Behind every pick on this site: data, models, and methodology — fully documented, no black box.
Every pick on this site goes through five stages of data collection, enrichment, and AI reasoning before it appears on your screen. No black box — here's exactly what happens.
We pull real-time odds from 28 bookmakers across 6 markets — 1X2, Over/Under, Asian Handicap, BTTS, Goalscorer, and Correct Score. For every selection, we identify the best available price so the AI always reasons against the strongest line on the market.

















Kickoff, venue, tournament stage, and lineups the moment they're released — starting XI, bench, formation, plus injuries, suspensions and anyone unavailable. The model goes in knowing exactly who's on the pitch and who isn't.
For both teams, we mine the last 10 matches across all competitions. Who started, who was a sub, who never made the squad. Goals scored, conceded, lead conversion, BTTS rate, late-game patterns. Per-player efficiency: minutes per goal, contribution rate, scoring under different game states. The AI sees more than form — it sees patterns.
We query Perplexity AI for the latest match-relevant journalism: manager press conferences, rotation hints, motivation angles, tactical comments, and off-pitch disruptions. Numbers tell you what; news tells you why. Both feed the AI's reasoning.
All of the above is bundled into a structured brief and sent to each AI model independently. Each model is instructed to find the 1-3 strongest edges, write its reasoning, and distribute exactly 10 units across its picks. One bigger play, smaller leans — the same way a sharp human bettor builds a portfolio.
Form tables tell you whether a team won. We tell the AI how, against whom, and with which players on the pitch. Here's a sample of what feeds every prediction.
Every player in the expected lineup gets a 10-match profile — 18 engineered metrics, not a stat line:
Beyond W-D-L — how a team actually behaves across its last 10:
Recent meetings, with full context:
We don't pick a single AI to trust. Every match goes to every model — frontier-tier and free-tier alike. The leaderboard then settles which one is actually adding value over time. No favouritism, just performance.
No cherry-picking. No deleted losses. Every prediction every model has ever made is here, with the original odds, the original units, and the result. Open data, public leaderboard, transparent ROI math.
Every tip is timestamped at creation, before kickoff. We can't (and won't) edit a pick after the fact. The full prediction history per model is browsable forever.
Profit = units × (odds − 1) for wins, −units for losses, 0 for voids. ROI = profit / total staked. Same formulas every pro tipster uses. No creative accounting.
Every model's win rate, profit, ROI, and best market — all visible. The good models float, the weak ones sink. You see exactly who's actually adding value.
Every pick links back to the bookmaker offering the highest price at the time. We don't pretend you got odds you couldn't actually take.
The experiment runs in public. Browse the live tips to see what the AI models are predicting, or check the leaderboard to see which ones are performing best over time.