How many Steam games go unplayed?
In the public Steam behavior dataset used here, 58,327 of 129,511 purchased user-game rows had no matching play-hour row. That works out to about 45% of purchase rows with zero recorded playtime.
Backlog Field Notes
Gaming backlogs
by the numbers.
Source · Public Steam behavior dataset (Kaggle / CaseRecommender mirror)
These are third-party aggregate findings, not Video Game Backlog user stats. The goal is to show how backlog behavior looks in public Steam data while our own first-party sample is still young.
The unplayed pile
Almost half the purchase rows never turn into play.
n = 129,511 purchases · Kaggle Steam behavior
Across the public behavior sample, 58,327 purchased user-game pairs had no matching play-hour row.
45%of every 100 games owned
Played at least onceNever launched
Library size
The bigger the library, the wider the played gap.
n = 12,393 users · Kaggle Steam behavior
Small libraries in this sample are often one-game accounts, but the gap becomes visible once libraries pass 10 owned games.
Median owned vs played games by library size
10-24 owned games
Owned
14
Played
9
25-49 owned games
Owned
33
Played
21
50-99 owned games
Owned
68
Played
41
100+ owned games
Owned
155
Played
83
Most abandoned
Bundle-era games dominate the never-launched list.
n = 5,155 games · Kaggle Steam behavior
Among games with at least 500 owners in the sample, old Valve pack-ins and free staples carry the highest unplayed rates.
Share of owners who never launched it
01
Ricochet96%
02
Deathmatch Classic94%
03
Counter-Strike CZ Deleted Scenes86%
04
Day of Defeat82%
05
Half-Life 2: Lost Coast82%
Where the hours go
One favorite game eats most of the actual playtime.
n = 4,791 multi-game players · Kaggle Steam behavior
For users with at least two played games, the top played title accounts for roughly two thirds of recorded hours on average.
Share of total hours played
65%
23%
Top game per library65%
Games ranked #2-#323%
Games ranked #4-#109%
Everything else played3%
The one-session shelf
A launch does not always mean a game escaped the backlog.
n = 70,489 play rows · Kaggle Steam behavior
More than a third of played rows still stop at two hours or less, while a much smaller slice crosses 100 hours.
Recorded play rows by total hours
25%
11%
56%
One hour or less25%
One to two hours11%
Two to 99 hours56%
100+ hours8%
Playtime skew
The average game is dragged upward by a few obsessions.
n = 70,489 play rows · Kaggle Steam behavior
The median played row is only 4.5 hours, but the average is 48.9 hours because the top time sinks are so large.
Hours recorded per played user-game row
Median played row4.5h0h
Average played row48.9h
90th percentile76h80h
Actually launched
Some popular games cut straight through backlog gravity.
n = 5,155 games · Kaggle Steam behavior
Among games with at least 500 owners in the sample, live-service staples and enduring classics have the highest owner-to-launch rates.
Share of owners with recorded playtime
01
Dota 2100%
02
Team Fortress 2100%
03
Counter-Strike: Global Offensive98%
04
The Elder Scrolls V: Skyrim94%
05
Sid Meier's Civilization V93%
Launch rate
Large libraries launch a smaller share of what they own.
n = 12,393 users · Kaggle Steam behavior
Users with 100+ owned games had a median 155 owned games and 83 played games; their average launched share was under 50%.
Average share of owned games with recorded play
1-9 owned games76%
10-24 owned games56%
25-49 owned games60%
50-99 owned games59%
100+ owned games49%
Sample shape
Public behavior data is lopsided before analysis even starts.
n = 12,393 users · Kaggle Steam behavior
The dataset has many tiny libraries, which is why the page separates broad sample counts from larger-library behavior.
User-library shape in the source sample
One-game libraries46%
Two to nine owned games36%
Ten or more owned games18%
8%
No recorded play rows
Some libraries have purchases but no matching play behavior at all.
Time sinks
The hours leaderboard is brutally concentrated.
n = 70,489 play rows · Kaggle Steam behavior
Dota 2 alone accounts for nearly one million recorded hours in this sample, more than the next six games combined.
Total recorded hours in sample
01
Dota 2981.7k h
02
Counter-Strike: Global Offensive322.8k h
03
Team Fortress 2173.7k h
04
Counter-Strike134.3k h
05
Sid Meier's Civilization V99.8k h
Method
Numbers below come from analysis of a public Steam behavior dataset originally published on Kaggle and mirrored by CaseRecommender. It contains purchase and play-hour rows from public Steam data. It is not a Steam-wide sample, it is historically skewed toward visible public libraries, and old bundle/default-library behavior shows up strongly in the results. Treat these as directional field notes, not platform-wide census figures.
Sample
12,393 users · 5,155 games · 200k behavior rows
Unplayed
Purchased, zero recorded playtime
Playtime
Recorded hours from public Steam behavior rows
FAQ
Backlog stats, plainly.
What does unplayed mean in these Steam backlog stats?
Unplayed means the dataset has a purchase row for a user and game, but no recorded play-hour row for that same user-game pair. It is a zero-playtime measurement, not a judgment about whether the game is good or worth playing.
Are these Video Game Backlog user stats?
No. These figures come from a public third-party Steam behavior dataset originally published on Kaggle and mirrored by CaseRecommender. They are directional field notes, not a platform-wide census and not first-party app usage data.
Why do large Steam libraries have lower launch rates?
Large libraries are more likely to include bundles, old sale purchases, free staples, and games bought speculatively. In this sample, users with 100 or more owned games launched a smaller share of what they owned than users with smaller libraries.
Your library, same view
See where your own hours go.
Curious how your library compares? The free Steam backlog analyzer shows your unplayed games, hours as work weeks, and a projected backlog clear date — no account needed. Once you have a public profile, the library comparison finds shared games between two players.