1 of 1

Riot Games Team Fight Tactics: Meta Shift Analysis

Joseph Waldron, Ching-yu Huang (faculty advisor)

School of Computer Science and Technology, Kean University, waldrojo@kean.edu

Team Fight Tactics (TFT) is a popular online strategy game developed by Riot Games [1]. With constantly evolving gameplay and a highly competitive ranking system, success in TFT requires multiple skills to create an effective strategy. However, the dominant team compositions, known as the meta, change over time due to game patches that include balance changes. This research explores how data mining techniques can be applied to TFT games gathered over the past few months. By analyzing several trends in gameplay, this research will provide players with essential information on what kinds of changes can shift the meta. Overall, this research is significant as it aims to offer valuable insights into the ever-evolving game meta of TFT.

This dataset has numerous benefits and potential for many use cases including

comparing values and predicting meta-compositions based on correlations between items and units. This study opens a small window of what is possible with this dataset, indicating its potential for many use cases.

​

Future research opportunities:

  • Developing a python package and website for graphing comparisons.
  • Creating a machine-learning model to predict a board’s placements based on items, traits, and units.
  • Additionally, this dataset can be used to study the impact of game balance adjustments on the TFT meta and player adaptation.

Data Collection: Python script, Riot API[2], match information from 500 North American players, SQL database

Data Cleaning and Transformation: Jupyter[3], pandas[4], SQL queries to filter, clean, and transform data.

Data Analysis / Data Mining techniques:

Normalization: Average daily occurrences / daily games played

Clustering: grouping similar team compositions based on units played

Association: Identity what characters and items are picked together

Data Visualization: Using matplotlib[5] create charts and graphs to visualize findings.

Tools and Technologies: SQL(database), python(data collection, data analysis, machine learning), pandas(data cleaning), matplotlib(data visualization), Riot API (data source)

[1] Riot Games official website [Online]. Available: https://www.riotgames.com/en

[2] Riot Games API TFT docs [Online]. Available: https://developer.riotgames.com/docs/tft

[3] Jupyter notebook [Online]. Available: https://jupyter.org/

[4] Pandas python library [Online]. Available: https://pandas.pydata.org/

[5] Matplotlib [Online]. Available: https://matplotlib.org/

Introduction

Methods and Materials

Results

Conclusions / Future Work

References

Frequency graph(traits) Admin and Space Corps

Win rate % graph(traits)

Top 4 % graph(traits)

Admin Buffed

Space Corps nerfed

Exploring correlation with Chi-Squared

pValue = 0.16 DOF 3906

Exploring correlation with Chi-Squared

pValue = 0.24 DOF 4030

Example of Power Creep

Dataset

Headers: matchID, gamedatetime, gameversion, tftgametype, participantId(primary key), placements, augments, traits, units, items

  • Each game has 8 players (participantId)
  • Each unit has up to 3 items

Results cont..

Dataset cont..