CS2CD/Context_window_256

Counter Strike 2 Cheat Detection Context Windows - Length 256

Overview

The Counter Strike 2 Cheat Detection Context Windows - Length 256 (Context_window_256) is a dataset conprised of "context windows". This dataset was created using extracted data from the CS2CD dataset. For more information regarding the data see AntiCheatPT: A Transformer-Based Approach to Cheat Detection In Competitive Computer Games by Mille Mei Zhen Loo and Gert Lužkov 1/6/2025.

Dataset Structure

⚠️ Warning: Any warnings that apply to the CS2CD dataset at commit 44e5129654508b22802a050a45bcdbb44b103d87 also apply to this project.

In the context of this project a context window(CW) is a sequence of ticks containing gameplay information. The CW in this dataset are centered around kills. These CWs capture the data of two parties: The attacker(the player performing the kill) and the victim(the player being killed). Given that the attacker was a cheater, the CW would be marked as a "cheater kill". Kills not performed by a cheater were marked as not-cheater kills regardless of whether the victim was a cheater or not.

Counter-Strike 2 gameplay data is recorded in 64 ticks/second. With the CW length being 256 ticks, the CWs capture 4 seconds of gameplay.

Ticks before kill	Ticks after kill	Total Number of ticks
224 (3.5 S)	32 (0.5 S)	256 (4 S)

Tick structure

A single tick in the CW is composed of the following information:

#	Data point	Description
1	attacker_X	X coordinate of the attacking player
2	attacker_Y	Y coordinate of the attacking player
3	attacker_Z	Z coordinate of the attacking player
4	attacker_vel	Velocity of the attacking player
5	attacker_pitch	Pitch angle of the attacking player
6	attacker_yaw	Yaw angle of the attacking player
7	attacker_pitch_delta	Change in pitch angle of the attacking player since last tick
8	attacker_yaw_delta	Change in yaw angle of the attacking player since last tick
9	attacker_pitch_head_delta	Pitch angle distance from the victim's head
10	attacker_yaw_head_delta	Yaw angle distance from victim's head
11	attacker_flashed	Is the attacker currently flashed
12	attacker_shot	Did the attacker shoot their weapon on this specific tick
13	attacker_kill	Did the attacker kill the victim on this specific tick
14	is_kill_headshot	Was the kill a headshot
15	is_kill_through_smoke	Was the kill through a smoke
16	is_kill_wallbang	Was the kill through some wall or surface
17	attacker_midair	Did the kill happen while the attacker was midair
18	attacker_weapon_knife	Is the attacker holding a knife
19	attacker_weapon_auto_rifle	Is the attacker holding an automatic rifle
20	attacker_weapon_semi_rifle	Is the attacker holding a semi-automatic rifle
21	attacker_weapon_pistol	Is the attacker holding a pistol
22	attacker_weapon_grenade	Is the attacker holding a grenade
23	attacker_weapon_smg	Is the attacker holding a submachine gun
24	attacker_weapon_shotgun	Is the attacker holding a shotgun
25	victim_X	X coordinate of the victim player
26	victim_Y	Y coordinate of the victim player
27	victim_Z	Z coordinate of the victim player
28	victim_health	Health of the victim player
29	victim_noise	Did the victim player make noise
30	map_dust2	Is the played map de_dust2
31	map_mirage	Is the played map map_mirage
32	map_inferno	Is the played map de_inferno
33	map_train	Is the played map de_train
34	map_nuke	Is the played map de_nuke
35	map_ancient	Is the played map de_ancient
36	map_vertigo	Is the played map de_vertigo
37	map_anubis	Is the played map de_anubis
38	map_office	Is the played map cs_office
39	map_overpass	Is the played map de_overpass
40	map_basalt	Is the played map de_basalt
41	map_edin	Is the played map de_edin
42	map_italy	Is the played map cs_italy
43	map_thera	Is the played map de_thera
44	map_mills	Is the played map de_mills

In order to capture the weapon used by the attacker without increasing the size of the tick vector too much, a weapon grouping was used as can be seen in row 18-24. The weapon groups were made based on similarity in gameplay.

#	Weapon group	Weapons
1	attacker_weapon_knife	Zeus x27, knifet, knife, Bayonet, Bowie Knife, Butterfly Knife, Classic Knife, Falchion Knife, Flip Knife, Gut Knife, Huntsman Knife, Karambit, Kukri Knife, M9 Bayonet, Navaja Knife, Nomad Knife, Paracord Knife, Shadow Daggers, Skeleton Knife, Stiletto Knife, Survival Knife, Talon Knife, Ursus Knife
2	attacker_weapon_autorifle	AK-47, M4A1-S, Galil AR, SG 553, M4A4, AUG, FAMAS, M249, Negev
3	attacker_weapon_semirifle	G3SG1, SSG 08, AWP, SCAR-20
4	attacker_weapon_pistol	CZ75-Auto, Desert Eagle, Dual Berettas, Five-SeveN, Glock-18, P2000, P250, R8 Revolver, Tec-9, USP-S
5	attacker_weapon_grenade	C4 Explosive, Decoy Grenade, Flashbang, High Explosive Grenade, Incendiary Grenade, Molotov, Smoke Grenade
6	attacker_weapon_smg	MAC-10, MP5-SD, MP7, MP9, P90, PP-Bizon, UMP-45
7	attacker_weapon_shotgun	MAG-7, Nova, Sawed-Off, XM1014

Usage notes

• The dataset is formated in UTF-8 encoding.
• Researchers should cite this dataset appropriately in publications.

Application

• Cheat detection

Acknowledgements

A big heartfelt thanks to Paolo Burelli for supervising the project.