Method version 1.1

How we score

Every score follows the same published steps, so you can check our work.

Every product on ChainMatch gets a score from 0 to 10. This page explains exactly how we get there, so you can check our work.

The short version

  • Each type of product has five or six criteria, and each criterion has a fixed weight.
  • We score each criterion from 0 to 10 against written descriptions of what a 0, a 5 and a 10 look like.
  • The overall score is the weighted average. Serious security problems cap it.
  • Every criterion score cites its sources. Commissions never enter the calculation.

Step by step

  1. Research. We collect facts from the company’s own documentation, independent security audits, incident reports and reputable news. Each fact is stored with its source.
  2. Score. A separate reviewer scores each criterion using only those facts and the descriptions for that product type. The reviewer doesn’t see whether we earn a commission.
  3. Check. A fact-checker confirms every claim against its source before anything is published.
  4. Publish. We show the overall score, the score for each criterion and a one-line reason for each.

Caps for security problems

SituationHighest possible score
A confirmed flaw in the product itself that has cost customers money, while people who used it are still at risk0.0
Any other unfixed security problem that puts customers’ money or personal data at risk4.0
A serious incident in the last 24 months that has been fixed, whether or not anyone independently checked the fix8.0

No product with an unfixed security problem can score a 5.

  • A flaw in the product itself. This means a fault in what the company designed or built: its hardware, its code, or the way it creates keys and recovery phrases. The 0.0 cap applies when three things are true. The company or an independent researcher has confirmed the flaw. Customers have lost money because of it. And people who used the product while it had the flaw are still at risk. A product under this cap scores zero whatever it does well, and it never appears in setup finder results. We keep its review up as a warning, marked “Not recommended” and with no button to buy it.
  • Attacks that need the device in hand. An attack that only works with physical access to the device and lab equipment, and hasn’t been shown to expose keys, doesn’t trigger the 4.0 cap. It lowers the security design score instead, and the review explains it. If researchers show it can expose keys or PINs, the 4.0 cap applies until it’s fixed.
  • Attacks that need the company’s servers hacked first. Some products encrypt your data so that even the company can’t read it. An attack that only works once the company’s own servers have been taken over, and hasn’t been used against customers, doesn’t trigger the 4.0 cap. It lowers the security design score instead, and the review explains it. If it’s used against customers or shown to have exposed their data, the 4.0 cap applies until it’s fixed.
  • Tampered devices and stolen signing keys. A report of devices tampered with before they reach buyers, or of a stolen key that lets fakes pass the maker’s authenticity check, counts as an unfixed problem once the company or an independent researcher confirms it, or customers lose money because of it. The 4.0 cap then applies until it’s fixed. This is a failure in how devices reach buyers, not a flaw in the device itself, so the 0.0 cap doesn’t apply, even when customers lose money. An unconfirmed report lowers the track record score instead, and the review explains it.
  • In setup finder results. A product under the 4.0 cap ranks below every product in its slot that isn’t, however well it fits your answers, and we never offer it as an alternative. It only appears if nothing else in that slot fits. A product under the 0.0 cap never appears.

Products we leave out of recommendations

Some problems are serious enough that we won’t recommend a product at all, whatever its score:

  • It ships with a recovery phrase already written down or set up. Real wallets always have you create your own.
  • The company has ever asked customers for their recovery phrase.
  • It claims to be “self-custody” but you can’t export or recover your keys without the company.
  • It requires a licence to operate where we recommend it, and doesn’t have one.

Hands-on or research-based

“Hands-on tested” means a person on our team set up and used the product. “Research-based” means we scored it from documentation, audits and public records. Every review says which one it is.

Keeping scores fresh

We re-check every score at least every six months, and straight away after a major security incident. Each review shows when it was last scored and which version of this method we used.

What we don’t score

We score networks on how well they work (fees, speed, decentralization, reliability), but we don’t score coins or tokens as investments, we don’t rate investments and we never predict prices.

Changelog

  • 1.1 (November 2026): Stricter caps for security problems. The highest possible score for a product with one drops from 5.0 to 4.0. A product with a confirmed flaw of its own that has cost customers money, while people who used it are still at risk, now scores 0.0 and never appears in setup finder results. For networks, the cost score may use the median fee for all transactions when no figure for simple transfers is published, or a worked estimate when no dollar figure is published at all. A short chain reorganization that settles within minutes, with no double spend, does not count against a network’s security score.
  • 1.0 (November 2026): First published version, covering exchanges, wallets, networks and safety tools.

Network data

Criteria by category