TennisEdge's AI model finds no betting value in Z. Pawlikowska vs L. Karatancheva (Warsaw on 2026-08-04) — the market is efficient, with L. Karatancheva favoured at 1.40.
Independent estimates of who wins, each from a different method. The blend is what the bot prices against the market.
| Method | Z. Pawlikowska | L. Karatancheva |
|---|---|---|
| Elo ratings-based, surface-weighted |
60.5% | 39.5% |
| Machine learning 29-feature gradient-boosted model |
43.6% | 56.4% |
| Serve / return point-by-point simulation |
41.9% | 58.1% |
| Blend what the bot actually prices off |
43.4% | 56.6% |
| Market implied de-vigged from the opening odds |
32.6% | 67.4% |
We publish how well these score against the closing market, including where they lose: how accurate are tennis predictions?
| Player | Surface | Elo | Record | Matches |
|---|---|---|---|---|
| Z. Pawlikowska | hard | 1399 | 55–47 | 102 |
| L. Karatancheva | clay | 1430 | 72–83 | 155 |
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 10, 2026.
| L | V. Allen | 2 - 0 | July 30, 2026 |
| W | S. J. Visscher | 2 - 0 | July 29, 2026 |
| L | V. Hruncakova | 0 - 2 | July 10, 2026 |
| W | M. Teixido Garcia | 2 - 0 | July 9, 2026 |
| W | K. Lazarenko | 2 - 0 | July 8, 2026 |
| W | N. Hurkacz | 2 - 0 | Aug. 2, 2026 |
| L | B. Ricci | 0 - 2 | July 17, 2026 |
| W | G. Maduzzi | 2 - 0 | July 16, 2026 |
| W | A. Zucchini | 2 - 0 | July 14, 2026 |
| L | L. Boskovic | 2 - 1 | July 11, 2026 |
All Aug. 4, 2026 results & picks →
Z. Pawlikowska (A) vs L. Karatancheva (B) - Warsaw, hard, 1/16-finals
MODELS vs MARKET
ELO A 60.5% | ML A 43.6% | blend 43.4% | market no-vig A 32.6% (book 2.9 / 1.4)
System value_pick A, edge +10.9% (blend-vs-implied). Models split hard: ELO makes
A a favourite, ML/market make B a clear favourite. Question is WHICH model is right.
ELO vs RANKING SANITY (decisive)
A ranked 406, B ranked 309 - B is the HIGHER-ranked player, yet ELO makes A a 60%
favourite. Classic high-ELO / modest-rank divergence => A's ELO is FARMED at weak
W35/W50 ITF fields. Discount that ELO edge; only back A if live form independently
backs it. It does not (below). The blend's +10.9% value is a stale-ELO artifact.
RECENT FORM
A (Pawlikowska, 21, #406): W35/W50 ITF grinder. Last: L V.Allen 1500 (6-3 7-5),
W Visscher 1500 (6-4 6-4), then L Hruncakova 1539 (1-6 4-6). Wins are DOM only vs
sub-1300 fields; loses to anyone 1500+. No sign of favourite-level quality.
Shape: STABLE-to-FADING (lost most-recent match to an avg opponent).
B (Karatancheva, 22, #309): higher-ranked, deeper wins. On clay beat A.Rus 1560,
Paquet 1520, made W75 Hague SF (L Boskovic 1535). Just won a Warsaw final on HARD
Aug 2, 6-1 6-1 vs Hurkacz. Match-sharp on surface, clear quality-of-win edge over A.
Shape: SHARPENING (title yesterday, straight-set demolition).
FATIGUE (Step 2)
A: last match Jul 30 (4d rest), 2 matches/155min in 14d - well-rested, no spot.
B: won a 2-SET match yesterday (53min, 6-1 6-1) => ~neutral (~-2pp), not the 3-set
hangover haircut; negligible. No extreme spots either side. Roughly equal.
NARRATIVE / SURFACE
B transitioned clay->hard but already re-adjusted (won on hard Aug 2). No let-down
(final win was a minor event and FUELS sharpness here). Nothing supports A.
PRICING
No betfair_ml / alt_markets => book-priced only, low-liquidity small event (the
segment the backtest loses on; Pricing #4 => prefer PASS). Nominal ML edge on A vs
book = +9.1% (43.6% - 34.5%), but that 43.6% inherits the farmed ELO and is
contradicted by rank + live form. B at 1.4 is short with no clear over-round value.
alt_markets: none at analysis time
GATES
G1 edge-origin: FAIL - the +9-11% edge is a farmed-ELO model artifact; live form
and ranking (B #309 > A #406, B holds the 1500+ wins) refute A being competitive.
G2 consensus: stake-tier only - models split, ML is the lone A-supporter and it
is contaminated by the stale ELO; no independent evidence for A.
G3 fatal-risk: n/a (ML, no margin bet).
VERDICT: PASS. The lone numeric value on A is manufactured by a farmed ELO that
rank and quality-of-win data directly contradict; the market/ML read (B favoured) is
the credible one. No exchange price, no validated edge on either side.