What CS2Scan actually measures — and what it doesn't
An honest introduction to how the scoring works, which signals carry weight, and where the limits are.
CS2Scan reads every behavioural and statistical feature the Steam Web API exposes for an account, scores each one against a population baseline, and combines the results into a single probability with an explicit confidence level.
This post explains what that means in practice, because a number between 0 and 100 invites more confidence than it deserves.
The eight engines
Each engine scores one family of signals independently, then a weighted vote
combines them. An engine that cannot read the data it needs reports N/A and
drops out of the vote entirely rather than guessing.
- Network — ban density across the friend graph
- History — VAC, game and economy bans, plus account provenance
- Precision — headshot conversion rate
- Accuracy — shot-to-hit rate and cross-weapon-class consistency
- Efficiency — K/D, kills per hour, MVP rate, win rate
- Opacity — how much of the profile is deliberately hidden
- Anomaly — composite outlier distance across every measured axis
What the data says
We calibrated against roughly 3,000 accounts sampled from the Steam friend graph. The finding worth publishing: the friend-graph signal works, and the individual stat lines mostly don't.
Ban density across someone's friends separated banned from clean accounts with an ROC-AUC of about 0.75. Most of the per-account statistical signals sat much closer to a coin flip. That matches intuition — a genuinely excellent player and a careful cheater produce similar-looking stat lines, but a friend list where a quarter of accounts are banned is hard to assemble by accident.
The limits, stated plainly
Valve's stats endpoint is unreliable for CS2. On roughly three quarters of accounts we cannot read CS2 statistics at all — either the player hid them, or the legacy endpoint simply has no record. When that happens, several of the eight engines go dark and the confidence figure drops to match. The result page tells you when this has happened.
A VAC ban is not attributable to a game. Steam reports that an account has a VAC ban; it does not say which title. A years-old ban in an unrelated game looks identical to a CS2 ban.
The score is not evidence. It says an account's public metrics sit at some distance from a population baseline. Skilled players, smurfs and deathmatch grinders all produce elevated scores for entirely legitimate reasons. The only body that can determine cheating is Valve.
Use it as a structured second opinion. Report suspected cheating in-game — that is the channel Valve actually acts on.