Assignment_6_Steam_Scraping

Steam Game Reviews Introduction

The purpose of this document is to understand the difference in reviews for multiplayer games versus single player games. In order to understand this difference, I will be using the popular gaming website called Steam. Steam is a website that houses games. It is where players buy games, play them, and talk within the community. The reason this is interesting to me is because I have spent around 100 hours of my life playing both single and multiplayer games. I am curious to see if my personal opinion will be seen in the majority of these reviews or if other ideas are more prevalent.I determined which games to focus on by looking at the top five multiplayer games under the “what’s being played” tab in the Steam store and the top five singleplayer games under that same tab. The games that I will be focusing on are Counter Strike: Global Offensive, Dota 2, APEX Legends, PlayerUnknown’s Battlegrounds, Naraka:Bladepoint, Lost Ark, Elden Ring, Grand Theft Auto 5, Vampire Survivors, and FIFA 22.

Below is a table explaining the data I will be using in my analysis. Each row is a variable that I will be using to complete the analysis. I got this data by scraping the community page on Steam for player’s reviews of the aforementioned games.

Variable	Description
review_game	Game that the player reviewed
reviewer_name	Steam Username of the reviewer
review_content	Text that the review contained
review_recommendation	Whether or not the reviewer recommends the game
helpful	Number of people who found the review helpful
reviewer_hoursplayed	Number of hours the reviewer has played the game they’re reviewing
reviewer_gamesowned	Number of games the reviewer owns
single_or_multi	Whether the game is single or multiplayer
words_in_review	Number of words written in review
level_of_detail	Level of detail of review (based on percentiles)
reviewdate	Date that the review was posted

Trends and Analysis

In order to understand how each type of game is reviewed, we will first be looking at whether or not that game was recommended as well as the amount of people that found the review helpful. Our first visualization will look at how many times each type of game was recommended or not.

Based on the above graph, we are able to see that there is a similar amount of reviews that recommend and don’t recommend the multiplayer games. We can also notice that there are more reviews that recommend a single player game rather than a multiplayer game. What this tells us is that single player game players are more likely to recommend the games that they play versus multiplayer game players. This helps point us in the right direction towards answering the question of whether or not single player games and multiplayer games differ in their reviews.

Something else interesting about these games is the number of words in the reviews. There are a number of reviews in here of people trying to be funny or just speaking their mind (ie, Onion’s review of Elden Ring). So, we will be looking at just how detailed these reviews are for each type of game.

From this visualization we can see that both types of games have reviews that are good sized (in the 50th percentile). We can also notice that our similarities end there, as multiplayer games have more short reviews while singleplayer games have more detailed reviews. This is to be expected, as someone is more likely to have more to say about a singleplayer game that gives them different experiences versus a multiplayer game that is the same thing over and over.

When understanding more about reviews, we need to understand who is writing reviews. The distributions shown below are going to show us which type of people are reviewing which game. This will help us understand which type of game engages its players for longer and also when players are reviewing these games.

This histogram shows us that there is a much more even distribution of hours played for people that are reviewing the multiplayer games. We can also note that there seems to be more seasoned players that are reviewing single player games. This could be due to possible frustration from players using a multiplayer game. Something else of note is that, over time played, people are more likely to be recommending a multiplayer game. This could be because there are people that get frustrated early in a multiplayer game, and they come to the review the game while frustrated. However, with more experience, they give recommending and less biased reviews.

Hours alone are not a determinant of how experienced a player is. You have to know what a bad game is before you can go online and slander games. That is why another important determinant is the amount of games that a reviewer has. From this, we will be able to see which game is getting reviews from more experienced players.

This shows us that people who review the singleplayer games typically own more games than the ones who review multiplayer games. This makes sense, as you finish one singleplayer game then you go onto the next. It is also interesting to see that singeplayer reviewers have more games because they have a more diverse experience of games. This could mean that their reviews are less biased than someone in the multiplayer realm who has been playing the same game for years.

Something else interesting about the reviews that we can look at is whether or not the reviews were found helpful. This will give us a good idea of whether or not the masses agree with if the game should or should not be recommended. First we will look at the distribution of users who found reviews helpful, next will be the number of people who found a review helpful by the type of game and then by the type of review recommendation.

This first histogram shows us that there is a right skewed distribution of the amount of people who were helped in reviews. With the most being 20 reviews having helped five people. This is interesting because I did not expect this result to come out this way. I fully expected this to be a normal distribution with around 40-50 people getting help from these reviews. Something interesting to note about the next visualizations is that there are many more people who find the reviews helpful when they recommend a game versus when a review does not. I expected this to be the case because I think about myself in this situation. If I didn’t like a game, I would just get rid of it and not look at it again. But if I had enjoyed a game I would be more likely to go back and mark the reviews I used as helpful because I agree with them. Lastly, there is an overwhelming amount of helpful reviews in the singleplayer genre versus the multiplayer genre. This could be due to the fact that most singleplayer games are less stressful than multiplayer games.

Overall, it would appear that singeplayer games have different reviews than multiplayer games. Singleplayer games have more helpful reviews than multiplayer games, and those reviewing the singleplayer games tend to be more experienced in terms of the amount of games that they own. Singleplayer games also tend to have more detailed reviews. The only place that multiplayer games have better reviews than singleplayer games is in terms of hours played. Considering there is less hours to play in games that are singleplayer, this is to be expected.

The statistical method that I would use in the future to validate my findings would be a hypothesis test to determine if the reviews for the singleplayer games are truly different than the ones for the multiplayer games. In my test, my null hypothesis would be that the singleplayer games get the same reviews as multiplayer games. My alternative hypothesis would be that there is a difference between single and multiplayer game reviews. Once I do this, I would run a z-test to determine whether or not there are differences because we have more than 30 samples. I would expect the outcome to be that the p-value is less than the significance level, meaning that there is a difference. This would supplement my findings because I would be able to tell whether or not I was right in saying that the reviews are different.