What is actually compared, and what each signal proves
A linkage decision is built from a short list of inputs, and each one proves something narrower than it looks. This page takes the nine signals in turn - identity fields, verified documents, contact details, payment instruments, device and network, and behaviour - and sets out what each can and cannot establish.
The three kinds of signal
joining Artefacts that should be unique to one person. A card number, a wallet address, a mobile number that has passed a verification text, a verified email address. Weak individually, decisive in combination - their whole value in a comparison is that one person is expected to have one.
separating Verified facts that differ. Two different legal names, two different passport numbers, two payout accounts in two different names. These are the only inputs that can positively establish that the accounts are not one person, which is why they are the strongest evidence a reader can supply.
ambiguous Shared merely because people share. A postal address, a home network, a family device, a surname. Real overlaps, no proof, and the reason a review band exists between the automatic outcomes.
The signals, one at a time
- Name and date of birth. A perfect match may be one person typing twice, a coincidence, or a relative who shares a birthday. A mismatch usually means two people - unless the second record was opened with another person's details, which is a different and more serious problem than duplication.
- Verified documents. The strongest separating input available. Two identities that both passed verification are two legal persons as far as the operator is concerned, whatever else overlaps, because verifying two different passports against two different faces is hard to do for one person.
- Contact details. An email address or a mobile number that passed a verification step is a mid-strength signal. Households do share a landline and sometimes an email; a verified mobile number is more often personal than not, which is where its weight comes from.
- The payment instrument. The signal that decides most cases on its own, because the terms separately require the holder's own funds and because a card number is a hard identifier issued to one person. It is also the signal a household most often shares by accident.
Behaviour, and why it is not evidence
The behavioural layer - session timing, stake shape, whether two accounts are ever used simultaneously - is genuinely useful to an operator and genuinely weak as evidence. Simultaneous activity is the closest thing to a proof of two people that a pure log file can produce, and it is not a proof: two people can play at different hours, and one person can play from two devices at once. The honest position is the one the comparison page takes: behaviour informs a review, and it does not decide one.
The corollary is worth stating because it is the reader's actual problem. Nothing in the signal set establishes who is typing. A decisive result means the operator's own rules have been met - a count of shared identifiers crossing a threshold it published - rather than that a human has been identified twice. That gap is not a loophole in the operator's position and it is not a defence for a reader; it is simply the shape of an inference built on identifiers, and it is why the mechanisms here read as data-handling rather than as detective work.
What a reader can see, and what they cannot
A reader can see their own declared details, their documents, their contact details and their payment instruments. They cannot see the device fingerprints, the network identifiers or the derived counts the operator holds. That asymmetry has one practical consequence and it runs through the whole desk: a reader cannot audit the comparison in advance, so the useful work is to know which of the nine signals they share with anybody else, and to be able to name the ones they do not.