Why we do not publish detection-rate numbers

Sooner or later somebody asks for a detection rate, and it is a reasonable thing to want. We are not going to give you one, because the number does not exist: not that we do not have it yet, but that no observer in this market can compute it, us included. Every figure you have seen quoted was produced by choosing it. Here is the argument, because refusing to publish something without saying why sounds exactly like evasion.

A rate needs a denominator

A detection rate is bans divided by users. The numerator is already bad and the denominator is unobservable, because what a seller can see is orders, while a rate would require a different list entirely.

• How many people who bought actually installed anything
• How many hours each of them played, in which modes, against which anti-cheat
• Which firmware, which reader version, which input device and which second-PC configuration each one ran
• Whether a ban happened at all, given that we learn only when somebody chooses to tell us
• Whether this product caused the ban, rather than a bought account, a shared login, a report-driven review, or other software running alongside

Every line on that list is invisible from the seller's side. A store sees a payment and, sometimes, a support ticket, which is not a sample but a self-selected fragment of a population whose size the store cannot measure. Dividing one unknown by another does not become a statistic, and a percent sign after it changes nothing.

Enforcement is delayed and batched

Even with a known denominator, the numerator arrives in a shape that defeats attribution. Action in this field is routinely deferred: a signal is collected, it sits, and enforcement lands later across a window rather than against a build. When a wave lands, it touches everyone who played in that window at once, across several products, configurations and kinds of behaviour. Assigning a share of that wave to one firmware image is not a hard measurement problem but a question the available data cannot answer at all.

Self-reported data does not rescue it

The obvious fix is to ask people, and it does not work for a structural reason that a better form cannot solve. Those most affected by a detection are the least likely to answer, because a banned account often means a lost identity, a burned platform and no appetite for a survey. Survivors answer, so a survey of survivors measures survival among people who answer surveys. Nothing in a self-report is verifiable either: not that the respondent owned the product, not which build they ran, and not that the configuration they describe is the one they used. Collecting more responses does not converge on the truth; it only narrows the interval around a biased estimate.

What we publish instead

We publish facts with dates on them and their limits stated beside them. Every build carries a published state and the date it was published, next to the game's own last patch date, which makes the gap visible: when we last looked, and when the ground moved. Where a signal is weak the card says so, and where nothing is published the card says no status instead of guessing. The page says in plain words that this is not a clean bill of health, and when the feed cannot be read, it shows nothing at all rather than green.

None of that is as satisfying as a percentage, but it is the strongest honest claim available: on this date, we had published no detection for this build. That is not a measurement of your risk, and it is not a statement about your account.

What to ask a seller instead

• Show me a dated state per build, not one badge for the whole catalogue
• Show me a row that says detected, and tell me when it last changed
• Tell me what your page does when your own database is unreachable
• Tell me where a detection is announced first, and how quickly the page follows
• Tell me what the word undetected means on your site, in one sentence

A seller who can answer those is telling you how they behave when they are wrong, which is the only thing genuinely worth knowing in advance. A seller with a percentage is telling you they were willing to invent one.