TennisEdge's AI model finds no betting value in T. Prozorova vs P. Stearns (Memphis (Usa) - Qualification on 2026-07-30) — the market is efficient, with P. Stearns favoured at 1.44.
Independent estimates of who wins, each from a different method. The blend is what the bot prices against the market.
| Method | T. Prozorova | P. Stearns |
|---|---|---|
| Elo ratings-based, surface-weighted |
46.4% | 53.6% |
| Machine learning 29-feature gradient-boosted model |
39.0% | 61.0% |
| Serve / return point-by-point simulation |
56.3% | 43.7% |
| Blend what the bot actually prices off |
43.4% | 56.6% |
| Market implied de-vigged from the opening odds |
34.4% | 65.6% |
We publish how well these score against the closing market, including where they lose: how accurate are tennis predictions?
| Player | Surface | Elo | Record | Matches |
|---|---|---|---|---|
| T. Prozorova | hard | 1728 | 119–58 | 177 |
| P. Stearns | hard | 1745 | 97–74 | 171 |
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 30, 2026.
P. Stearns leads 1–0 over 1 meeting. Full record.
| Date | Event | Surface | Winner | Score |
|---|---|---|---|---|
| July 30, 2026 | Memphis (Usa) - Qualification | hard | P. Stearns | 2 - 0 |
| L | P. Stearns | 0 - 2 | July 30, 2026 |
| W | S. Sierra | 1 - 2 | July 27, 2026 |
| W | K. Drenser-Hagmann | 2 - 1 | July 27, 2026 |
| L | A. Zakharova | 2 - 0 | July 17, 2026 |
| W | D. Khomutsianskaya | 0 - 2 | July 16, 2026 |
| L | G. Ruse | 2 - 0 | Aug. 3, 2026 |
| L | E. Kalieva | 2 - 1 | July 31, 2026 |
| W | T. Prozorova | 0 - 2 | July 30, 2026 |
| W | M. Stoiana | 1 - 2 | July 28, 2026 |
| L | N. Bartunkova | 1 - 2 | June 29, 2026 |
All July 30, 2026 results & picks →
═══ T. Prozorova (A) vs P. Stearns (B) — WTA Memphis Qualification, hard, 1/8-finals ═══
MODELS vs MARKET (player A)
ELO 46.4%
ML 39.0%
model 43.4%
sim (resim) 56.3% <-- lone outlier
Betfair ML A 2.54 -> 39.4% implied | B 1.64 -> 61.0%
Book A 2.75 / B 1.44
Consensus (ML + market + ELO-ish) = Stearns ~55-61% favourite. Only the
serve/return sim dissents (A 56%), and it dissents for a KNOWN, identifiable
reason (see sim validation) — so it is not independent support for A.
RECENT FORM (last 5)
Prozorova (A): won Memphis qualifying + main-draw R1 (07-27, two matches
that day). R1 vs Sierra a 3-set grind, qual final 4-6 6-3 6-1 comeback.
Opponents all ~1500 ELO (weak field). Shape: match-sharp but UNDER-
performing — grinding 3-setters past weak opposition. Earlier grass run
(Wimbledon SF label) but ELOs pinned ~1500 = noisy/data-thin ratings.
Stearns (B, #49, 24): Memphis R1 07-28 scrappy 3-set win vs Stoiana
(dropped a 0-6 set). Recent schedule BRUTAL: FO/Rome/Strasbourg vs Keys
1892, Bencic 1816, Kasatkina 1790, Sakkari 1709, Snigur 1718, Kenin.
Eastbourne final (lost). Higher-quality player on a mixed run.
PATH / SHARPNESS
A is 3+ matches deep at venue (match-sharp, positive). B direct main-draw
entry, 2nd match, 2 days rest — fresh-ish, minor rust only.
FATIGUE (Step 2 calibration)
A: last match 07-27 = 3 days ago; no day-before 3-set hangover. 340min/14d
is CONTEXT only, not an edge (cumulative-minutes = noise). Age null.
B: last match 07-28 = 2 days ago; won a 3-setter but NOT yesterday -> no
-4.5pp haircut applies. Neither side carries an extreme fatigue spot.
NARRATIVE
No first-Final let-down (A's "Final" wins are qualifying finals, not tour).
Surface: both on hard, no transition. No motivation signal in the data.
SIM VALIDATION (resimulate baseline)
Inputs: A spw .59/rpw .427 vs B spw .573/rpw .432. Sim -> A 56.3%.
DIVERGES from ELO/ML/market by a lot. Cause: sim inputs are opponent-
UNADJUSTED. Prozorova's serve/return numbers were built vs ~1500 weak
fields (inflated); Stearns' were built vs a genuinely elite schedule
(Keys/Bencic/Kasatkina/Sakkari, 1700-1890 ELO) which DEFLATES her raw
SPW/RPW. So the sim systematically OVER-rates A and UNDER-rates B. A
justified correction (shade B's SPW up) would only pull the sim back toward
the market — killing the apparent A edge, not creating one. Trust consensus,
not the biased sim.
MARKET CHOICE
ML A @ Betfair 2.54: eff 2.463, break-even 40.6%. Credible win prob for A
is at best ~43-46% (ELO ceiling, itself suspect); consensus lower. Edge
~2-3%, below 5% bar. No bet.
Handicap: Betfair A -2.5 @2.04 / B +2.5 @1.89. Sim (biased toward A) gives
B+2.5 covers 52.7% -> vs eff 1.846 break-even 54.2% = NEGATIVE edge; and
the bias means true B-cover is even higher, still no A-side value. No edge.
AH: B +2.5 @ 1.89
Totals: sim over21.5 .491 / under .509 vs Betfair over2.02/under1.92, both
below their break-evens. Efficient, no edge.
GATES
G1 edge-origin: FAIL — the only pre-narrative model-vs-market gap favouring
A comes from the sim, which is identifiably biased in A's favour
(opponent-unadjusted inputs); ML+market+ELO cluster on B. No clean 5%
exchange edge exists on any market. Candidacy dies here.
G2 consensus: n/a (stake-tier only; no qualifying bet)
G3 fatal-risk: n/a (no margin bet staked)
VERDICT: PASS. Every credible signal (ML, exchange, ELO-lean) makes Stearns a
~55-61% favourite; the lone dissenter is a sim biased toward Prozorova for a
concrete reason. No market clears the 5% exchange-edge bar. Efficient price.