What happens when heroes play alone: playtest data

Key findings

• 12,500+ players and about 127,000 hours played across the whole August 2026 Steam playtest.
• 43% of players used offline play (whole August playtest).
• About 60% of fights happened without the player (whole August playtest).
• 2.5 times more heroes than human players were online at peak (whole August playtest).
• 57% of hours played were unattended (7–23 August 2026).
• 68% of heroes online had no player, as an hourly average, with hourly values between 63% and 74% (18–23 August 2026, UTC, 6-hour offline limit).
• Raising the offline limit from 2 to 6 hours on 18 August 2026 raised the share of heroes playing alone from 31% to 68% (7–23 August 2026, before vs after the change).

The chart above shows the last of these windows hour by hour.

What did we test, and how?

Dominus Automa is an automated multiplayer idle RPG. Each player has a hero that acts on the game server, not on the player’s device. The player sets a plan: where to go, what to fight, what to gather. The hero then carries out that plan by itself, whether the player is watching or not. We describe the game in more detail in Dominus Automa: the game built to prove the tech.

Because the hero runs on the server, it can keep playing after the player closes the game. During the playtest this was capped by an offline limit: the number of hours a hero kept playing after its player left. When the limit ran out, the hero stopped until the player came back.

We use these terms throughout:

• Attended: the hero is online and its player is in the game.
• Unattended: the hero is online and playing, but its player is not in the game.
• Offline play: a session in which the hero kept playing after the player left.

The figures come from three windows. They are not interchangeable, and we name the window each time:

1. The whole August playtest (a month-long open playtest on Steam). These are the figures we presented publicly at the Strefa Inwestorów Demo Day on 15 October 2026.
2. 7–23 August 2026, the window of our internal playtest report. The offline-limit experiment sits inside it.
3. 18–23 August 2026, UTC, the hourly view in the chart above. It starts on the day the limit went up, so every hour in it ran under the 6-hour offline limit. We do not average the full 7–23 August window by hour, because that would blend the 2-hour and 6-hour limits.

What did the whole August playtest show?

Across the whole August playtest:

• 12,500+ players took part and played about 127,000 hours in total.
• 500+ heroes were online at peak.
• 1 in 4 players played more than 2 hours, 2% played more than 50 hours, and the top player reached 447 hours.
• Steam wishlists grew from 5,500 to 16,500.
• Day-1 retention was above 36% among players from Tier 1 markets.
• A survey of 270 players was 91% positive.

The survey figure needs care. 270 responses is a small sample, and the people who answer a playtest survey are usually the ones who liked the game enough to stay. Read it as a signal from engaged testers, not as a measure of how all players felt.

How much of the game was played without the player?

This is the question the playtest was built to answer. We have five measures. Each one counts something different, so they should not be compared as if they were the same number.

Two things stand out.

First, offline play was not a niche behaviour. 43% of players used it, and by fight count most of the game, about 60% of fights, happened without anyone watching (both whole August playtest).

Second, the hourly data shows that unattended heroes outnumbered attended ones in every hour of the day. In the 18–23 August window, the share of heroes with no player was highest in the early UTC hours, roughly midnight to 6:00, and lowest in the UTC afternoon, when more players were at their screens. Even at its lowest hourly value, 63%, almost two in three heroes online were playing alone.

What happened when we raised the offline limit?

On 18 August 2026 we raised the offline limit from 2 hours to 6 hours. Within the 7–23 August window, the share of heroes online playing alone was 31% before the change and 68% after it.

What this shows: when heroes were allowed to keep playing longer after their players left, the share of heroes playing alone more than doubled. Players used the extra time.

What it does not show:

• It is not a controlled test. All players moved to the new limit on the same day. There was no group that stayed on 2 hours, so later changes in the player mix, news about the playtest, or the point in the playtest when players joined could all play a part.
• It does not show that more unattended play made players happier or kept them longer. It measures what heroes did, not what players thought about it.
• It does not tell us where the limit should be. We tested two values, not a range.

Why does this matter?

The idea behind Dominus Automa is that many people still want to play games but have less time to do it. Public data on players supports that starting point:

• The average US video game player is 37 years old, according to the Entertainment Software Association, 2026 Essential Facts (3 June 2026).
• Two in three PC and console players play less than 10 hours a week, according to MIDiA Research (Q4 2024), via GameDev Reports.
• Of gamers who enjoy gaming less than before, 47% cite work or family responsibilities “a lot” and 45% cite less free time, according to YouGov (2 July 2024).

Rafał Bieniek, our CEO, put it this way at the Demo Day:

“Players aren’t losing interest in games. They’re losing the time to play them.” — Rafał Bieniek, Strefa Inwestorów Demo Day, 15 October 2026

The playtest gives a first, partial answer to whether a hero that plays on its player’s plan fits that situation: given the option, most of the playing happened without the player present. For how this differs from smarter non-player characters, see AI NPCs vs AI player agents. For the wider category, see agentic AI in games.

What are the limitations of this data?

• One playtest. All figures come from a single month-long playtest of one game. They have not been repeated.
• Self-selected testers. Playtest players chose to join an early test of an automated RPG. They are likely more open to automation than players in general.
• Small survey. The 91% positive figure comes from 270 responses.
• Different windows. Some figures cover the whole August playtest, others 7–23 August or 18–23 August. Comparing numbers across windows can mislead, which is why we state the window each time.
• Company-reported. These are our own figures from our own server data and analytics. They have not been audited by a third party.

What do we test next?

Our next playtest is planned for November 2026. It adds:

• Player-set hero paths, so players can set the route their hero follows.
• More maps.
• Teams.
• Chat with heroes on Telegram, so a player can talk to their hero outside the game. This is in closed testing now.

We will publish the results in the same way, with the window and definition stated for every figure. To follow the game itself, visit dominusautoma.com.

How to cite this

Metashift Labs, “What happens when heroes play alone: playtest data”, 21 October 2026, https://metashiftlabs.com/insights/playtest-data-heroes-playing-alone