CS:GO Overwatch FAQ
The Overwatch lets the CS:GO community regulate itself by allowing qualified and experienced members of the community (‘investigators‘) to review reports of disruptive behavior, determine whether those reports are valid, and apply temporary bans if appropriate.
Prospective Overwatch Investigators are presented with an active Overwatch button in the main menu, which indicates that there is a pending case for them to evaluate. The investigators can then choose to participate by watching a replay (eight rounds’ worth or roughly 10 minutes) and selecting a verdict.
If the investigators collectively agree that an offense has occurred, a ban will be issued. The ban duration will depend on the severity of the offense and the suspect’s history of convictions, if any.
Investigators are selected based on their CS:GO activity (competitive wins, account age, hours played, Skill Group, low report count, etc.) and, if applicable, prior Overwatch participation level and score (a function of their accuracy as an investigator). Community members who maintain both a high level of activity and high Overwatch scores will receive more cases to elect to participate in.
Investigators are presented with a replay of a randomly selected eight-round segment from an accused player’s match, and their task is to determine whether or not that player has committed any offenses during that replay. The suspect is referred to as “The Suspect” and the other players’ names have been replaced. All text and voice chat has been omitted. The investigator is expected to make a determination solely based on the actions of the suspect.
The system contains replays of players whose pattern of being reported exceeds or otherwise stands out from those of their peers. They may have a sudden spike in reports, or they may have slowly built up many reports over a long period of time. Additionally, a player may appear in a test case replay that has previously received a not-guilty verdict. These are randomly inserted into a case load and are used exclusively to help the system score investigators (the verdict in test cases will not result in any action against the suspect).
There are four distinct charges to be evaluated that cover the major forms of disruption. Each charge has two possible verdicts, ‘Insufficient Evidence,’ and ‘Evident Beyond a Reasonable Doubt.’ Additionally, an investigator may choose to ‘Postpone’ a review and restart it at a later time without a resolution.
For each charge, if the system determines that the investigators’ collective judgment converges on an overwhelming verdict, then the case will be closed and all decisions rendered on the case will determine the final verdict. If the verdicts are overwhelmingly “Insufficient Evidence” or are inconclusive, the case will be thrown out.
Yes. A higher-scoring investigator’s verdict will carry more weight than a lower-scoring investigator.
An Overwatch score represents an investigator’s ability to consistently and accurately judge the evidence they review, per charge, in both real and test cases. Investigators score positively for agreeing with the majority of other investigators’ verdicts for the same replay, and score negatively for being in the minority. The resulting change in score is larger when most investigators are in agreement, and smaller when they disagree. Because of the occasional test case inserted into the case load, the only way to improve an Overwatch score is to carefully watch the entire replay and make an informed and accurate judgment of the evidence provided.
During the beta, the system will issue fewer cases and the results will be reviewed and analyzed before any bans go into effect. The system will be adjusted for reliability and accuracy. As the system is tuned, more cases will be made available to more prospective investigators. Eventually the system will become entirely community-driven.
Update May 29
So far, Overwatch investigators have found evidence of disruptive behavior in 90% of highly-reported cases, with the decision being unanimous in the vast majority of cases. In addition to our standard test cases, we recently added demos of pro CS:GO players (specifically NiP and Quantic) to the test pool. Every single test case has been correctly dismissed as “Insufficient Evidence”, with no false convictions made.
Currently, the best approach is to play lots of matches in our official Competitive Matchmaking. We are slowly adding players to the pool of investigators, and randomly pick them with consideration to their playtime and skill level. The goal is to invite as many skilled reviewers as possible.
Report disruptive behavior. Reports feed the pool of cases that get reviewed, and good reports result in efficient case reviews. To file a report, open the scoreboard, use the arrow keys to navigate to a player, and press enter to open their player panel.
It is not necessary for a player to be exceptionally skilled in order to be a good Overwatch investigator. For example, we would expect casters to be good investigators regardless of their skill levels, given their experience in observing highly skilled play.
Every new investigator will enter the pool with a low score, which only increases based on the accuracy of their verdicts. This process gives everyone a chance to participate while ensuring that only the best investigators drive case outcomes.
None of our test cases have led to a false conviction, and we are still calibrating our system to ensure that highly-skilled players are not at risk. One way we are calibrating Overwatch is by sending investigators evidence from matches played by top CS:GO players (including NiP and Quantic) in Classic Competitive Matchmaking. So far, Overwatch Investigators of all skill levels have correctly reviewed these cases as “Insufficient Evidence”.
Update June 11