What the lists mean

Every list on this site ranks something. Here is what each one shows, how it is worked out, and why it is worth tracking. The exact formulas, with the numbers this version of the site uses, are on how the numbers are worked out.

Best beliefs

How it is worked out. Every reason, finding and prediction under a belief is a row, and every row gets a signed score: its direction, times how far its own page has been argued from 50-50 (two times its truth score minus one), times how much work stands behind that page (confidence), times how relevant it is (linkage), how much it matters (importance) and how little it repeats another row (uniqueness). Each of those numbers is argued on a page of its own. A row argued true adds weight to its side; a row argued false counts against it; a row nobody has argued sits at 50-50 and adds nothing. Ties go to the belief with more scored rows beneath it.

Why it is tracked. So the best supported positions come first, and so that volume cannot buy a place: listing twenty reasons nobody has argued moves a belief exactly as far as listing none. The exact rule, with its numbers →

What it shows. Beliefs ranked by belief score: the weight of the rows for a belief minus the weight of the rows against it.

Most argued over

How it is worked out. Every reason under the belief is counted, then every reason under each of those, and so on down, each row once. A reason that appears in two places beneath the same belief is counted once.

Why it is tracked. To show where the debate has actually been worked through in depth, and where it has barely started. It measures how much has been argued, not how well, which is what Best beliefs is for. Hours spent arguing are not recorded anywhere, so they are not counted; the count of reasons is the part that can be checked.

What it shows. Beliefs with the most reasons for and against, counted at every level beneath them.

Hardest to resolve

How it is worked out. Each belief is classified from its own tables. A factual dispute: both sides have real weight in the evidence, so the fight is over what is true. A linkage dispute: most of the reasons are argued to be beside the point, so the fight is over whether the facts bear on the belief at all. A values conflict: supporters and opponents rank the same values in a different order (the gap in their rankings is the largest signal). A mixed dispute: no one of those three leads the others by at least 0.1. The order on the list (values, mixed, linkage, factual) is a judgment about which is hardest to settle, not a computed number.

Why it is tracked. Because the kind of dispute says what kind of work would move it. A factual dispute needs a measurement both sides accept in advance. A linkage dispute needs an argument about relevance. A values conflict needs a trade both sides can live with, because no new data will change which value somebody ranks first.

What it shows. Beliefs ordered by the kind of disagreement they are, from the kind no measurement can settle to the kind one measurement could.

Most relied on

How it is worked out. Picture a reader who starts at a belief, follows one of its reasons to that reason's page, then follows one of the reasons there, and so on, picking each next step in proportion to how relevant, important and distinct that row is argued to be. Now and then (15% of the time) the reader jumps back to a belief picked at random and starts again. Relied on is the share of that reader's time spent on each claim; all the shares add up to 1. Whether a claim is true is deliberately left out, so a claim does not drop off the list the moment it is argued false.

Why it is tracked. Because a claim many beliefs rest on is where one piece of work moves the most: settle it and every belief above it moves. It is also the honest stand-in for "popular". Nobody's votes or views are counted here, so this says how much rests on a claim, not how many people like it. The exact rule, with its numbers →

What it shows. The claims the most of this site depends on: the reasons, findings and assumptions that the most beliefs rest on, directly or through other claims.

What to argue next

How it is worked out. Relied on times the share of the work still undone (one minus confidence). A claim everything rests on but nobody has worked on comes first; a claim everything rests on and that is already well argued drops down.

Why it is tracked. To point the next contributor at the work that matters, instead of at whatever is most talked about. The exact rule, with its numbers →

What it shows. The claims where one more hour of work would move the most.

Who has a stake

How it is worked out. Every interest on the site is written as a group and a need, "<who> need <what>", so the group is read off the sentence. Groups are ranked by how many beliefs list one of their interests, then by how many interests they have; under each group, its needs are ranked by how valid each is argued to be on its own page. Validity asks whether the need should be honored, and is settled by the mirror test: would the group accept the same need honored for its opponents, and what happens if everyone pursues it? It is never set by who holds the need, how many of them there are or how much power they have. On each belief page a second number, Drives, asks whether the need is really why that side holds its position, settled by the side's record rather than its rhetoric (the scoring methodology calls it Linkage Accuracy), and Validity x Drives names the one interest actually driving each side. A need that is real and does not hold up is scored, not deleted, and the same suspicion is applied to both sides. A group that is only a belief's supporters or opponents is a side rather than a party, so it is kept but listed last.

Why it is tracked. Positions are what people say they want; interests are why. A compromise is built out of interests, so knowing who has a stake, what they actually need, and which of those needs hold up is where a way through starts.

What it shows. The groups of people with something at stake in the beliefs here (voters, investors, officials, parents, workers), and what each group needs, like "voters need enough information to judge an official".

Best books, studies and reports

How it is worked out. Best, so far, means impact: how much of these pages the work moved, which is its contribution to every belief that cites it, counted through the row it sits on. A work is listed by what it changes. Whether it deserved to is a separate number, quality, argued on the work's own page: is it accurate, well reasoned, well made, built on primary sources? The two are kept apart because a widely read work that is wrong and an unread work that is right are different problems. Other yardsticks for best (most important, most accurate, most entertaining, most cited) are each a criterion anyone can propose on a work's page and argue; a yardstick nobody has argued is not used to rank anything.

Why it is tracked. Because a work reaches people in a way an argument never does: one film teaches its ethics to more people than every analysis of those ethics put together. Knowing which works moved the beliefs here, and whether they earned it, is how a reader finds the best thing to read on a question, and how a work that moved a lot of people with a claim that does not hold up gets noticed.

What it shows. Every book, study, report, article and film cited on a belief page, by kind, each kind ranked. Under each work: what it is, and the belief it bears on most.

Changed in this revision

How it is worked out. The tables behind this version are compared with the previous version, row by row; the pages are listed most relied on first.

Why it is tracked. So a returning reader can see what is new without rereading everything, and so every change to a number can be traced to the row that caused it.

What it shows. Pages edited, or whose numbers moved, since the last published version of the site.

Votes vs the analysis

How it is worked out. A vote is one GitHub account saying agree or disagree with a claim, and only that account's latest vote counts. The share of the votes that agree is compared with the claim's truth score, and the list is ordered by the size of the gap, widest first. A claim enters the list once three or more people have voted on it; fewer votes are shown on the claim's own page but rank nothing. The card appears only once a claim has that many.

Why it is tracked. Because a claim the analysis rates low that many people agree with, or the other way round, is where a reason is missing from the page: either the people know something the tables do not, or the tables know something the people do not, and the next argument goes there. The votes themselves never move a score, because a claim nobody has argued is worth nothing however many people like it.

What it shows. Claims people have voted on, ranked by how far the share of votes to agree sits from the truth score.

Topics

How it is worked out. Each belief is filed under one topic. A topic page places every belief on three scales: which way it points, how strongly it is worded, and how general or specific it is. The "Most argued" line at the top of the shelf orders topics by the reasons counted beneath their beliefs, at every level, the same count Most argued over uses; a topic and the topic above it are never both listed.

Why it is tracked. So that two ways of saying the same thing land in the same place, and so a reader can see the whole range of positions on a subject at once.

What it shows. Beliefs filed by subject, like a library shelf.

Built from revision 42c8344, committed 2026-10-04. 2173 pages. Every number above is computed from the two tables in that revision, so a reader who fetches it gets these numbers and not different ones.