TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

What I learned reviewing 35k reviews from Steam

13 pointsby AdamJMarshabout 3 years ago
Since I was a kid theorycrafting in WOW, I&#x27;ve been interested in games and data. Over COVID, I’ve had this itch I&#x27;ve wanted to scratch to find out what exactly gamers complain about the most in which games.<p>Let&#x27;s face it AAA game developers doubled down on their Wall Street corporate mentality. This has seen them grow increasingly out of touch with what gamers really want, as they see them as nothing more than cash cows to extract dollars out of.<p>And no place demonstrates that better than within the reviews of Steam.<p>I wanted to see if I could view &amp; analyse Steam reviews at scale to get different understanding of what people really care or complain about the most.<p>Out a dataset of 35000 scraped reviews gathered from a randomized selection out of a larger dataset of 380k, (provided thanks to Luthfi Mahendra) I gathered the following insights.<p>TL;DR<p>- PUBG is a love it or hate it game which people constantly dump on for being filled to the brim with hackers.<p>- GTA V has a captive audience of people who hate the game, hate hackers, can&#x27;t understand how to use mods as well as micro-transactions.<p>- Rocket League fans are super happy, loving the competitiveness and easy to pick up game mechanics. It&#x27;s constantly referred to as fun, enjoyable, addictive and entertaining.,<p>- Rust fans are gush like fanboys over the game. It exceeds their expectations without exploiting them.<p>- In general, Steam reviewers hate hate hate hackers, bans, microtransactions, glitches, crashes and lag.<p>PUBG<p>- This game had the highest number of reviews from this dataset (Roughly 10% of reviews were for PUBG).<p>- There were 1916 analyzed not recommended reviews (from clusters of 204+159+153+147+114+136+134+133+132+125+123+121+121+114 recommended reviews)<p>- There were 1064 not recommended reviews (from clusters of to 33+132+100+58+142+65+68+55+41+48+40+105+57+120 not recommended reviews)<p>- This gives it a ratio of roughly 1.9 not recommend reviews for every 1.06 recommended reviews.<p>- Most complaints generally mentioned “hackers”, “hacking” &amp; “cheaters”<p>- Less quantities of complaints revolved around “greed”, “buggy”, “unplayable”, “microtransactions”, “banned”, with references to other games and developers like “fortnite”, “Battlefield”, “Capcom” and “EA”<p>- A high number of negative reviews made reference to connectivity &amp; server issues with mentions of “lag” “latency” “laggy”<p>- Another segment of complaints revolve around “crashing”, “crashes”<p>- Further review clusters included food references such as the Winner Winner Chicken Dinner phrase, with mentions of potato aiming and in game states&#x2F;food like meat, pan, apple, hungry, food.

