TennisEdge's AI model finds no betting value in E. Micic vs G. Garcia-Perez (Warsaw on 2026-08-02) — the market is efficient, with E. Micic favoured at 1.37.
Independent estimates of who wins, each from a different method. The blend is what the bot prices against the market.
| Method | E. Micic | G. Garcia-Perez |
|---|---|---|
| Elo ratings-based, surface-weighted |
72.2% | 27.8% |
| Machine learning 29-feature gradient-boosted model |
85.1% | 14.9% |
| Serve / return point-by-point simulation |
61.1% | 38.9% |
| Blend what the bot actually prices off |
71.1% | 28.9% |
| Market implied de-vigged from the opening odds |
67.1% | 32.9% |
We publish how well these score against the closing market, including where they lose: how accurate are tennis predictions?
| Player | Surface | Elo | Record | Matches |
|---|---|---|---|---|
| E. Micic | hard | 1503 | 73–73 | 146 |
| G. Garcia-Perez | hard | 1327 | 34–44 | 78 |
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.
E. Micic leads 1–0 over 1 meeting. Full record.
| Date | Event | Surface | Winner | Score |
|---|---|---|---|---|
| Aug. 2, 2026 | Warsaw | hard | E. Micic | 2 - 0 |
| L | V. Podrez | 0 - 2 | Aug. 3, 2026 |
| W | G. Garcia-Perez | 2 - 0 | Aug. 2, 2026 |
| L | L. Fruhvirtova | 0 - 2 | July 18, 2026 |
| L | Q. Zheng | 0 - 2 | July 15, 2026 |
| W | M. Matoula | 0 - 2 | July 13, 2026 |
| L | E. Micic | 2 - 0 | Aug. 2, 2026 |
| L | E. Vedder | 2 - 1 | July 19, 2026 |
| L | E. Pridankina | 2 - 0 | July 13, 2026 |
| W | A. Prisacariu | 2 - 0 | July 12, 2026 |
| L | L. Bronzetti | 2 - 0 | July 7, 2026 |
All Aug. 2, 2026 results & picks →
E. Micic (A, age 22, rank 345) vs G. Garcia-Perez (B, age 34, rank 945) - Warsaw, HARD. Status: pending.
MODELS vs MARKET (A = Micic)
ELO 72.2% | ML 85.1% (outlier high) | blend/model 71.1% | market raw 1/1.37 = 73.0% (de-vig 67.1%)
Odds: A 1.37 / B 2.80. No exchange (alt_markets null) - illiquid WTA/ITF-level event.
RECENT FORM (last 5 each; all fetched matches used for classification)
Micic: 07-18 L Fruhvirtova(1478) clay 3-6 5-7; 07-15 L Zheng(1500) hd 4-6 1-6; 07-13 W Matoula hd;
07-12 W Kulikova hd 6-3 3-6 6-4 (GRIND, Athens Q Final); 07-11 W Ninomiya 6-0 6-2 (DOM).
Won an Athens hard-court qualifying event mid-July, then Prague SF on clay. Hard-comfortable.
GGP: 07-19 L Vedder(1600) clay 4-6 6-2 6-4 (Hamburg SF); 07-13 L Pridankina(1643) clay 6-1 6-1 (Iasi
Final, blown out); 07-12 W Prisacariu clay; 07-07 L Bronzetti hd 6-3 6-4; 06-23 L Shaikh clay.
Deep clay runs (Iasi Final, Hamburg SF) - a clay grinder.
SHAPE / PATH / SURFACE
Both idle ~14-15d (last matches 07-18 / 07-19) - rust risk both sides, neither fatigued; no extreme
fatigue spot, no haircut. Micic 22 (fast recover), GGP 34.
TRANSITION: GGP clay -> hard (Warsaw); recent form clay-built, weaker on hard. Favours A directionally
but already broadly priced; cannot manufacture numeric edge.
ELO/rank roughly consistent (A higher on both) - ELO credible, mildly ITF-farmed but not decisive.
GATES
G1 edge-origin: FAIL - at executable book price 1.37 (raw 73.0%), my A estimate ~71% (ELO 72 / blend
71) gives NEGATIVE edge. Only ML (85%, distrusted outlier) disagrees. B at 2.80: my ~29% vs 35.7%
implied, also no edge. No model-vs-market edge survives against the takeable price.
G2 consensus: stake-tier only - moot (no qualifying edge).
G3 fatal-risk: n/a (ML market, no bet).
PRICING: no exchange ML price (alt_markets null) -> book-priced segment needs edge >=8% vs book AND is
the segment the backtest loses on. Edge is negative, not +8%. Automatic PASS.
alt_markets: none at analysis time
VERDICT: PASS. ELO and blend agree with the market on A (~71-73%); the lone high ML print is not enough,
there is no exchange liquidity, and neither side prices as value. Efficient market - no stake.