Recent posts

Visualizing The DotA 2 The International Metas

DotA, Dimensionality Reduction, The International
29 May 19 05:01 -0400

In this blog post, I will do very basic visualizations of various matches from previous The International tournaments of DotA 2. DotA 2 is a video game by Valve. In this game, two teams compete in a virtual arena over objectives until one wins. The International is a tournament held every year by Valve, and it is arguably the most important and largest tournament of the scene. We will employ dimensionality reduction techniques to make the visualizations possible.

continue reading...

Sequencing What Items Invoker Will Use

DotA, Neural Network, RNN, Invoker
07 Sep 17 19:55 -0400

In this blog post, I attempt to explore the use of recurrent neural networks (RNNS) in predicting what items Invoker players will purchase in standard DotA 2 matches. DotA 2 is a game managed by Valve (and lead developed by IceFrog, hallowed be his name) with team-based, player versus player, and role-playing elements. As with most role-role playing games, there exists a diegetic component of a character build where a player must choose how the character he or she plays is developed. One of these elements of construction includes what items a player will purchase with in-game rewards in order to strengthen their character, and I attempt to predict this particular element with RNNs. The problem was limited to one of the characters of DotA 2 in higher level matches. This blog post will not focus on particularly difficult techniques; the network is actually a simple LSTM cell, and the code runs on most computers within the past 5 years.

continue reading...

Using Neural Networks to Detect Bots

DotA, Neural Network
26 Jun 17 19:55 -0400

Hi, in this blog post, I hope to explore the performance of neural networks, particularly auto-encoders, in detecting anomalous DotA matches using the feature pipeline (if it can even be called such) I have published on my github. Essentially, the idea was to collect a bunch of matches from Patch 7.06c and then feed them into a neural network to detect weird matches.

continue reading...