About
Methodology
No black boxes. Every number on this site is either a stored fact, a documented calculation, or an explicitly labeled inference.
Card fit score
Each position has an attribute weight model. A card's attributes are multiplied by those weights, averaged, then normalized to 0-100 so a 40-99 attribute range spreads across the scale. Fit answers “is this card good at its job”, which is why an 86 OVR speed back can outrank a 90 OVR power back in a spread offense.
Ability fit score
Every ability carries fit dimensions (deep pass, quick game, man beating, edge rush…). Your playstyle produces weights for those same dimensions — first from onboarding answers, then replaced by your film. Ability fit is the weighted overlap, 0-100.
Ability point optimization
AP allocation is a bounded knapsack: each candidate has a cost in AP and a value equal to ability fit times positional importance. We solve for maximum value within your AP budget, limited to three abilities per card, and report the alternatives that lost.
Chemistry
Chemistry is a threshold system. We sum your equipped components per chemistry, find which tiers you have crossed, and apply the boosts to the affected attributes before recomputing fit. Recommendations target the cheapest path across the next threshold.
Improvement measurement
Every recommendation records the metrics it intends to move before you apply it, then re-measures them across your next games. A change is only called WORKING once the underlying numbers move, not because you won.
Film and Madden DNA
Uploaded games are broken into snaps with formation, concept, coverage faced, result and what MUTIQ flagged. Aggregated, those snaps become your Madden DNA: pass depth distribution, run rate, pressure rate, missed reads and formation predictability. DNA then overrides the onboarding priors used by every engine.
Confidence labels
Facts, calculations and inferences are never mixed. Verified data comes from confirmed sources, calculated values come from the models above, and inferred values are labeled so you know when the platform is estimating. Community-submitted data stays unverified until corroborated.