TennisEdge's AI model finds no betting value in G. Knutson vs S. Costoulas (Warsaw on 2026-08-03) — the market is efficient, with S. Costoulas favoured at 1.50.
Independent estimates of who wins, each from a different method. The blend is what the bot prices against the market.
| Method | G. Knutson | S. Costoulas |
|---|---|---|
| Elo ratings-based, surface-weighted |
39.6% | 60.3% |
| Machine learning 29-feature gradient-boosted model |
23.9% | 76.1% |
| Serve / return point-by-point simulation |
45.8% | 54.2% |
| Blend what the bot actually prices off |
39.6% | 60.3% |
| Market implied de-vigged from the opening odds |
36.6% | 63.4% |
We publish how well these score against the closing market, including where they lose: how accurate are tennis predictions?
| Player | Surface | Elo | Record | Matches |
|---|---|---|---|---|
| G. Knutson | hard | 1531 | 131–86 | 217 |
| S. Costoulas | hard | 1570 | 131–79 | 210 |
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. 4, 2026.
Level at 1–1 over 2 meetings. Full record.
| Date | Event | Surface | Winner | Score |
|---|---|---|---|---|
| Aug. 4, 2026 | Warsaw | hard | G. Knutson | 2 - 0 |
| April 2, 2025 | W50 Nantes | hard | S. Costoulas | 2 - 1 |
| W | S. Costoulas | 2 - 0 | Aug. 4, 2026 |
| L | R. Sramkova | 0 - 2 | July 26, 2026 |
| W | S. Zhang | 1 - 2 | July 25, 2026 |
| L | D. Salkova | 2 - 0 | July 19, 2026 |
| W | A. Ito | 0 - 2 | July 18, 2026 |
| L | G. Knutson | 2 - 0 | Aug. 4, 2026 |
| L | M. Inglis | 0 - 2 | July 29, 2026 |
| W | L. Cabrera | 0 - 2 | July 29, 2026 |
| L | A. Aksu | 2 - 1 | July 20, 2026 |
| W | A. Aksu | 2 - 0 | July 19, 2026 |
All Aug. 3, 2026 results & picks →
=== G. Knutson (A) vs S. Costoulas (B) - Warsaw, hard ===
MODELS vs MARKET
Market: Knutson 2.6 (impl 36.6pct) | Costoulas 1.5 (fav ~66.7pct)
ELO: Knutson 39.65pct / Costoulas 60.35pct
ML: Knutson 23.88pct / Costoulas 76.12pct
value_pick A, value_edge 3.06pct (ELO vs market only). Models SPLIT hard on
the dog: ELO 39.65 vs ML 23.88 for Knutson - no consensus; the only edge is a
thin 3pp ELO-over-market read that ML flatly contradicts.
RECENT FORM (last 5)
Knutson (29, rk182): 07-26 Washington F Lost Sramkova(1672) 3-6 2-6 (DOM loss);
07-25 Washington SF Won Zhang(1733 Strong) 3-set grind; 07-19 Prague F Lost
Salkova(1567); 07-18 Prague SF Won Ito(1197 weak); 07-15 Istanbul-Q Lost
Tararudee(1500) 3-set. Two finals reached, both lost; ELO likely partly
farmed on ITF W50/W75 fields.
Costoulas (21, rk146): 07-29 Vancouver Won Cabrera(1539) + Lost Inglis(1646)
same day; 07-20 Prague Lost Aksu(1470) 3-set; 07-19 Prague F Won Aksu(1485)
7-5 6-2 (TITLE, clay ~14d ago); 07-18 Prague SF Won Hibino(1501) grind.
French Open finalist (May). Young, higher-ranked, deeper pedigree.
SHAPE/PATH: Both mixed/STABLE. Both on hard recently (Washington / Vancouver) -
no meaningful surface transition either way.
FATIGUE: Knutson last 07-26 -> 7d rest, fresh. Costoulas last 07-29 -> 4d rest,
recovered. No extreme spot, no 3-set-hangover. Context only, not an edge.
NARRATIVE: Costoulas TITLE Prague ~14d ago - beyond letdown window. None either.
PRICING: alt_markets none at analysis time (all null). Book-priced only -> needs
edge >=8pct. Staked ELO edge 3.06pct. min_odds Knutson = 1.05/0.3965 = 2.65 >
book 2.60 -> below min-odds automatic PASS regardless.
GATES
G1 edge-origin: FAIL - only 3.06pct ELO-vs-market on Knutson, below 5pct floor
and far below the 8pct book-priced bar; ML contradicts it entirely.
G2 consensus: stake-tier only - lone-model (ELO) split, ML opposes.
G3 fatal-risk: n/a (ML market).
VERDICT: PASS. Tiny 3pp edge on the dog, models split, price below min-odds, no
exchange liquidity. No stakeable edge.