATIGUE (Step 2)
Walton: 5 matches/678min/14d, played 08-01 — but a 2-SET win yesterday =
~neutral hangover (no -4.5pp extra-set toll; that applies only to 3-set wins).
Age 27 baseline. Brooksby: 260min/3 matches, last 07-31. Cumulative gap is
context only, not an edge (calibration: minutes alone are noise).
NARRATIVE (Step 3)
LETDOWN: Walton won a Final 08-01 (0-1d ago) -> let-down spot, cuts Walton.
This confirms fading the ELO number; it does not create an opposing edge
large enough to back Brooksby (market already prices him ~57%).
MARKET CHOICE
alt_markets: handicap/totals null; only ML priced.
AH: not quoted.
Backing Walton @ Betfair 2.26: eff odds 2.197, 1/eff = 0.4552. Trusted-prob
read (ML+market, ~44-45%) shows NO edge; the 7% "edge" exists ONLY off the
discounted/farmed ELO. Backing Brooksby @ 1.75: eff 1.7125, 1/eff = 0.584;
min_odds 1.05/0.58 = 1.81 > 1.75 -> market efficient, no edge either.
GATES
G1 edge-origin: FAIL — the only >5% edge (Walton) comes from a farmed-ELO
number the ranking sanity + form read explicitly discount; reliable models
(ML+market) show no edge on either side.
G2 consensus: n/a (stake-tier only)
G3 fatal-risk: n/a (no margin bet)
VERDICT: PASS. Models split with the reliable ones (ML + Betfair + Pinnacle)
siding with the market on Brooksby; the lone dissenting model (ELO on Walton)
is a suspected farmed rating that the live-form and let-down reads undercut
rather than support. No trustworthy edge on either side. Efficient -> pass.