Loading NullCS
INDEPENDENT RESEARCH / CS2BEHAVIORAL ANALYSIS SYSTEM
FIG. 01 / THE ENCOUNTER / CONCEPTUAL VIEW
Look closer.
Beyond
the score.
Find unusual player behavior in your CS2 replays.
Open a demo. Compare players. Know where to look.
SEE HOW IT WORKSRESEARCH SIGNALS. HUMAN JUDGMENT.
01 / OBSERVATIONFROM PLAY TO PATTERN
The moments
between moments.
A fast flick tells only part of the story. NullCS looks at how the crosshair moves, when a player fires, and what they could see before the shot.
RECORDED ENCOUNTERS / DE_INFERNO
ENCOUNTER 0164
Minimum aim error2.04 degrees
Real measurements from a test replay. Each line uses its own scale so you can see its shape; these lines are not player scores. View the recorded data
INPUT → RESPONSE → CONTEXT
Timing. Aim. Movement. Visibility.
No single measurement tells the whole story.
02 / THE ANALYSISFROM REPLAY TO REPORT.
01/ 06
.DEM / INPUT
A match becomes a record.
Choose a Counter-Strike 2 demo file: a saved replay containing the actions from a match.
Read all six steps
- A match becomes a record. Choose a Counter-Strike 2 demo file: a saved replay containing the actions from a match.
- Every action has context. Read where players moved, where they aimed, when they fired, and what they could see.
- Isolate the encounter. Look at short sequences around player encounters, including what happened before and after the action.
- Read behavior over time. A model trained on gameplay data looks for patterns in how aim, movement, and timing change during an encounter.
- Bring the signals together. Combine measurements from a player’s encounters to build a picture of their behavior across the match.
- Give the reviewer a starting point. See which players the model puts first for replay review. Use the scores and match context to decide what to examine.
03 / THE INSTRUMENT, IN YOUR HANDSWINDOWS DESKTOP / BETA
Your match.
Your machine.
Drop a replay into the Windows app. Analysis runs on your computer. Compare the players and save a report.
Download NullCSNULLCS / LOCAL ANALYSISWINDOWS x64
ACTUAL APPLICATION CAPTURE
Local validation demo / de_inferno
LOCAL PROCESSING
No demo uploads. No account.
READABLE OUTPUT
Ranked players. Plain-text reports.
04 / RESEARCH RECORDMEASURED. DOCUMENTED. OPEN TO SCRUTINY.
Evidence before
certainty.
NullCS is an ongoing research project. Testing checks both missed cases and legitimate players who receive high scores. The newest model is still being tested and is not yet in the download.
01 / Replays in the research dataset894
02 / Matches scored in internal testing860
03 / Player results in internal testing6,886
04 / Replay analyzed at a time1
RECORD / 001Tests and results
See what has been tested.
GitHub explains how the models work, how they are tested, and where they make mistakes. The results include the limitations.
05 / INTERPRETATIONTHE HUMAN STAYS IN THE LOOP
Measure the behavior.
Examine the evidence.
Some replays show clear cheating. Others need a closer look at player skill, timing, and what was visible. NullCS helps you choose where to look; it can miss cases or flag legitimate play. The app analyzes one match at a time. Player history is planned, not available yet.