3 comments

magicalhippoabout 3 years ago
I don&#x27;t know much about PUBG, but I did play a lot of other online shooters. Based on those experiences it seems there&#x27;s a fairly large number of players who thinks that anyone substantially better than them are cheating.<p>They played with poor sound setup so didn&#x27;t realize I knew exactly where they were thanks to a good headset and could blast them in the face as they rounded that corner.<p>Or they didn&#x27;t know about drag&#x2F;flick shooting[1] so they thought the snap that landed the headshot was a cheat rather than a deliberate, physical mouse movement.<p>Or just they just significantly overestimated their own abilities, thinking anyone being that much better must be cheating.<p>[1]: <a href="https:&#x2F;&#x2F;themeta.com&#x2F;aiming-essentials-3-flick-shot-technique&#x2F;" rel="nofollow">https:&#x2F;&#x2F;themeta.com&#x2F;aiming-essentials-3-flick-shot-technique...</a>
评论 #30467761 未加载
评论 #30470180 未加载
muzaniabout 3 years ago
Have you spotted a better way to put ratings on a game based on this data?<p>I feel like there are more mediocre games with &quot;Overwhelmingly positive&quot; reviews than &quot;mostly positive&quot;. I barely play most games that I buy, so I don&#x27;t really want to waste time on meh games with high reviews. I&#x27;m looking for the type of games that I might love but everyone else hates.
AdamJMarshabout 3 years ago
GTA V<p>- Grand Theft Auto V had the 2nd highest amount of total reviews<p>- There 927 total negative reviews gathered (from clusters containing 193+173+164+130+124+31+113 negative reviews)<p>- There were 243 positive reviews (from clusters containing 54+43+27+22+15+58+24 positive reviews.)<p>- This gives it a rough ratio of 3.8 negative reviews for every 1 positive review. This was the largest amount in the dataset.<p>- A large number of complaints revolved around banning. Whether it was a general complaint on bans, unfair banning or the ban appeal.<p>- A large number of reviews were orientated around general mods references or specific mods. Words that came up were related around “mods” “modding” “openiv”<p>- Certain missions, vehicles, or purchases were mentioned such as “Royale” &amp; “Alpha”<p>- There was a less quantity of negative reviews regarding hackers.<p>- There were a less quantity of negative reviews relating to microtransactions with specific mentions of “rockstar”, “greedy”, “gtx”, “currency”, “pay”, “paying” “overpriced”<p>Rocket League<p>- Rocket League had a total of 517 recommended reviews (in clusters of 49+351+117 ) to 10 not-recommended reviews (in clusters of 3+6+1).<p>- The largest cluster of reviews made specific references to “addictive” “addiction” “fun” as qualities of the gameplay.<p>- Many positive mentions of the game being “entertaining”, “fun” &amp; “enjoyable”<p>Rust<p>- Rust had a distribution of 569 recommended reviews (from clusters of 64+323+182)<p>- It had a not recommended reviews of 103 not recommend reviews (from clusters of 34+52+17)<p>- Many references were made mentioning “cancer” with regards to the game itself or the community.<p>- Other references included “gaming” “gameplay” “ammo” “community” “Gamers” “ultimate” “simulator”<p>- This game received many mentions of other games including “skyrim”, “fortnight”, “battlefield”, “minecraft”<p>- Specific items or gameplay mechanics were mentioned like “Death” “health” “dying” “war” “toxic” “survival” “pain” “fire” “build” “nake” “run” “kills” “looting” “gunplay” “spawns”<p>Overall positive recommended reviews<p>- 378 recommended reviews skewed towards co-operative video games.<p>- The most positive and popular recommendations (within two clusters of 351 &amp; 323 reviews) specifically referenced the video games gameplay.<p>- 129 positive reviews were made making general mentions of the video game as being fun to play (a large portion of these making specific reference to Rocket League)<p>The negative sentiment reviews gave much meatier insights though.<p>Overall negative sentiment reviews<p>- 204 negative reviews that were made in relation to bad game play or poor game play quality.<p>- In 193 reviews, users complained about being banned by Steam.<p>- 173 negative reviews were made with reference to “mods”. These reviews were related to mods causing bugs that made their game crash or become unstable.<p>- 159 users complained about experiencing lag and latency issues within multiplayer video games.<p>- 153 negative reviews blamed games for crashing and glitching that rendered them unplayable or disruptive.<p>- In two different clusters that contained 147 &amp; 142 each, negative reviews were posted, voicing complaints regarding hackers, which is common in online multiplayer games.<p>If I had my time again, I would...<p>- Like to spend more time mining for insights<p>- Work on updating or scraping my own dataset with a larger and more updated dataset.<p>- Track which game developers as well as video games had the highest amount of negative reviews, positive reviews and positive to negative review ratios.<p>- Begin to visualise each genre of video game, their review totals, and negative to positive review ratios.<p>- Map out repeatable criticisms or issues with larger marquee games, over a wider &amp; more up to date dataset. (i.e Specific mentions of Fortnite Skins being too expensive, or negative sentiment with reference to GTA V Shark Cards)<p>End Credits.<p>Source: <a href="https:&#x2F;&#x2F;www.kaggle.com&#x2F;luthfim&#x2F;steam-reviews-dataset" rel="nofollow">https:&#x2F;&#x2F;www.kaggle.com&#x2F;luthfim&#x2F;steam-reviews-dataset</a><p>Tool for Visualisation and Analysis <a href="https:&#x2F;&#x2F;relevance.ai" rel="nofollow">https:&#x2F;&#x2F;relevance.ai</a><p>For those that want to dive deeper or reproduce this experiment:<p>Broader Clustering results - <a href="https:&#x2F;&#x2F;cloud.relevance.ai&#x2F;dataset&#x2F;steam_reviews_35k_zeroshot&#x2F;deploy&#x2F;cluster&#x2F;517b37aee3ebc3920275&#x2F;bHcwVEgzOEJGd0pIb0s3SUZRYWw6RzIzU042cFpUU2loM0NUOUU5aFcwQQ&#x2F;QDUKH38BQZpB7BTDVb_v&#x2F;us-east-1&#x2F;?sort=sum_Not%20Recommended&amp;order=desc" rel="nofollow">https:&#x2F;&#x2F;cloud.relevance.ai&#x2F;dataset&#x2F;steam_reviews_35k_zerosho...</a><p>Game specific clustering results – <a href="https:&#x2F;&#x2F;cloud.relevance.ai&#x2F;dataset&#x2F;steam_reviews_35k_zeroshot&#x2F;deploy&#x2F;cluster&#x2F;517b37aee3ebc3920275&#x2F;dTEwOUwzOEJfU3VrT0lSOWpuUW86SFBqRS16WnVSSGU5UFM5aWwwdnJFdw&#x2F;r107L38B_SukOIR94XSh&#x2F;us-east-1&#x2F;" rel="nofollow">https:&#x2F;&#x2F;cloud.relevance.ai&#x2F;dataset&#x2F;steam_reviews_35k_zerosho...</a><p>For those that know Python, how to reproduce: <a href="https:&#x2F;&#x2F;drive.google.com&#x2F;file&#x2F;d&#x2F;11EAQN_xYhIBmjU0ItAs5hxvW0AlqVe92&#x2F;view?usp=sharing" rel="nofollow">https:&#x2F;&#x2F;drive.google.com&#x2F;file&#x2F;d&#x2F;11EAQN_xYhIBmjU0ItAs5hxvW0Al...</a><p>Cleaned dataset source – <a href="https:&#x2F;&#x2F;github.com&#x2F;RelevanceAI&#x2F;michelangiolo_experiments_repo&#x2F;tree&#x2F;main&#x2F;220119_automatic_tweet_scraper&#x2F;apps&#x2F;steam_reviews_35k_zeroshot" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;RelevanceAI&#x2F;michelangiolo_experiments_rep...</a>