TennisEdge's AI bot is backing J. Lu (moneyline) at 3.95 in M. Hontama vs J. Lu (Vancouver on 2026-07-29).
M. Hontama (A) vs J. Lu (B) - Vancouver, hard, 1/8-finals
MODELS (win prob A): ELO 63.3pct | ML 65.8pct | blend 63.4pct | SIM 47.0pct (matchup-adj)
MARKET: Betfair A 1.32 (76pct) / B 3.95 (26pct); book A 1.25 / B 3.70. Value_pick B, edge ~11pct.
MODEL-vs-MARKET GAP: ELO and ML (both opponent-quality adjusted) put A at only ~63-66pct,
i.e. B at 34-37pct - well ABOVE the exchange 26pct. The market is pricing A ~13pp above
even her own opponent-adjusted ELO, an inflation off her marquee name/results.
SIM DIVERGENCE (resolved): point-by-point sim makes B the fav (53pct), diverging hard from
ELO/ML. Cause = input bias, NOT a real signal: Lu serve/return (spw .583) was farmed on
W35 fields (opponents elo 1241-1417), while Hontama (.561) were earned vs Slam-level
returners (Krueger 1792, C.Liu 1710, clay 1686-1697). The sim is not opponent-adjusted, so
it over-rates Lu / under-rates Hontama. I discount the sim toward the opponent-aware
ELO/ML (~37pct B). Both methods still put B ABOVE market 26pct, so B is underpriced on the
conservative read too.
RECENT FORM (last 5):
Hontama (A): W vs Day 1692 GRIND 3set (7.7-6.3 3-6 4-6, hard, YESTERDAY) | L Tagger 1697
(clay) | W Parry 1686 (clay) | W FINAL Ryser 1436 (clay, title) | W SF Palicova 1444.
Read: marquee edge (Wimbledon SF grass, French SF clay, Prague title clay) is largely
OFF-surface; on hard she is mixed. Transitioned clay->hard; won an ugly 3-set grind
yesterday. Shape STABLE-to-FADING on hard.
Lu (B): W vs Dolehide 1628 DOM (7-5 6-4, hard, YESTERDAY) | L Chavez 1353 | W Marton
1241 | W Plekhanova 1417 | W FREY 1282. Read: pure hard-court grinder, match-sharp
(deep in W35 Saskatoon then straight scalp of a legit 1628 hard-courter at Vancouver).
Shape SHARPENING on surface; Dolehide win is a real opp-quality data point she plays
above her W35 baseline. Serve stats inflated by weak fields (caveat).
PATH/SHARPNESS: Lu match-sharp on this exact surface; Hontama transitioned off clay, big
results off-surface. Both played yesterday (back-to-back today).
FATIGUE (Step 2 calibrated): Hontama WON a 3-SET match yesterday -> -4.5pp hangover
haircut (age 26, baseline) - weakens A / confirms B. Lu won a 2-set match yesterday
(~neutral) but carries heavy load (matches_7d 4, age 36 -> hardens). Net: A extra-set toll
slightly exceeds Lu load -> fatigue CONFIRMS (does not create) the B lean. Not used to
enlarge the staked edge.
ELO/RANK: A rank 242, B rank 387; A is ELO-fav, consistent. No material divergence.
MARKET CHOICE: alt_markets present.
- Total 20 (over 1.94): sim over~0.556 vs implied 0.528 -> ~3pct edge, thin -> PASS.
- Handicap A -4.5 / B +4.5 (2.08): abs line 4.5 HIGH-VARIANCE tier. Sim margin basis is
BIASED toward B, and Lu loss distribution is bimodal blowouts (0-6, 3-6, 1-6) - a
self-negating risk for a +4.5 cover. G3 (margin bet = HARD VETO) FAILS -> PASS.
- ML B @ Betfair 3.95: cleanest, lowest-contamination edge -> STAKE.
GATES
G1 edge-origin: PASS - B 37pct (ELO/ML, opp-adjusted) vs exch effective 26.3pct, +10.7pct
BEFORE any fatigue/narrative; pure model-vs-market disagreement.
G2 consensus: PASS (stake tier) - ELO AND ML both put B above market (2-model);
full ladder eligible, tempered to 1.5u for wide sim-divergence and dog variance.
G3 fatal-risk: PASS - ML bet, G3 never blocks. Favorite-strength risk (A = 2x Slam SF)
noted as risk_factor; stake kept moderate.
PRICING: stake Betfair B 3.95. Effective = 1+(2.95*0.95)=3.80, implied 26.3pct.
Edge = 0.37 - 0.263 = +10.7pct. min_odds = 1.05/0.37 = 2.84; 3.95 clears comfortably.
VERDICT: BET B (J. Lu) ML @ 3.95, 1.5u. Predicted match winner remains Hontama (models
majority + market), but B is a value dog: market inflated A ~13pp beyond opponent-adjusted
ELO, Lu is match-sharp on hard with a real 1628 scalp, and A off-surface edge plus a
3-set hangover confirm the fade.