World Cup Results
Every MAX VII match-result call, graded against the final score. Full honest accounting — every pick counted, wins and losses shown.
⚙ Important context — draw handling
The engine's draw-detection module was not integrated until 20 June (the Australia match), after development issues. Before that date, the engine could not call draws — it picked a team in every fixture, and drawn matches counted as losses. This is the single biggest factor in the early record. From 20 June onward the engine began correctly identifying draws, and accuracy lifted sharply — see the split below. The tournament-wide figure therefore understates the current engine, which carries the draw module.
Full Match Ledger
Reading the numbers
The table above is the raw, full-inclusion truth: every match-result call, counted, with the draw-module handicap left in. Below is what the record shows once examined more closely. Each figure is reported with its caveat — this is a development log of a system still being refined, not a marketing sheet.
Remove drawn matches entirely (the engine's known weak point) and look only at fixtures a team won where the engine picked a team: 33 of 38 correct.
Caveat: this is a conditional figure — it excludes draws by design, so it is an upper bound on winner-discrimination skill, not an all-in accuracy claim. Quoted only alongside the draw context. 95% CI [73%, 94%].
Backing every engine call flat-stake at the market's own prices returns a positive ROI — the technical definition of beating the market, which a market-following model cannot do. The standout contrarian wins: Australia @ 6.00, USA @ 2.00, Mexico @ 2.10, Norway @ 2.30, and the Egypt–Iran draw @ 2.70.
Caveat: a thin edge on n=48 — directionally the thing that matters commercially (beating the closing line is the real moat), consistent with genuine edge, but not yet proven at statistical significance.
From 20 June, once draw-handling was integrated, the all-in figure (draws included) sits at 67.9% — above the 55–60% band that defines elite pre-game soccer prediction. The tournament-wide 60.4% understates the engine as it now runs.
The same engine, across every code
The World Cup is one proving ground. The same architecture has been ported to every sport ClutchTip covers, reaching competitive accuracy each time:
Is Resonance even possible?
That was always the real thesis — whether a unique AI model composition, running on consumer hardware, could find the natural frequency of a domain and predict it. On the evidence to date: not only is the architecture possible — it is already reaching the accuracy band of the best commercial engines, at a fraction of their resource and infrastructure. The established models (XGBoost, the large ML systems) were refined over years with billions in research behind them. This is one developer, on a phone-accessible engine, with a fortnight-handicapped draw module still in beta. If the architecture can be refined to that same level of maturity, the question may no longer be does it work — but what can't it do.
Methodology & Notice
Match-result calls graded against verified final scores from multiple authoritative sources. Predictions are pre-game only and do not account for in-game variables. ClutchTip is a prediction and analytics service, not a gambling operator. This record reflects group-stage fixtures; the draw rate in group play is higher than in knockout rounds.