NBA Finals Game 5: Unlocking Trends with Sentiment Detection

Going into game 5, the Raptors led the series 3-1 and were looking to capture their first NBA title in franchise history. Since it was a Do-or-Die match, there was a lot of action on Twitter for Game 5 of the NBA Championship. We recorded north of 1.2 million tweets including pre and post-game time.

Volume Distribution

We plotted the number of tweets made in the intervals of 60 seconds over time. Interestingly, the tweet volume was higher right before the game, after a slight dip, tweet volumes took off during the game.

 

Volume Distribution

Geographical Distribution

Some observations about geographical distribution:

  • The NBA is garnering attention all around the world and we can clearly see that by observing the origins of tweets.
  • People from at least 150 countries were tweeting about the game. Below is a geographical heat map of tweets made during the game.

 

Geographical Distribution

Hashtag Trends

Kevin Durant made a comeback after 9 games in a celebrated fashion and it’s reflected on Twitter. We can observe that Kevin Durant is the most talked about player even trumping Steph Curry.

 

Hashtag trends
Hashtag Activity

Sentiment Analysis

One interesting observation we made with the average sentiment of tweets over various sessions is that there were more sessions where tweets were negative.

 

Sentiment Analysis

 

Things get really interesting if we break down sentiment trends based on hashtags. For example #Raptors, the average sentiment on this hashtag was positive pre-game but the sentiment starts to shift slightly downwards during and end of the game.

 

Raptor Sentiment

 

On the contrary, if we look at #Warriors, we can observe a mixed spread of sentiment across tweets.

 

Warriors Sentiment

 

Unlocking Trends with Sentiment Detection

Textual data like tweets can be very subjective and due to the large of tweets, it can be very difficult to make any conclusions or gain insights. Using algorithms like we have done in this post as well as the entire blog series, we have been able to gain valuable insights. Working on such data at scale is now possible, thanks to Google BigQuery, Google Kubernetes Engine, and Google Dataflow.  Click here for the details on the methodology we have been using for the Twitter Sentiment Analysis.

Series Continues with Game 6 on Thursday

If you have anything you are interested in seeing for the Game 6 Twitter sentiment analysis on Thursday, drop us a line at info@SpringML.com or tweet us @springmlinc. We’d love to have you follow this blog series online and on Twitter.