How many players reach the ending? Here is what Steam counts.

Steam publishes the unlock rate of every achievement in every game, and almost nobody reads it. Find the achievement that marks the end of a game and you know what share of its players, as Valve counts them, ever got there.

Finishing is the exception

The best public measurement of this is Bailey and Miyata, 2019. They took a random sample of 725 Steam games and identified the completion achievement in each one by hand, which matters more than the sample size does. Automated detection gets this wrong constantly, and this guide used to carry numbers that proved it.

Their distribution:

• Median: 9.7 percent
• 75th percentile: 20.5 percent
• 90th percentile: 33.1 percent
• 95th percentile: 41.1 percent
• Highest in the sample: 56.1 percent

So in the typical game about one in ten of the players counted reaches the end, and a 20 percent finish rate beats three quarters of the sample.

Sony's own trophy system says the same thing in different words. PlayStation labels an unlock rate between 15.1 and 50 percent Rare, and anything above 50.1 percent Common. A fifth of players reaching the end is, by the platform holder's own taxonomy, a rare event.

A quarter of the counted players never start

There is a floor under all of this that explains most of it. Take the single most-unlocked achievement in a game, usually something you trip over in the first minute. Across 286 games Bailey measured it at a median of 74.5 percent. An independent scan of 886 games in 2026 put it at 73.8 percent.

For the median game, about a quarter of the counted population never unlocks the easiest achievement in it. They are counted and never really start. Every finish rate you see is measured against that same denominator, so it is carrying all those people too.

Valve does not publish who is counted. Its documentation says only "the percentage of Steam players of the game" and never defines a player. That is why this site states the number and its source, the share of players Valve counts at the ending achievement, and never turns it into a share of buyers or owners.

What high actually looks like

Some real ones, read from Valve's data:

• Cuphead, "Good Ending": 17.5 percent
• Dishonored, "Complete the game": 40.2 percent
• The Stanley Parable, "Beat the game": 79.4 percent

The last one is the exception that shows the rule. Values above 76 percent exist, and on October 31, 2026 they were about 3 percent of the finish rates this site could measure. Every case anyone has verified is a game you can finish in about two hours. Length is doing the work. A two-hour game gets finished because it is two hours long.

Ubisoft's Jason VandenBerghe put the practitioner bar at the same place in 2013: when a studio sees finish rates over 30 to 40 percent, they celebrate.

Genre is length in disguise

Endless games do not have a finish line at all. PUBG and Counter-Strike 2 have no ending achievement to read, and a roguelike like Vampire Survivors shows a low number that means nothing. They are the wrong shape for the statistic, and lumping them in is how this number usually gets abused. The score on this site treats multiplayer and roguelike games on population instead and pays them no finish credit.

Among games that do have an ending, the pattern is that players finish what respects their time. A short narrative game gets finished several times more often than an eighty-hour role-playing game. When a store page advertises eighty hours of content, the honest reading is that those hours count only for the players who reach the end, and the achievement data says how many do.

What this does to the price math

This is why dollars per advertised hour flatters long games. If a game advertises 40 hours and one player in sixteen reaches the end, then 40 hours is the denominator for that one player. Everybody else is dividing by a number they never got near.

The BANG Score uses the finish rate where one exists, and full credit now lands at 40 percent, which on September 25, 2026 was about the top sixth of what this site could measure. It used to require 76 percent, which sounded demanding and was really a bonus only very short games could earn.

A correction

Until August 24, 2026 this guide carried per-game finish rates from this site's own detector, including about 10 percent for ELDEN RING and figures for Cyberpunk 2077, HELLDIVERS 2 and Black Myth: Wukong. Those were wrong. The detector had a fallback that, when it could not identify a completion achievement, returned the rarest achievement in the game instead and published that. It produced about two thirds of the finish rates on this site, and one of them was 88 percent for Detroit: Become Human against a real ending achievement of 4.2 percent.

The fallback is gone. The detector now declines to answer rather than guessing, so the site publishes far fewer finish rates and the ones it publishes are real detections. The genre averages and the review-score comparison this guide used to carry were computed over the bad data and have been removed rather than restated.

The method: the finish signal is the global unlock rate of the achievement that best marks a game's ending, from Valve's public data. It is absent for games without achievements, unreliable where a developer hides the ending achievement, and meaningless for endless genres, which the scoring treats separately.

Sources: Bailey and Miyata, "Improving video game project scope decisions with data", dataset at doi.org/10.17632/jc37n76wyt.1, sampled 2018 and covering Steam through 2016. Per-game rates read from Valve's public achievement data on August 23, 2026. Sony trophy rarity bands from PlayStation's own documentation.