Reliability (how much to trust the pick)
RobustTrust. Both players well-rated on fresh data, no guard fired. The only tier worth betting.
ThinUnproven. A player's rating hasn't converged (few rated matches).
StaleBlind. A player's recent results are invisible to the data (lower-tier gap / long absence) — form unproven.
OverratedOpponent rising. The opponent's current ranking says they're far better than their match history shows — % capped toward the ranking.
DoubtedMarket disagrees. The betting market prices this pick far below the model — the market is usually right. % capped toward the market.
Model vs market chip (on picks with a live line)
✓ ALIGNEDBoth signals agree (within 3pp) — the strongest picks.
△ MODEL +6ppModel runs 3–10pp hotter than the market — mild caution.
▼ DISAGREE 15ppBig gap (10pp+). The market knows something the stats don't — avoid.
▲ MARKET +5ppMarket is even more confident than the model.
Match status
LiveIn progress now.
UpcomingNot started — the bettable window.
Won LostFinished — did the pick win?
The numbers on each card
78%Calibrated probability — the model's number after self-calibration to real results, the 90% ceiling, and any guard caps. This is the honest one.
model 91% →Raw model output before corrections — shown when it differs from the headline.
market 64% (+147)The live exchange line: implied win % and American odds for the pick.
🌱 grass-adjGrass-specific ratings used for this match.
Underdog-v1 paper strategy
DOG +150The model-selected player is also a strict market underdog: decimal odds >2.00, longer than the opponent, and market probability below 50%. Thin and stale data are excluded; Doubted is allowed because disagreement is the hypothesis being tested.
Every qualifier is frozen as a $1 paper single. Five or six qualifiers open a $10-total By-2 Round Robin. Fewer than five means no RR—favorites never fill empty places.
Golden rules
1. General board: trust Robust + ✓ ALIGNED. 2. Never pay odds shorter than the pick's fair line. 3. Underdog-v1 is a separate paper experiment that intentionally studies market disagreement. 4. Who-wins reads and underdog patterns are not proven +EV.