Let's go over methodology and why it's really, really unlikely that this was a false positive. There's a higher chance of winning the jackpot in a lottery (7.1511238420185E-6%) than a false positive in this situation.
Data Collection
Approximately 90,000 kills were collected since August. These were split into two datasets - labeled and unlabeled.
Labeled Dataset
- Kills made by users that were signed up on the forums were automatically labeled as legitimate.
- Kills made by me and a few other users on the dev server using various hacks were labeled as hacking.
- 30% of the labeled dataset was excluded from training the detector, and was instead use to provide an accurate measurement of the detector's real accuracy. This is called a validation dataset.
Unlabeled Dataset
- Kills made by users that were not signed up on the forums. 45749 kills made by 2042 different players.
Detection Methods
I ran two different models on the unlabeled dataset - one tuned for blatant aimbots, another tuned for subtle aimbots. They both returned a lot of positive hits on your kills.
Subtle Hacking
Validation Dataset Results
True Positive %: 321 / 464 (69.181%) of hacking kills were correctly identified as hacking
True Negative %: 12118 / 12162 (99.638%) of legit kills were correctly identified as legit
Unlabeled Dataset Results
45490 kills were labeled as legitimate
259 kills were labeled as hacking
0.5661325930621435% of all kills were labeled as hacking
Here's the number of detections for everyone in the unlabeled dataset that set off more than one detection. The format is the number of detections, the number of total kills, and the probability of that many detections happening across that many kills.
Statistically, the more kills you get, the higher the chances of a false positive. A low percentage implies a high confidence by the detector.
Someone 2 / 274 ~= 26.12481417110864127276902648%
Someone 2 / 421 ~= 45.04969553050558835682131520%
Someone 8 / 842 ~= 1.281158394365624780613896410%
Someone 6 / 814 ~= 7.827600371123071985139007220%
Someone 2 / 156 ~= 11.01722262770741636476110179%
Someone 2 / 504 ~= 54.48471257135807943971915430%
Someone 6 / 956 ~= 13.68758722408053363204875611%
Someone 7 / 3320 ~= 95.49193693366735784711596619%
Someone 2 / 38 ~= 0.8448830523287402531368268500%
Someone 2 / 107 ~= 5.789701955609884059409463370%
Someone 2 / 316 ~= 31.71109866838154822690851285%
Someone 2 / 715 ~= 73.09075780886082901474743276%
Someone 2 / 517 ~= 55.85639375750722737574379914%
Someone 2 / 1078 ~= 90.14149613577330637101381496%
Someone 2 / 9 ~= 0.0463853427893176095225232300%
Someone 4 / 355 ~= 4.131875839842128099642541360%
Someone 2 / 60 ~= 2.018631845414636415879097280%
Someone 2 / 117 ~= 6.768203886574358553004438130%
Someone 2 / 151 ~= 10.43957375753245954252819570%
Someone 3 / 443 ~= 21.74013715350598276528435169%
Acrasia 18 / 553 ~= 5.034779104174968600E-10%
Someone 2 / 198 ~= 16.14708744387407504144498877%
Someone 2 / 65 ~= 2.344212678020780373351517170%
Someone 2 / 84 ~= 3.756341488970316793673624620%
Someone 4 / 929 ~= 43.36769708690454545593476023%
Someone 3 / 212 ~= 4.255019235187668383131969480%
Someone 2 / 57 ~= 1.833203861972019497911954060%
Someone 3 / 5 ~= 0.0000471807130383102081217800%
Someone 9 / 34 ~= 5.160789033429100E-13%
Someone 3 / 1072 ~= 74.44103435451574644417436005%
Someone 2 / 49 ~= 1.376734664920089606006894170%
Someone 2 / 21 ~= 0.2628772444731194773937163400%
Someone 2 / 51 ~= 1.485541873326783114695531310%
Someone 2 / 39 ~= 0.8884267988692136490431041000%
Someone 4 / 28 ~= 0.0003279982443945032856773900%
Someone 2 / 6 ~= 0.0194676195610228971736262400%
Someone 2 / 313 ~= 31.31437137901927005135792778%
Someone 2 / 41 ~= 0.9784589311363078528991128400%
Someone 2 / 127 ~= 7.796510683219792898043820830%
Someone 4 / 66 ~= 0.0103466925731434828288346300%
Someone 3 / 23 ~= 0.0079573952411673512518047700%
Someone 3 / 86 ~= 0.3880735773011146906912752100%
Blatant Hacking
Validation Dataset Results
True Positive %: 1049 / 1168 (89.812%%) of hacking kills were correctly identified as hacking
True Negative %: 10586 / 10594 (99.924%%) of legit kills were correctly identified as legit
Unlabeled Dataset Results
45603 kills were labeled as legitimate
146 kills were labeled as hacking
0.31913265863734724% of all kills were labeled as hacking
Someone 2 / 814 ~= 3.900018653190190268999015810%
Someone 4 / 504 ~= 0.0047604864164614274615736300%
Someone 6 / 3320 ~= 0.19187327248470814 46531351800%
Someone 2 / 461 ~= 1.364140124089468477280673500%
Someone 2 / 1078 ~= 6.412469747458562277630455230%
Someone 4 / 60 ~= 9.996253997415272111600E-7%
Acrasia 8 / 553 ~= 7.4574457854969364100E-9%
Someone 12 / 34 ~= -1.00E-25%
Someone 2 / 14 ~= 0.0013100515737227744564381700%
Someone 2 / 22 ~= 0.0033187850989585743181414300%
Someone 2 / 5 ~= 0.0001442902872738706091616500%
Someone 2 / 3 ~= 0.0000433090256000000047681100%
Someone 2 / 21 ~= 0.0030178414222042359486123300%
Someone 2 / 10 ~= 0.0006484843848373056438465700%
Someone 2 / 28 ~= 0.0054224956341558663494228600%
Someone 2 / 4 ~= 0.0000865961086554080095326200%
Someone 3 / 8 ~= 3.068455875858720408500E-7%
Someone 2 / 13 ~= 0.0011231857901402758806470900%
Someone 2 / 3 ~= 0.0000433090256000000047681100%
Someone 2 / 5 ~= 0.0001442902872738706091616500%
Someone 6 / 19 ~= 8.1347504864900E-15%
Someone 3 / 5 ~= 5.48407277141100890400E-8%
Someone 7 / 21 ~= 1.32424484300E-17%
Someone 2 / 16 ~= 0.0017266657070106395765361600%
As such, based on a combination of anti-cheat data and the videos provided by
@helix, I believe you are guilty beyond a reasonable doubt of hacking.
Appeal: Denied, and extended to forums + Discord since hacking bans are community-wide bans. Also, not voided.
Thread: Locked
If you have questions, you can reach out to me via Discord.
~Pierogi