Schedule scanning
Eligible matches are collected from the configured leagues and the data required for analysis is verified.
The CornerEngine workflow narrows the match universe into ranked evaluations and a small set of PLAY candidates eligible for publication, while retaining WATCH and AVOID for analysis.
Eligible matches are collected from the configured leagues and the data required for analysis is verified.
The engine normalizes six signals: prediction, recent form, trend, home context, H2H and consistency.
Each evaluated candidate receives a weighted Score and calibrated Confidence. Configured thresholds, risk and decision guards produce the final PLAY, WATCH or AVOID verdict.
Candidates are ranked primarily by Confidence and Score. Under the current policy, only PLAY for Over 8.5 is eligible for publication; WATCH and AVOID remain evaluation decisions.
After the match, the observed corner total factually settles Over 8.5: 9+ corners = WIN, 0-8 = LOSS. The result then enters history and performance evaluation.
Score, Confidence and the verdict are different indicators and none is a guarantee.
A 0-100 score built from the weighted contributions of the six analysis factors.
An internal 0-100 score calibrated from Score plus consistency/risk adjustments. It is not a guaranteed win probability.
PLAY passes policy and decision guards; WATCH indicates mixed or insufficient evidence; AVOID indicates exclusion or high risk.
Public settlement uses only the observed corner total: 9 or more = WIN, at most 8 = LOSS.
Statistics tracks published picks. Performance evaluates a broader public-safe universe, so percentages can use different denominators.