How the numbers are actually built.
The full technical reference — the six-axis system, how consensus is blended from up to 40 sources, when everything updates, and exactly what a fixture needs to become a Prime Pick.
The six-axis rating system
Attack
Chance creation and finishing quality, weighted toward recent matches over the full season, so a club's current attacking output is reflected faster than a season-long average would allow.
Defense
Chances conceded and defensive solidity in open play — set-piece defense is deliberately tracked as its own separate axis rather than folded in here.
Form
The most reactive axis — a five-match lookback window, updated after each fixture, designed to move faster than the other five when something genuinely changes.
xG
Expected goals for and against, capturing chance quality rather than just chance volume — a team can generate plenty of shots and still read low here if those shots aren't good ones.
Set Pieces
Corner and free-kick routines, both ends of the pitch — kept separate from Attack and Defense because set-piece performance often doesn't correlate with open-play performance.
Discipline
Cards and disciplinary record — a quieter axis that rarely drives a headline number on its own, but genuinely moves a club's Overall score over a season.
How the consensus number is built
Update cadence
Prime Pick qualification criteria
Calibration bands how accuracy is actually checked
| Band | Expected hit rate | Checked against |
|---|---|---|
| 50–59% | Roughly 50–59% across the band | Every settled fixture in that range |
| 60–69% | Roughly 60–69% across the band | Every settled fixture in that range |
| 70–79% | Roughly 70–79% across the band | Every settled fixture in that range |
| 80–89% | Roughly 80–89% across the band | Every settled fixture in that range |
| 90%+ | Roughly 90%+ across the band | Every settled fixture in that range |
Common questions
Why does Overall use equal weighting but Power Rating doesn't?
Overall is meant to answer "how strong is this club, in general" — equal weighting treats all six axes as equally informative for that question. Power Rating answers a narrower question, "who's trending right now" — which is why it leans on Form and xG specifically, the two most reactive axes, rather than treating all six as equally relevant to momentum.
Why is Form a five-match window specifically, not more or fewer?
We tested several lookback windows before settling on five — shorter windows were too noisy match to match, longer windows moved too slowly to reflect genuine recent form. Five matches was the point where the axis stayed reactive without swinging wildly on a single result.
Does a new league start with full calibration confidence?
No — a newly added league is held to the same minimum-sample standard as everything else on the platform. Its numbers are published, but treated with more caution until enough matchdays have accumulated to check calibration honestly.
Is the methodology the same for every league?
Yes — the six axes, the consensus blend, and the calibration process are applied identically regardless of league. What varies is how quickly a specific league's numbers stabilize, which depends on how much verified history exists for it, not a different formula.
Where do I go to see this methodology applied to a real result?
The results page grades every fixture against its pre-match consensus. AI Reports aggregates that into weekly and monthly summaries. This page is the reference for how the numbers work; those two are where you'd check whether they actually held up.
Why this page exists as one reference, not scattered across every other page
Every fact here has already been stated somewhere else on the site — this is where they're gathered into one place, so you don't have to go hunting page by page for the piece you need.
Almost every page on FDVS explains a piece of how its own numbers work — team analysis explains the radar, AI Research explains calibration, Prime Picks explains its own selection bar. This page exists to put the complete technical picture in one place, for anyone who wants the whole reference rather than piecing it together page by page. Nothing here is new information — every fact on this page has already been stated somewhere else on the site; this is where they're all gathered in one place, cross-referenced against each other rather than scattered.
Why the six axes are fixed, not adjustable
Letting anyone reweight the six axes to their own preference would make every rating mean something different depending on who's looking at it — which defeats the purpose of a shared reference number. Attack, Defense, Form, xG, Set Pieces and Discipline are fixed and equally documented for everyone, so a rating means the same thing regardless of who's reading it. A personalized weighting system might feel more tailored, but it would also mean no two people could compare notes on the same club's rating with any confidence they were looking at the same thing — which is exactly the shared reference this page exists to protect.
Why Overall and Power Rating use different formulas on purpose
These two numbers are frequently compared side by side on Club Rankings and Power Rankings, and it matters that they're not secretly the same calculation wearing two names. Overall is a general strength read; Power Rating is a momentum read. When they agree, that's a genuinely stronger signal than either alone. When they diverge, that gap is informative precisely because the two numbers were never supposed to move in lockstep — a club with a strong Overall score but a cooling Power Rating is telling you something specific that a single number never could, and treating the two as interchangeable would erase exactly that signal.
Why calibration bands, not a single accuracy percentage
A single "FDVS is X% accurate" headline number would flatten five genuinely different situations into one figure that doesn't actually describe any of them well. A 90%+ consensus and a 55% consensus should behave completely differently in practice, and checking them against separate bands is the only honest way to confirm each one is doing what it claims — a single blended accuracy figure would hide exactly the kind of miscalibration this process is built to catch. Five bands checked honestly are worth more than one impressive-sounding number that can't actually be verified against anything specific, and a single figure would also make it much harder to notice if one particular confidence range started drifting while the others stayed accurate.
Why Prime Pick's criteria are published in full, not kept vague
A selection process that isn't explained invites the assumption that it's arbitrary. Publishing the exact odds band, the model-confidence requirement, and the editorial sign-off step means anyone can judge for themselves whether a specific pick's inclusion makes sense — rather than being asked to simply trust that it does. The criteria on this page are the same criteria applied to every single pick, not a simplified version written for public consumption while something different happens behind the scenes, which is the same standard applied to every other number described on this page.
How to use this page alongside the rest of the platform
This page won't tell you anything about a specific fixture — for that, the fixtures board, match analysis, and team analysis pages are where the methodology described here actually gets applied. Treat this as the reference you'd come back to when a number somewhere else on the platform raises a "wait, how is this calculated" question, not as a page you'd check before every fixture. It's built to be looked up occasionally, not read start to finish before every visit to the site — a technical reference works best when it's there for the moment you need it.
What this page won't tell you
Deliberately absent here: any claim about future accuracy, any promise about specific outcomes, anything resembling betting advice. This is a description of process — how a number gets built and checked — not a case for why any specific number should be trusted blindly. The results archive and AI Reports are where you'd go to judge whether the process described here actually delivers, not this page, which stays intentionally limited to explaining the mechanics rather than making a case for them.