Who won E3 2019 on Twitter?

No comments

The 2019 edition of the Electronic Entertainment Expo (E3) took place in Los Angeles from June 11 to 13. As most gamers know, E3 is one of the most prominent gaming events in the world. I’m pretty active on Twitter and every day I see E3 2019 on Twitter trending with hundreds of thousands of tweets. It was a time where the gaming community gets together to rave, celebrate and fervently discuss what’s upcoming.

E3 2019 marks the 25th edition of the premier video gaming trade event.

Developers, publishers and manufacturers have long relied on this medium to release never before seen teasers, instantly reaching out to thousands of enthusiastic gamers in an unparalleled hyped up fashion. However, it is notable that this is the first year where Sony chose not to participate.

e3-2019-expo

Turns out it doesn’t quite matter even if Sony doesn’t have a physical presence in E3 2019. They made their statement through coveted exclusives.

Just 2 weeks before E3 kicked off, the highly anticipated and bizarre Death Stranding trailer dropped exclusively on the PlayStation Twitch channel. During E3, a new Final Fantasy VII Remake trailer was revealed to explosive responses thanks to the showcase of beloved familiar characters.

My Twitter feed was bombarded with everything E3. Newly released trailers, unexpected collaborations (looking at Keanu Reeves), breakout industry personalities. Even games that made no appearances had people discussing about why didn’t it. Given the tremendous amount of buzz E3 is generating on Twitter, I wanted to conduct a sentiment analysis of the tweets to see who got most of the positive engagement.

twitter-api

To find out who won E3 2019 on Twitter, we need to do 2 things. One, mining the tweets, and two, analyzing them.

Enter Twitter APIs. Twitter provides an extensive collection of APIs for developers who are interested to communicate with the Twitter interface. There are some available for public use, though more premium APIs and functionalities typically will require some form of paid subscription.

For our mining purposes, we will be using the Search API. This API searches against a sample of recent Tweets published in the past 7 days. Twitter explains that the API focuses on relevance rather than completeness, so some tweets are expected to be missing. It’s not an issue for us considering we’re just gathering a sense of the community sentiment.

e3-2019-ghostwire-tokyo-ikumi-nakamura
Ikumi Nakamura got most of the attention on E3 Day 1 with her enthusiasm and down-to-earth presentation of GhostWire: Tokyo (Photo: Risa Media)

Sentiment analysis, a subset of Natural Language Processing (NLP), is a deep topic in its own right and we are just about to scratch its surface.

It focuses on identifying, extracting and studying the opinions of information. I’ll be using the VADER sentiment analysis tool for the analysis. VADER (Valence Aware Dictionary and sEntiment Reasoner) is a lexicon and rule-based Python library that is specifically attuned to sentiments expressed in social media. The research paper that comes with it is a good read and I recommend you to go through it if you’re interested. More of that in the link above.

import pandas as pd
import tweepy
import time
import sys
from collections import OrderedDict
from pandas.io.json import json_normalize
from vaderSentiment.vaderSentiment import SentimentIntensityAnalyzer

My Github post covers the steps to install and import the libraries, authenticate and interface with Twitter, data mining process and various data manipulation required.

To start off with, we will be importing the above libraries required. The rest of the code can be found in my Github.

We will crawl for tweets using specific keywords pertaining to notable announced projects. The API is rate-limited, meaning we can only mine for a finite amount of tweets each time. Once mined, we will pass these tweets into the VADER sentiment tool. The tool will qualify the tweets and determine whether it is of a positive, neutral or negative sentiment. We will then analyze the results.

Based on my Twitter feed during E3, I chose a couple of projects that stood top of my mind for this experiment.

1. Elden Ring by Bandai Namco
2. GhostWire: Tokyo by Bethesda Softworks
3. Cyberpunk 2077 by CD Projekt Red
4. Star Wars Jedi: Fallen Order by Electronic Arts
5. Pokémon Sword and Shield by Nintendo
6. Final Fantasy VII Remake by Square Enix
7. Watch Dogs: Legion by Ubisoft
8. Halo Infinite by Xbox Game Studios

Firstly, I ran a non geo-specific search using the project titles as keywords. I kept the limit of mined tweets to be 500 as I did notice any significant differences in the sentiment results beyond this amount. Then, I’ve sorted the titles with the highest % of positive tweets first:

Title% positive% negative% neutral
GhostWire: Tokyo47%24%29%
Elden Ring44%23%33%
Cyberpunk 207740%30%29%
Final Fantasy VII Remake40%24%36%
Pokemon Sword and Shield40%35%25%
Halo Infinite39%30%31%
Watch Dogs: Legion38%21%41%
Star Wars Jedi: Fallen Order17%75%8%

To visualize the information better, I’ve constructed a stacked bar chart below. I’ve also plotted the sums of positive and neutral tweets as a green line and sorted the chart in descending order of this sum.

Initial findings seem to indicate that Watch Dogs: Legion leads the pack in terms of both positive and neutral sentiments.

It is notable however that neutral tweets form a significant proportion of this lead. Also, GhostWire: Tokyo deserves a special mention for having the highest number of positive tweets, possibly fueled by the popularity of Ikumi Nakumura.

From the data, it also seems obvious that Star Wars Jedi: Fallen Order is somewhat of an anomaly. I sorted and pulled the negative tweets to investigate further.

DOOM ETERNAL, Watch Dogs Legion, Gears of War 5 and Star Wars: Jedi Fallen Order are the games on my watch
I’ve been seeing a lot of unfair and undeserved criticism of Star Wars Jedi: Fallen Order over the last few days
Star Wars Jedi: Fallen Order – I just cannot get excited for EA Star Wars. My kids were not in school at all the
Star Wars Jedi: Fallen Order is not the game I was expecting, but it fits Star Wars so damn well. More tomorrow.

The simple experiment above highlights several challenges in text analysis particularly in understanding sentiments.

Almost immediately, anyone should be able to tell that the negatively classified tweets does not seem to be that negative. My hunch is that the name ‘Fallen Order‘ contains the word ‘fallen’ that’s usually perceived negatively. Therefore, tweets that should have been other classified as neutral have been unfairly penalized.

While it provides a lighthearted implication to product naming, we cannot deny the repercussions analysts would face if they wish to conduct sentiment analysis for a product name that has a strong feeling or emotion attached to it.

Social media sentiments is definitely an important source of information for brands to assess their audience reactions. I’m glad that this has cut through multiple industries, more so for gaming where the community is very active and vocal online. Hopefully in the near future, we will be able to see more accessible sentiment analysis tools to learn more about what gamers want and make greater games.

Featured photo by E3expo

Ever wondered how much money esports make? Read this post to learn more.

What was YOUR favourite moment of E3 2019? Comment below to share your thoughts.

Don't be a stranger, leave a comment.