How we estimate owners

What the number is

Steam publishes a game's total review count, but not its sales or owner count. Independent analysts, most notably Mike Boxleiter in 2014, noticed that reviews track owners at a roughly stable ratio for a given game. Multiply a game's total reviews by that ratio and you get a rough owners estimate, commonly called the "Boxleiter number" or "Boxleiter method" after him. It has been revised many times since, and we do the same here: this is our own estimate, refreshed alongside a game's ratings, never a number Valve, a publisher, or a data partner gave us directly.

Why a range, not one number

The review-to-owners ratio moves with a game's price and how long it has been on sale. Older games have had more time for owners to leave a review, and a review from ten years ago came from a much smaller, more review-prone Steam audience than a review today. Cheaper games convert a larger, more casual owner base into reviews at a lower rate per owner. We fold all of that into a low, mid, and high multiplier for each combination of release year and price, so the estimate always reads as three numbers, not one that looks more precise than it is.

The floor: 10 reviews

Below 10 total reviews we don't show an estimate at all. A handful of reviews is too noisy for a ratio to say anything useful, and a made-up-looking number there would do more harm than a blank field.

The ratio table

Each cell is the low, mid, and high multiplier we apply to a game's total reviews, based on its release year and price. Older and cheaper games sit on higher multipliers. There is no column for free games: free-to-play titles are not estimated at all, see below.

Each cell reads low / mid / high. A game with 10,000 reviews in a cell reading 20 / 35 / 60 gets an estimate of roughly 200,000 to 600,000 owners, with 350,000 as the midpoint.

Games with very many reviews

Above 200,000 reviews the multipliers shrink. Every doubling of the review count cuts them by about a fifth, so at one million reviews the table values are multiplied by roughly 0.57.

How we checked it

We compared the estimate with 10 sales figures that studios or publishers published themselves. Steam does not publish sales, so these are the only firm numbers available, and only 1 of them are stated for Steam alone with a review count from the same article. That is too few to say how accurate the estimate is in general. The others are used for narrower checks: 1 is a Steam figure with a newer review count, 5 cover more than Steam and only cap the estimate from above, and 3 are launch weeks, when reviews trail purchases.

• Valheim, 10,000,000 units, 2022-04-25. Copies sold since the Steam Early Access launch, no breakdown by store published. NME, 10 million copies and more than 293,000 Steam reviews
• V Rising, 3,900,000 units, 2024-02-01. Copies sold since the Steam Early Access launch, PC only at the time. Game World Observer, 3.9 million copies and 67k Steam reviews
• Kenshi, 2,300,000 units, 2024-04-10. Copies sold since launch, no store breakdown given; the Epic Games Store version launched the same day. Game World Observer, 2.3 million copies and more than 69k Steam reviews
• Core Keeper, 2,000,000 units, 2024-08-28. Units sold during Early Access, as reported; this entry does not record a platform split. Game World Observer, 2 million units and 26.3k Steam reviews
• Phasmophobia, 20,000,000 units, 2024-12-03. Publisher boilerplate: "almost 20 million copies on Steam" (22 million on all platforms). Kinetic Games press release, 3 December 2024
• Terraria, 32,000,000 units, 2024-07-15. About 32 million copies on PC, 58.7 million on all platforms. PC Games Insider, over 32 million copies on PC alone
• Stardew Valley, 26,000,000 units, 2024-12-30. 26 million copies on PC, 7.9 million on Switch, 41 million on all platforms. Game Developer, 26 million on PC and 41 million overall
• Megabonk, 1,000,000 units, 2025-10-02. Developer announcement of one million copies on Steam, 15 days after release. Game World Observer, one million copies, 14.9k Steam reviews
• PEAK, 4,500,000 units, 2025-07-08. Studio figure at about three weeks after the 16 June release. 80.lv, 4.5 million copies, 39,655 reviews at time of writing
• Sledding Game, 100,000 units, 2026-05-05. Developer post, more than 100,000 copies in five days. Notebookcheck, 100,000 copies in five days, more than 1,600 reviews

On the Steam-only figures the middle estimate is off by 1.9 times at the median, and the published figure sits inside the low to high range for 0 of 1. The lagged figures are scored on their own: 1 of 1 inside the range. Our low end is under every bound figure. We never use figures from other estimation sites to set the table.

Where the starting ratios come from

The table itself is our own reading of public write-ups on the Boxleiter number, which differ widely, not a single agreed formula and not a citation of any one of them. Treat it as a labelled estimate.

Where it breaks down

• Free-to-play games are not estimated. For a free game, owners really means installs, and the share of installs that ever leaves a review varies by an order of magnitude between titles, so no single multiplier gives a comparable number. We show "not estimated for free games" instead of a figure that would look precise while being a guess.
• Games in their first weeks on sale read low. Reviews trail purchases, so a game that sold a million copies in two weeks can have a review count that implies a fraction of that. In our launch-week figures the published sales were between 2.0 and 3.8 times the middle estimate.
• Games sold mostly through bundles (Humble Bundle, Fanatical, and similar) accumulate owners who never intended to play the game, so review rates run far below what a normal sale would produce. The estimate reads low for bundle-heavy titles.
• A game that changed price significantly during its lifetime (permanent price cuts, long early-access discounting) only has one price band applied for its whole review history, even though different reviews were left at different price points.
• This is not partner data. We have no access to Steam's, Valve's, or any publisher's actual sales or wishlist figures. Every number under "Est. owners" is derived entirely from the public review count using the table above.