TennisEdge's AI model finds no betting value in M. Zheng vs K. Nishikori (Montreal on 2026-08-01) — the market is efficient, with M. Zheng favoured at 1.57.
Independent estimates of who wins, each from a different method. The blend is what the bot prices against the market.
| Method | M. Zheng | K. Nishikori |
|---|---|---|
| Elo ratings-based, surface-weighted |
55.6% | 44.4% |
| Machine learning 29-feature gradient-boosted model |
61.6% | 38.4% |
| Serve / return point-by-point simulation |
65.5% | 34.5% |
| Blend what the bot actually prices off |
55.6% | 44.4% |
| Market implied de-vigged from the opening odds |
58.9% | 41.1% |
We publish how well these score against the closing market, including where they lose: how accurate are tennis predictions?
| Player | Surface | Elo | Record | Matches |
|---|---|---|---|---|
| M. Zheng | hard | 1772 | 67–35 | 102 |
| K. Nishikori | hard | 1721 | 55–28 | 83 |
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. 4, 2026.
Level at 1–1 over 2 meetings. Full record.
| Date | Event | Surface | Winner | Score |
|---|---|---|---|---|
| Aug. 1, 2026 | Montreal | hard | M. Zheng | 2 - 0 |
| June 18, 2023 | Palmas Del Mar | hard | K. Nishikori | 2 - 0 |
| L | M. Kecmanovic | 1 - 2 | Aug. 4, 2026 |
| W | K. Nishikori | 2 - 0 | Aug. 1, 2026 |
| L | A. Gea | 2 - 1 | July 30, 2026 |
| W | M. Landaluce | 2 - 0 | July 29, 2026 |
| L | J. Fearnley | 2 - 0 | July 25, 2026 |
| L | M. Zheng | 2 - 0 | Aug. 1, 2026 |
| L | R. Jodar | 2 - 0 | July 30, 2026 |
| W | J. Shang | 2 - 1 | July 27, 2026 |
| L | J. Kennedy | 2 - 1 | April 22, 2026 |
| W | C. Smith | 2 - 0 | April 21, 2026 |
All Aug. 1, 2026 results & picks →
M. Zheng (A, 22, #182) vs K. Nishikori (B, 36, #238) — Montreal Final, hard.
MODELS vs MARKET
ELO A 55.6% | ML A 61.6% | model A 55.6% | market implied A 58.9% (odds 1.57 / 2.25).
B side: ELO 44.4%, ML 38.4% (SPLIT), market implied 41.1%.
B edge: ELO +3.3%, ML -2.7% — models disagree on direction. A side no edge (models ~= market).
RECENT FORM (hard, most-recent 5)
Zheng: L Gea(1807) 0-6 6-4 6-0; W Landaluce(1776) 6-4 6-3 (solid DOM-ish over Strong);
L Fearnley(1677) 6-3 6-4; W Walton(1767) 4-6 7-6 6-3 GRIND; W Magadan(1438) 7-5 6-3.
Shape STABLE/mixed — beats Strong but also drops straight sets to Strong. Match-sharp (played 07-30).
Nishikori: L Jodar(1873, Elite) 6-3 6-2; W Shang(1723) 6-7 6-3 6-4 GRIND over Strong.
Thin 90d sample (comeback, #238 but competitive vs Strong). Match-sharp (played 07-30).
FATIGUE (context only)
Both last played 07-30 (~1d rest). Zheng 6/14d, 603min, b2b 2. Nishikori 2/14d, 230min.
3-set-hangover haircut applies to yesterday's WINNERS only — both LOST 07-30, so no haircut.
Cumulative-minutes gap is context, not an edge (Step 2 calibration). Zheng 22 recovers fast.
ELO/RANK sanity: both ranks modest (182/238) yet mid-table ELO; comeback vet Nishikori's ELO may
understate — do not back opponent on ELO gap. ELO here is credible-enough but no lean either way.
DATA NOTE: neither player's history contains any Montreal match — histories are Los Cabos /
Washington / Wimbledon. Venue path unverifiable from brief; treat metrics with extra caution.
GATES
G1 edge-origin: FAIL — best numeric edge is B +3.3% (ELO only), below 5%/8% bar; ML contradicts (-2.7%).
G2 consensus: stake-tier only — would be lone_signal (models split), moot given G1 fail.
G3 fatal-risk: n/a (ml market).
PRICING: alt_markets null -> no Betfair exchange price. ML bets require edge >= 8% vs book quote
(book-priced) or prefer PASS. Actual edge 3.3% and split -> auto PASS.
VERDICT: PASS. No exchange liquidity, sub-bar split edge, no qualitative signal strong enough to
override an efficient market. Predicted winner: M. Zheng (favourite).