TennisEdge's AI model finds no betting value in T. Gibson vs E. Cocciaretto (Toronto on 2026-08-03) — the market is efficient, with E. Cocciaretto favoured at 1.65.
Independent estimates of who wins, each from a different method. The blend is what the bot prices against the market.
| Method | T. Gibson | E. Cocciaretto |
|---|---|---|
| Elo ratings-based, surface-weighted |
50.0% | 50.0% |
| Machine learning 29-feature gradient-boosted model |
39.0% | 61.0% |
| Serve / return point-by-point simulation |
55.1% | 44.9% |
| Blend what the bot actually prices off |
50.0% | 50.0% |
| Market implied de-vigged from the opening odds |
42.3% | 57.7% |
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. Gibson | hard | 1825 | 179–94 | 273 |
| E. Cocciaretto | clay | 1690 | 94–47 | 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. 3, 2026.
T. Gibson leads 1–0 over 1 meeting. Full record.
| Date | Event | Surface | Winner | Score |
|---|---|---|---|---|
| Aug. 3, 2026 | Toronto | hard | T. Gibson | 1 - 0 |
| W | E. Cocciaretto | 1 - 0 | Aug. 3, 2026 |
| L | T. Maria | 0 - 2 | July 27, 2026 |
| L | M. Bouzkova | 1 - 2 | June 30, 2026 |
| L | M. Keys | 0 - 2 | June 23, 2026 |
| L | Ka. Pliskova | 2 - 0 | June 19, 2026 |
| L | T. Gibson | 1 - 0 | Aug. 3, 2026 |
| L | N. Osaka | 1 - 2 | July 31, 2026 |
| W | E. Navarro | 1 - 2 | July 29, 2026 |
| W | C. Tauson | 2 - 1 | July 27, 2026 |
| L | Xin. Wang | 2 - 1 | June 29, 2026 |
All Aug. 3, 2026 results & picks →
T. Gibson (A, 22, #62) vs E. Cocciaretto (B, 25, #41) - Toronto WTA-1000, hard, 1/64.
MODELS
ELO: A 50.0% / B 50.0%
ML: A 39.0% / B 61.0%
Sim: A 55.1% / B 44.9% (win_prob resim)
Market: A 43.5% / B 59.9% (Betfair 2.30 / 1.67, commission-adj B 0.611)
The sim favours GIBSON (55%), contradicting ML, market AND the form read - the classic
opponent-unadjusted-input trap. Gibson sim inputs spw .599/rpw .424 are farmed off a May
Strasbourg CLAY title run vs weak fields (Zhang 1485, Lys 1646) plus long layoffs; they do
not reflect her current hard-court level. Sim diverges from ML/market/qualitative -> inputs
suspect -> discount the sim. ELO 50/50 is likewise stale/inflated for Gibson (rank 62 vs 41,
ELO overrates her). No re-run staked: a correction only confirms B is correctly priced.
RECENT FORM
Gibson (FADING, rusty): last match 2026-07-27 Memphis QUALIFYING, hard, LOST 0-6 3-6 to
T. Maria (1500) - a blowout loss to an Avg player in qualies. Before that a ~1-month gap
back to Wimbledon (Lost 1-6 6-3 2-6 Bouzkova). Grass season mostly losses (Keys, Ka.
Pliskova). Only bright spot is May clay (Strasbourg title) - off-surface, stale. Currently
out of form and match-rusty on hard.
Cocciaretto (SHARPENING, match-sharp): 3 hard-court matches this week at Washington -
Won vs Tauson (1795) 4-6 7-6 7-6, Won vs Navarro (1725) 3-6 6-4 3-6, competitive Loss to
Osaka (1929) 6-4 4-6 3-6. Quality wins over Strong opponents on THIS surface; grinder who
wins tight (two TBs vs Tauson).
FATIGUE
Cocciaretto 3 matches/9 sets/352min this week, last match 2 days ago (07-31, a LOSS - no
3-set-hangover haircut, which applies only to a WON 3-setter the day before). Heavy load is
CONTEXT only (calibration: cumulative minutes are not an edge; market over-prices visible
fatigue). Age 25 baseline. No extreme spot -> roughly neutral. Gibson well-rested (1 match
in 14d) but that "rest" is rust, not sharpness.
Note: the fatigue narrative would drift B's price up; per calibration the value on such a
drift is usually ON the tired side (B) - but B is not underpriced (see gates), so no bet.
NARRATIVE
No let-down (no recent Final for either in window bar Gibson's May clay title, 2+ months,
irrelevant). Surface: both now on hard; Gibson's recent baseline is grass/clay -> mild
transition/rust risk against her. Cocciaretto already grooved on hard (Washington).
GATES
G1 edge-origin: FAIL - the only numeric "edge" (sim/ELO on A, or value_pick A +7.7%) comes
from farmed clay stats & stale ELO, not credible model-market disagreement. B's ML edge vs
Betfair is ~1% (0.62 vs implied 0.611) - below the 5% bar. No numeric edge stands.
G2 consensus: n/a (no candidate clears G1).
G3 fatal-risk: n/a.
MARKET CHOICE
ML: B fully priced (edge ~1%), A "value" is an artifact -> PASS.
Handicap: A +2.5 @1.76 sits on the out-of-form side (self-negating cover) and rests on the
farmed sim; B -2.5 @1.89 needs a blowout but Cocciaretto grinds tight (TBs, 3-setters) ->
both PASS.
Total: line 21.5, sim over 0.506/under 0.494 vs market over 0.538/under 0.500 - no edge ->
PASS.
AH: A +2.5 @ 1.76 (shadow - passed, out-of-form side, sim-biased).
VERDICT: PASS. B (Cocciaretto) is the correct side on form/surface/sharpness but the market
already prices her at ~60% (ML agrees). The apparent A value is a farmed-ELO/farmed-sim
mirage. No market clears the edge bar.