Datasets:

Jamie-Packer
/

cs2-playstyle-dataset-2024

@@ -1,97 +1,70 @@
-# CS2 Playstyle & Roles Dataset (2024)
 [![DOI](https://zenodo.org/badge/1069609087.svg)](https://doi.org/10.5281/zenodo.17264782)
-**Single-file dataset** capturing player playstyle features and roles across professional Counter-Strike 2 matches (2024).
-It’s designed for quick exploration (clustering, similarity search, role analysis) and to be easy to load in Python/R.
-- **File:** `data/cs2_playstyle_roles_2024.csv`
-- **Rows:** players (unique by `steamid`)
-- **Key features:** opening attempts (OAP), trading behaviour (PODT/POKT), time-alive-per-death (TAPD), and positioning **team-relative ranking** metrics (ADNT/ADAT), plus per-side and overall role labels.
-> **License:** CC BY 4.0 (attribution required).
-> **Credit:** dataset engineered with the help of the [`awpy`](https://github.com/pnxenopoulos/awpy) library; role labels derived from Harry Richards’ [(@NER0cs)](https://x.com/NER0cs) [Positions Database](https://public.tableau.com/app/profile/harry.richards4213/viz/OLDPositionsDatabaseArchived/PositionsDatabaseNER0cs) (credited below). Special Thanks!
 ---
-## What’s inside
-- **Identifiers:** `steamid`, `player_name`, `team_clan_name`, `map_count`
-- **Playstyle features (per-side + overall):**
-  - `tapd_*` – Time Alive Per Death (seconds)
-  - `oap_*` – Opening Attempt % (0–100)
-  - `podt_*` – % of deaths that were traded (0–100)
-  - `pokt_*` – % of kills that were trades (0–100)
-- **Positioning (team-relative ranking, 0.2–1.0):**
-  - `adnt_rank_*` – relative rank for distance to nearest teammate
-  - `adat_rank_*` – relative rank for distance from team centre
-- **Roles:** `role_t`, `role_ct`, `role_overall`
-See the full column descriptions in **[`docs/DATA_DICTIONARY.md`](docs/DATA_DICTIONARY.md)**.
 ---
-## How the features were built (short version)
-- Demos were parsed with **awpy** to produce per-round/player events and pre-plant tick snapshots.
-- **TAPD**: average seconds alive at time of death.
-- **OAP**: opening attempt rate = (opening duels participated) / (team rounds) × 100.
-- **PODT/POKT**: traded-death% / trade-kill% within a 5-second window.
-- **Positioning (ADNT/ADAT)**: computed from 3D player positions every 3s, starting 10s after freeze end, **pre-plant only**, **alive-only**, and **only when a full 5v5 is alive** (to focus on structured setups).
-  - For each map and team, per-player map-level means are converted to **team-relative ranking scores** in (0.2–1.0], where 0.2 ≈ closest to teammates and 1.0 ≈ furthest.
-  - Ranks are then **round-weighted across maps** to produce `*_t`, `*_ct`, and `*_overall`.
-> Rationale for design: we target **initial round tendencies** (pre-plant, 5v5) and avoid map-size/veto bias by using team-relative ranks instead of absolute distances.
-## Data coverage & limitations
-- **Demo selection:** Demos were collected from professional Counter-Strike 2 matches played in 2024.
-  Only `.dem` files that were publicly available and successfully parsed by `awpy` were included.
-  Some matches may be absent if demos were missing, corrupted, or caused parse failures.
-- **Role coverage:** Not every player has a role label.
-  Role data come from [Harry Richards’ Positions Database](https://public.tableau.com/app/profile/harryrichards);
-  players or teams not covered by that source are left blank in the `role_t`, `role_ct`, and `role_overall` columns.
 ---
-## Load & use
-**Python (pandas)**
 ```python
 import pandas as pd
 df = pd.read_csv("data/cs2_playstyle_roles_2024.csv")
-```
-**R (readr)**
-```r
-library(readr)
-df <- read_csv("data/cs2_playstyle_roles_2024.csv")
 ```
 ---
-## Licensing & credits
-- **License:** Creative Commons **CC BY 4.0** (see `LICENSE`). You may use, share, and adapt with attribution.
-- **Please attribute:**
-  - Jamie Packer (this repo).
-  - [**awpy**](https://github.com/pnxenopoulos/awpy) (demo parsing).
-  - [**Harry Richards’ Positions Database**](https://public.tableau.com/app/profile/harry.richards4213/viz/OLDPositionsDatabaseArchived/PositionsDatabaseNER0cs) (source for role labels).
-Suggested attribution line:
-> “This work uses the CS2 Playstyle & Roles Dataset (2024), created by Jamie Packer, generated using awpy; role labels adapted from Harry Richards’ Positions Database. Licensed under CC BY 4.0.”
----
-## How to cite
-If you create a paper, post, or tool using this dataset, please cite the repository.
-Please also credit [**Harry Richards’ Positions Database**](https://public.tableau.com/app/profile/harry.richards4213/viz/OLDPositionsDatabaseArchived/PositionsDatabaseNER0cs) when roles are used.
-```
-Jamie Packer (2024). CS2 Playstyle & Roles Dataset (2024). GitHub repository.
-```

+---
+license: cc-by-4.0
+language:
+- en
+tags:
+- cs2
+- counter-strike
+- esports
+- gaming
+pretty_name: CS2 Playstyle Dataset (2024)
+size_categories:
+- n<1K
+---
+# CS2 Playstyle Dataset (2024)
 [![DOI](https://zenodo.org/badge/1069609087.svg)](https://doi.org/10.5281/zenodo.17264782)
+**Author:** Jamie Packer
+**License:** [CC BY 4.0](LICENSE)
+**Source:** [GitHub Repository](https://github.com/Jamie-Packer/cs2-playstyle-dataset-2024)
 ---
+## Overview
+This dataset captures **player playstyle features and role labels** from professional Counter-Strike 2 matches in 2024.
+It includes aggregated player-level metrics describing opening attempts, trading behaviour, time-alive-per-death,
+and **team-relative positioning features** (ADNT, ADAT) derived from parsed demo files using
+[`awpy`](https://github.com/pnxenopoulos/awpy).
+Designed for clustering, player profiling, and role-based analysis.
 ---
+## Contents
+| File | Description |
+|------|--------------|
+| `data/cs2_playstyle_roles_2024.csv` | Main dataset (player-level records). |
+| `docs/DATA_DICTIONARY.md` | Feature definitions and schema. |
+| `LICENSE` | Creative Commons Attribution 4.0 International. |
+---
+## Data coverage
+- **Demos:** Professional matches played in 2024, parsed with `awpy`.
+  Some demos are excluded if missing or failed to parse.
+- **Roles:** Adapted from [Harry Richards’ Positions Database](https://public.tableau.com/app/profile/harry.richards4213/viz/OLDPositionsDatabaseArchived/PositionsDatabaseNER0cs).
+  Players not covered in that source are left blank in `role_t`, `role_ct`, or `role_overall`.
 ---
+## Quick start
 ```python
 import pandas as pd
 df = pd.read_csv("data/cs2_playstyle_roles_2024.csv")
+df.head()
 ```
 ---
+## Citation
+If you use this dataset, please cite:
+> Jamie Packer (2024). *CS2 Playstyle Dataset (2024).* Zenodo. [DOI: 10.5281/zenodo.17264782](https://doi.org/10.5281/zenodo.17264782)
+Also credit [Harry Richards’ Positions Database](https://public.tableau.com/app/profile/harry.richards4213/viz/OLDPositionsDatabaseArchived/PositionsDatabaseNER0cs)
+when using the role labels.