TennisEdge's AI model finds no betting value in S. Shimabukuro vs R. Pacheco Mendez (Los Cabos on 2026-07-27) — the market is efficient, with S. Shimabukuro favoured at 1.62.
═══ S. Shimabukuro (A) vs R. Pacheco Mendez (B) — Los Cabos, hard, 1/16-finals ═══
MODELS vs MARKET (the whole story)
ELO A 87.9% | ML A 85.1% | model A 78.2% vs Betfair 1.74/2.16 → A ~57%, B ~46%
App flags value_pick A, edge +19%. That edge is MANUFACTURED, not real.
Resimulate: player_b matches_analyzed = 0, spw/rpw/hold = null → "Insufficient
serve/return data", sim win_prob = null. Pacheco Mendez is UNRANKED with an EMPTY
match history. ELO/ML are farmed against a phantom/default opponent rating — the
models have literally zero information on B. The 78-88% is a data artifact.
WHO IS PACHECO MENDEZ (path/context)
Unranked (ranking null, age null), no data anywhere. Los Cabos is a Mexican event;
an unrated local at 2.16 on the exchange is a home wildcard the market treats as a
LIVE underdog (~46%). The market prices knowledge the model cannot see. On an
unrated opponent the exchange is the sharp estimate, not ELO.
SHIMABUKURO — real negatives the inflated model ignores
Recent form (last 5, most-recent first):
06-30 Wimbledon (grass) L Faria (1688) 6-7 3-6 7-6 3-6
06-17 Halle (grass) L Tiafoe (1674) 6-4 7-5
06-16 Halle (grass) W Griekspoor (1672) 3 sets
06-13 Stuttgart (grass) L Shelton (1779) 3 sets
06-11 Stuttgart (grass) W Kyrgios (1820) 4-6 7-6 6-4
Won Stuttgart grass title 06-07. Strong GRASS run — but:
- TRANSITION grass→hard: every recent match is grass; this is hard. Depressed level
+ wider margin variance in match one on the new surface.
- RUST: last match 06-30, today 07-27 = 27 days idle (>21d threshold).
- Weak HARD form: last hard outings were April Jiujiang CHALLENGER QUALIES — beat
Shimizu (1569), lost to K. Smith (1587). His hard-court level is Challenger-tier,
not the 88% the ELO implies.
The market has already priced ALL of this down to ~57%. The model has not.
FATIGUE: both well_rested (0 matches 14d). Neutral. (Shimabukuro's 27d layoff = rust,
not fatigue.)
ELO/RANKING SANITY: A ranked 147, ELO-fav at 88% vs an UNRATED opponent → textbook
farmed/unreliable ELO. Discount entirely; lean on the exchange.
ALT MARKETS: model probs null; sim cannot run (no B inputs). No cover/total estimate
possible → alt markets un-priceable, excluded.
GATES
G1 edge-origin: FAIL — the +19% "edge" comes from ELO/ML built on a phantom
opponent (B has 0 matches, null stats). Strip the artifact and there is no
trustworthy model-vs-market disagreement; the exchange is the sharp number.
G2 consensus: FAIL — no hard independent evidence either way. B has zero data
(can't build a dog thesis); A's model edge is unreliable and off-surface/rusty.
G3 fatal-risk: FAIL — backing A's inflated number carries the negating risk that
B is a legit home prospect (market confirms via 2.16) while A is rusty + off his
surface with Challenger-tier hard form. Units cannot buy that back.
VERDICT: PASS. Classic manufactured-edge trap — the value_pick is generated by an
unrated opponent inflating the models, against a sharper exchange that already
discounts A's transition + rust + weak hard-court level. No priceable alt market
(sim can't run). No side has an articulable edge over the exchange. Pass.