TennisEdge's AI model finds no betting value in A. Zakharova vs R. Zarazua (Memphis (Usa) - Qualification on 2026-07-31) — the market is efficient, with A. Zakharova favoured at 1.60.
Independent estimates of who wins, each from a different method. The blend is what the bot prices against the market.
| Method | A. Zakharova | R. Zarazua |
|---|---|---|
| Elo ratings-based, surface-weighted |
62.0% | 38.0% |
| Machine learning 29-feature gradient-boosted model |
67.9% | 32.1% |
| Serve / return point-by-point simulation |
46.6% | 53.4% |
| Blend what the bot actually prices off |
58.4% | 41.6% |
| Market implied de-vigged from the opening odds |
57.9% | 42.1% |
We publish how well these score against the closing market, including where they lose: how accurate are tennis predictions?
| Player | Surface | Elo | Record | Matches |
|---|---|---|---|---|
| A. Zakharova | hard | 1771 | 168–89 | 257 |
| R. Zarazua | hard | 1720 | 110–84 | 194 |
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 31, 2026.
R. Zarazua leads 1–0 over 1 meeting. Full record.
| Date | Event | Surface | Winner | Score |
|---|---|---|---|---|
| July 31, 2026 | Memphis (Usa) - Qualification | hard | R. Zarazua | 2 - 0 |
| L | R. Zarazua | 0 - 2 | July 31, 2026 |
| W | C. Osorio | 0 - 2 | July 29, 2026 |
| W | L. Fruhvirtova | 2 - 0 | July 27, 2026 |
| L | H. Vandewinkel | 0 - 2 | July 18, 2026 |
| W | T. Prozorova | 2 - 0 | July 17, 2026 |
| W | T. Korpatsch | 2 - 1 | Aug. 3, 2026 |
| L | D. Vidmanova | 2 - 0 | Aug. 1, 2026 |
| W | A. Zakharova | 0 - 2 | July 31, 2026 |
| W | T. Maria | 0 - 2 | July 29, 2026 |
| W | M. Kessler | 2 - 0 | July 29, 2026 |
All July 31, 2026 results & picks →
A. Zakharova (A, #81, age 24) vs R. Zarazua (B, #75, age 28) — WTA Memphis Qualification QF, hard, 2026-07-31
MODELS (prob A): ELO 62.0% | ML 67.9% | sim/composite 58.4% | Betfair implied ~60.6% (odds_a 1.65).
Commission-adj effective A price 1.617 -> breakeven 61.8%.
The models SPLIT rather than agree: the point-by-point sim (58.4%) sits BELOW the market, ELO (62%)
sits essentially AT the market, and only ML (67.9%) is meaningfully above. When sim + ELO + exchange
cluster ~58-62% and the lone dissenter is ML (known break-even/overconfidence failure mode), the read
is efficient market, not edge.
RECENT FORM (last 5):
Zakharova: 07-29 beat Osorio (1500) 6-4 6-3 DOM; 07-27 beat Fruhvirtova (1500) 7-6 6-1 GRIND;
07-18 LOST Vandewinkel (1749) 3-6 3-6; 07-17 beat Prozorova (1500) 6-3 6-3 DOM; 07-16 beat
Shubladze (1500) 7-6 7-6 GRIND. Deeper: Eastbourne title (grass, 06-21). Shape STABLE, wins over
weak quali fields; legit #81 pedigree.
Zarazua: 07-29 beat Maria (1500) 6-4 6-4 AND beat Kessler (1500) 6-4 6-1 (SAME-DAY DOUBLE); 07-25
LOST Dolehide 5-7 4-6; 07-24 beat Naito 6-4 4-1; 07-23 beat Stephens 2-6 6-3 6-2. Level-dropper
(W100 Evansville ITF) with poor tour-level shape (grass/clay littered with bad losses to sub-1500).
Shape FADING at tour level.
PATH/SHARPNESS: Both match-sharp at this venue. Zarazua's opponent ELOs are all 1500 (unknown/weak
ITF fields) -> her rating context is thin, but that would bias the SIM not ML.
FATIGUE: Zarazua played a same-day DOUBLE on 07-29 (extreme spot), back_to_back 3, most court time
(451min). Zakharova won only 2 sets yesterday (~neutral). Real differential favouring A, BUT fatigue
is confirmation only (cannot create the edge) and market already drifted B book 2.2 -> Betfair 2.34.
ELO/RANK: B ranked #75 (above A's #81) yet ELO-dog -> ELO credible given B's poor recent tour results.
ALT MARKETS:
- Totals 21.5: sim over 51.7%/under 48.3%. Over edge 0.517 - 1/1.770 = negative; Under edge
0.483 - 1/1.798 = negative. No edge.
- Handicap A -2.5 (1.78): high-variance tier (laying fav -2.5 needs blowout), model handicap null,
no margin thesis. Pass.
- AH shadow: not staking a dog ML >=3.0, n/a.
PRICING: For the credible ~62% estimate, min_odds = 1.05/0.62 = 1.694 > Betfair 1.65 -> automatic
PASS on price. Only ML's 67.9% would clear min (1.546), but ML is the lone dissenter contradicted by
the sim and ELO.
GATES
G1 edge-origin: FAIL — no 5% exchange edge pre-narrative; ELO ~0.2%, sim NEGATIVE, only lone ML
6% and it is contradicted by the point-by-point sim. Executable price below min_odds for the
credible estimate.
G2 consensus: stake-tier only — would be lone_signal (ML alone), capped; moot given G1 fail.
G3 fatal-risk: n/a (ML market; no margin bet).
VERDICT: PASS. Sim + ELO + exchange cluster ~58-62%; only ML disagrees and price sits below min for
the credible read. Fatigue on B is real but is narrative confirmation the market has already partly
priced — it cannot manufacture an edge that the sim and ELO do not support.