From model judgement to settlement, every KQ Prism AI step is reviewable.
KickQuant is an AI-based football match data analysis and quantitative prediction platform. KQ Prism AI brings match data, model probabilities, market prices, value signals, and post-match results into a single record chain, helping users understand how a signal is formed, when it is triggered, and how it is ultimately settled. The platform also provides team data views, risk planning tools, and a public historical audit.
About KickQuant
We connect model decisions, price verification, and settled results in one reviewable data chain.
KickQuant is an AI-powered football quantitative analysis and prediction platform. KQ Prism AI is its engine for probabilities, value signals, post-match verification, and performance records. We bring 300+ competitions and cups, market structure, prices, team form, xG, historical performance, and post-match settlement into one system. The goal is not only to show which direction the model favors, but also to verify whether market prices actually meet the model's requirements.
We aim for verifiable backtest metrics based on real trigger records, rather than using a high hit rate to create a short-term illusion.
Research Criteria Matter More Than a Single-Match Hit
An analysis can meet the research criteria only when the model probability is higher than the probability implied by the market price. A high hit rate does not necessarily mean a positive result: if low-price selections around 1.15 are chosen over time, even a very high hit rate can still be offset by a single drawdown event. KickQuant therefore evaluates price, hit rate, backtest metrics, maximum drawdown, and sample size together.
The model first freezes the candidate and model reference price. A member research record is triggered only when a verifiable price appears in the designated monitored market.
Historical performance includes only signals that were actually triggered and fully settled, and publishes hit rate, backtest metrics, maximum drawdown, sample size, and recorded price together.
A game is broken down into thousands of individual clues.
The backstage won’t just ask “who will win?” The same game will be divided into multiple sets of clues such as price, xG, ELO, main handicap, secondary handicap, handicap depth, main line of big and small goals, handicap movement, water level changes, team status, schedule pressure, home and away difference, H2H, injuries, lineup and league stage, etc., and enter different model lines in parallel.
Each line is scored independently based on historical performance, research criteria, backtest metrics, price stability, and drawdown risk.
Directions that pass data integrity and risk filters first become candidates; only after the price condition is met do they enter the member research record.
Football is unpredictable, so we expose our boundaries.
Football can be changed by red cards, penalties, injuries, weather, in-game tactics, and random events. KickQuant cannot promise that every prediction will be correct, and does not make such a promise. What we can do is preserve the full record of candidate generation, price observation, actual triggering, subsequent better observed prices, and post-match settlement, so long-term performance can be reviewed.
All supported markets are evaluated independently using the same process. Candidates that do not reach the reference-price threshold are clearly marked as not triggered and are excluded from member research records and performance statistics.