PureSkill.gg CS:GO Data Science Development Kit
Counter-Strike helpers for CSDS match data (a manifest plus one pandas DataFrame per channel). scrub_csds_pii anonymizes a match before it leaves the platform, pop_overtime removes overtime rounds from a channel, and add_player_vector_derived_columns computes player_vector’s velocities and movement angles. It works on data you have already loaded, usually with pureskillgg-dsdk.
Installation
Requires Python 3.11 or later.
$ uv add pureskillgg-csgo-dsdk
It does not install pureskillgg-dsdk. The example below loads the match with it, so add it too:
$ uv add pureskillgg-dsdk
Usage
Scrub a match, then pop overtime:
from pureskillgg_dsdk import DsReaderS3, GameDsLoader
from pureskillgg_csgo_dsdk import (
csds_pii_channel_instructions,
pop_overtime,
scrub_csds_pii,
)
loader = GameDsLoader(
reader=DsReaderS3(bucket="my-csds-bucket", manifest_key="path/to/match/csds")
)
# Load only the PII channels this CSDS has, then scrub them.
data = loader.get_channels(csds_pii_channel_instructions(loader.manifest))
manifest = scrub_csds_pii(loader.manifest, data) # data is scrubbed in place
# CS2 regulation is 24 rounds (MR12). The default, 30, is CS:GO MR15.
deaths = loader.get_channel({"channel": "player_death"})
overtime = pop_overtime(deaths, max_rounds_csgo=24) # deaths keeps rounds 1-24scrub_csds_pii returns the new manifest. It replaces the job id with the anonymous id, redacts names, clan tags, chat, share codes and demo ids, maps each Steam id to a letter, zeroes pings, caps inflated win and commend counts, and marks every changed column in the manifest. docs/scrub-csds-pii.md lists every field.
pop_overtime returns the removed rows with their original index labels. Rows with a missing round stay. A channel without a round column raises MissingColumns.
Derived player_vector columns
Ten player_vector columns are computed from the columns read from the demo: second, x_vel, y_vel, z_vel, speed_2d, movement_angle, movement_angle_diff, phi_vel, theta_vel and ang_vel. Compute them after loading:
from pureskillgg_csgo_dsdk import (
add_player_vector_derived_columns,
player_vector_source_columns,
)
# Everything, on a whole player_vector.
player_vector = loader.get_channel({"channel": "player_vector"})
add_player_vector_derived_columns(player_vector) # added in place
# Or load only what two of them need.
wanted = ["speed_2d", "z_vel"]
player_vector = loader.get_channel(
{"channel": "player_vector", "columns": player_vector_source_columns(wanted)}
)
add_player_vector_derived_columns(player_vector, columns=wanted)The values are the ones csgo-ppp writes:
Velocities are differenced per player per round, so each player’s first sample in a round reads 0. A sample where any axis moved faster than 3,500 units a second (the engine’s cap) is a teleport and reads 0 too.
movement_angle is the direction of movement, 0 to 360. movement_angle_diff is where the player looks minus where they move, -180 to 180, and missing when they stand still.
second is tick over the tick rate: 64 in CS2. Pass tick_rate for a CS:GO match; it is in the match’s header.
A derived column the frame already has is replaced, never read, so a file that stores them comes out the same as one that doesn’t. Rows must be in tick order within each player and round, as csgo-ppp writes them. For a frame holding more than one match, such as a tome, pass group_by=["match_key", "player_id", "round"]. A frame that lacks a column they are computed from raises MissingColumns.
Exports
Export |
Used by |
|---|---|
scrub_csds_pii, SCRUB_CSDS_PII_CHANNEL_INSTRUCTIONS |
csgo-ppp’s scrubber |
csds_pii_channel_instructions |
no consumer yet; prefer it to the constant, which names channels an older CSDS may not have |
pop_overtime |
csgo-coach’s economy course |
add_player_vector_derived_columns, player_vector_source_columns, PLAYER_VECTOR_DERIVED_COLUMNS |
readers of player_vector that need its velocities, movement angles or second |
pureskillgg_csgo_dsdk.player_vector’s calc_* functions |
csgo-ppp’s converter, which computes the same columns step by step |
MissingColumns, UnsupportedChannelStructure |
raised when a DataFrame lacks a required column |
Development
You need Python 3, uv and Git LFS (the test fixtures are in LFS).
$ git clone https://github.com/pureskillgg/csgo-dsdk.git $ cd csgo-dsdk $ git lfs install $ git lfs pull $ uv sync
The tasks are in the Makefile: make lint, make test, make watch (tests on every change) and make format.
Publishing
Set the new version with uv version <version> (or uv version --bump patch), then run make version. It commits pyproject.toml and uv.lock and pushes a signed v* tag, which triggers the publish workflow. Or run the version workflow by hand with a version number or a bump (patch, minor, major); it does both steps.
Publishing needs the PYPI_API_TOKEN repository secret. The version and format workflows also need GH_USER, GH_TOKEN, GIT_USER_NAME, GIT_USER_EMAIL, GPG_PRIVATE_KEY and GPG_PASSPHRASE.
License
MIT. See LICENSE.txt.
1 maintainer
Metadata
1 maintainer
Release files for pureskillgg-csgo-dsdk 3.3.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| pureskillgg_csgo_dsdk-3.3.0.tar.gz | 14.4 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| pureskillgg_csgo_dsdk-3.3.0-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 34.5 kB
Release files / pureskillgg_csgo_dsdk-3.3.0.tar.gz
| Download URL | pureskillgg_csgo_dsdk-3.3.0.tar.gz |
|---|---|
| Size | 14.4 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
9123c7b0bb2ae07e9d6a99f6b7c8bafa2303f41445a1c4d730f1fc339d09b455
|
|
BLAKE2b-256 checksum How to use checksums |
09dcc8717cd05f65f345b4fa7e7f3d4c4d411704237589e2d085b37251496a15
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.11.21 {"installer":{"name":"uv","version":"0.11.21","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}
|
Release files / pureskillgg_csgo_dsdk-3.3.0-py3-none-any.whl
| Download URL | pureskillgg_csgo_dsdk-3.3.0-py3-none-any.whl |
|---|---|
| Size | 20.1 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
57d85fdb03f511b123d617e537c066eea9659dfbd85926264fa57cb7be62ec8b
|
|
BLAKE2b-256 checksum How to use checksums |
85424c39b1b2daa3952d82b4920f641cdf19f052149022bdba7146f9c5aca86d
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.11.21 {"installer":{"name":"uv","version":"0.11.21","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}
|
