Can A VAC or Overwatch Ban Be Removed?

A conversation on the CSGO Discussion Section excited me today. It came from a user who asks: “Through Overwatch, can a player be incorrectly banned?

The short answer is “yes, but it’s extremely unlikely.” User Pestilence   summed it up quite well in the first reply: “if u really dont use hacks u will be fine”.

User funnyS   gives a more technical explanation, saying that it’s “[v]ery unlikely, Overwatch is 99.8% accurate, and there’s a huge difference between luck/skill, and cheats.”

Then, of course, I came along and decided to fuck shit up by saying the same thing but with slightly different wording.

From experience monitoring players’ ban status, the answer is yes, they can be incorrectly issued. Game bans that have been issued by Overwatch (or another source, since the game you’re banned from isn’t public) are retracted when the bans were erroneously issued. On the other hand, I’ve seen “game ban” statuses convert to VAC bans.

This post struck fear into the hearts of many of the thread’s participants, no doubt causing rapid onset of tachycardia and foaming of the mouth for Valve’s most ardent defenders of the CSGO forum. However, the conversation that ensued has shown a spotlight on just how vague the wording is for the treatment of VAC (Valve Anti Cheat) and Overwatch bans after they are issued.

The fact of the matter is, once a VAC ban or Overwatch ban (listed on profiles as a more general “game ban”) is issued on an account, it is very unlikely to be removed.

On June 18th, I issued a public retraction on social media regarding three Steam profiles that had been listed as “game banned”. In the CSGO Thread, user AzKat  pointed out that a PlayerUnknown’s Battlegrounds ban wave (which does not use VAC or Overwatch) had resulted in innocent players getting a “game ban” status on their account, resulting in mass ban retractions. This incident was covered in IGN and

The newest iteration of VAC, named “VACnet”, utilizes deep learning and social credit scoring (CSGO Trust Factor) to hasten the onset and efficiency of VAC bans. It allegedly works in tandem with the Overwatch system, which is a player-led organization of game investigators who have been proven to accurately identify cheating players in a CSGO match.

The trade-secret nature of VAC’s inner workings lead official explanations few and far in-between, but the VACnet trade presentation demystify some of the most basic concepts that cheating programs fail to account for. Below you’ll find 3kliksphilip’s analysis of the VACnet 2018 GDC Conference, whose points I’ll add on.

Two highlights to take from the video are at 3:09 and 4:21.

The claim of VACnet set forth in 3:09 is that it will rectify a mistakenly issued ban “clearly and immediately”, which I have witnessed for at least 3 game bans (though they were not VAC bans).

The other claim of VACnet made in 4:21 is with regard to Overwatch cases being completed. When an Overwatch case is being investigated, Overwatch investigators must determine whether or not an accused player (called “The Suspect”) is cheating beyond a reasonable doubt or dismiss the case as having not enough evidence.

Overwatch’s mechanisms are described by 3kliksphillip summarizing a piece of the Q&A section by quoting as follows, “if judged by a particularly reliable Overwatchers, then it can be done in as little as 7 viewings. But if the case is a little more difficult, then it can take 50 to 100 viewings, all for the same demo.”

In the realm that I cover with CSGO Competitive Extreme, which is primarily composed of competitive matches in the Silver rank, the chances of players being convicted using Overwatch by a jury of 7 is not unheard of. In fact, the social science done regarding smaller juries, and issuing incorrect verdicts, is well recognized. A study in the Journal of Applied Social Psychology (doi:10.1111/j.1559-1816.1985.tb02262.x) found that smaller juries with experienced jurors tend to convict slightly more than non-experienced jurors. Similarly, a 1975 study by Valenti and Downing (doi:10.1037/0022-3514.32.4.655) indicate than criminal cases led by a strong prosecution will convict more people with a smaller group of jurors (a six-person jury) versus a larger group of jurors (a 12-person jury).

Obviously, matters concerning incarceration come with a greater deal of magnitude than a negative stamp on a gaming account. However, just as it is reasonable to consider the possibility of a jury issuing a false verdict, it is also possible for processes modeled by humans (e.g: VACnet) to get it wrong, too.

With that being said, if you are an innocent player, please do not be alarmed. I am not trying to suggest you will be banned randomly or by mistake. However, as the technology improves, their ideal efficiency and certainly in detecting cheaters will approach a value of 1.