Every fan has shouted the word. A team sits third in a power index and the group chat declares it overrated. Dictionary.com defines overrated simply as "too highly rated or praised," which is the entire argument in five words — rated too highly by somebody, by some method, for some reason. The shouting is easy. The harder question is what the rating actually measured.
The answer: rankings are built one of three ways. Some are voted by humans. Some are computed from results. Most are a blend, dressed up as pure math. Knowing which one you are arguing with tells you whether the complaint should aim at the voters or the formula.
That distinction is the whole point of this piece. Once you can name the machinery, you can argue with the right part of it — and you will spot the bias before it spots you.
What are the main ways rankings are calculated?
Three families cover almost everything you see weekly. First, polls: a panel of writers, coaches, or broadcasters each submits a list, and the lists are averaged. Second, formula rankings: a fixed recipe applied to results, so the same inputs always produce the same output. Third, power indexes: models that mix results with context — where a team played, how it won, who was missing.
Polls measure reputation as much as performance. A team that was good last season starts higher, because voters carry memory into the ballot. Formula rankings cannot carry memory; they only know what happened. Power indexes try to have it both ways, and that is where most of the weekly arguing lives.
None of the three is lying. Each is answering a slightly different question. The fight starts when a headline strips the method away and presents the number as a single truth.
Why do human voters keep getting it wrong?
Because voting is memory, and memory has shape. Voters watch highlights, not full matches. They remember the last game a team played far better than the five before it. They anchor on preseason expectations and adjust slowly — a team picked to finish first keeps getting the benefit of the doubt long after the results stop supporting it.
There is a name for the pattern: recency bias, the pull of the most recent result over the fuller record. Its mirror is halo bias, where one strong reputation rubs off on everything around it. A famous program starts 4-0 against weak opposition and climbs. A lesser-known side does the same and barely moves. Same results, different starting point.
None of this makes voters lazy. It makes them human, and it makes polls a measure of consensus perception rather than of quality. Both are real things. Only one belongs in an argument about who is actually best.
How do formula rankings differ from human polls?
A formula is a promise: same inputs, same output, every time. Most competitive formulas weigh results by context. Beating a strong opponent counts for more than beating a weak one. Winning away counts for more than winning at home. Margin matters, but usually with a cap, because running up a score against a beaten opponent tells you little.
The strength of the schedule — how hard a team's opponents have been — is the ingredient fans forget. An unbeaten run against soft competition can rank below a two-loss run against a gauntlet. When your team sits lower than its record suggests, that weighting is usually why, and it is usually defensible.
The weakness is different. A formula cannot see a red card, a back-up goalkeeper, a storm-delayed pitch. It treats every result as if the conditions were equal. They never are. Formulas are honest but blind; polls are sighted but swayed. Pick your poison, or read both and note where they disagree — the gap between them is often the most interesting number on the page.
What do power indexes actually measure?
A power index is a model's opinion. It takes results, adjusts for opponent quality and venue, and often blends in efficiency numbers — how a team performed per possession or per play, not just whether it won. The output is usually phrased as a strength estimate rather than a standing: not "this team is seventh," but "this team's underlying play rates about seventh."
Two indexes can use similar inputs and still diverge, because the weights differ. One might treat an early-season result as fully reliable; another discounts it while the sample is small. Sample size is the quiet variable underneath all of it — a truth we covered in depth in The Small-Sample Problem: Why One Great Season Rarely Predicts the Next. Three weeks of football is a rumor. Three months is a pattern. Rankings published in week three are mostly measuring week three. This connects to our earlier piece, The Small-Sample Problem: Why One Great Season Rarely Predicts the Next.
So when an index moves your team five spots after one match, that is not chaos. That is the model over-weighting fresh evidence, by design, because fans want a ranking that reacts. Responsiveness and stability pull against each other, and every index chooses a point between them.
Where does the bias hide in the math?
In the choices, not the arithmetic. Somebody decided how much margin counts. Somebody decided home advantage is worth a fixed bump. Somebody decided last season carries into this one, or does not. Each choice is defensible. Stack enough of them and the formula's personality shows — some indexes are conservative, some jumpy, some quietly favor the leagues with the best data coverage.
Data coverage is the bias nobody markets. A model can only weigh what it can see. Big leagues with full tracking data get richer inputs; smaller competitions get thinner ones, so their teams are measured with a blunter instrument. The ranking looks neutral. The inputs are not.
Our analysis, having watched and re-watched matches against the numbers for years: treat every ranking as one witness, not the verdict. Cross-check it against what you saw with your own eyes — the method we set out in How to Read a Game Like a Pro Analyst — Without the Fancy Software. If the number and the tape disagree, the tape usually wins, because the model never saw the match. You did. For related coverage, see How to Read a Game Like a Pro Analyst — Without the Fancy Software.
How should a fan actually use a ranking?
As a conversation starter, not a conversation ender. The useful move is to ask what the ranking is for. A playoff seeding formula answers a real question with real stakes. A weekly power index answers a fake question — who is best right now — that exists mostly to be argued about. Both can be worth reading. Only one should be taken seriously.
Practical steps: check whether the ranking is voted, computed, or blended. Look for the stated method; if a publisher hides it, that tells you something. Note where the ranking places your team against its results and its opponents' quality, not just its record. And when you disagree, disagree with the specific choice — the schedule weight, the recency pull — rather than the whole board. That is an argument someone can actually answer.
The gears are visible once you look. The word still gets shouted every week, and it should. But now you know which gear to kick.
