A midnight curfew on slot autoplay cut average next-day session returns by 22% among UK players who had been using the feature, according to operator-side data drawn from 4.2 million sessions across six brands between October 2024 and March 2025. The drop is measured against a matched control group of manual-spin players on the same titles over the same period, and it holds after adjusting for stake size, time of day, and game volatility. In plain terms: removing autoplay didn't just change how people played — it changed what they got back.

The finding matters because the policy debate around autoplay has been conducted almost entirely in terms of harm reduction, with the implicit assumption that restricting the feature is a costless good. This data suggests there is a measurable cost, and that it lands on the players least likely to be flagged as at-risk.

What the data actually shows

The headline figure comes from a cohort of 38,000 accounts that used autoplay on at least 60% of their slot sessions in the three months before the restriction took effect. After midnight autoplay was disabled across the six participating brands, those accounts' next-day returns — defined as net position after a session, expressed as a percentage of total staked — fell from a mean of −4.1% to −5.0%. That's the 22% relative drop in the title, and it's a figure about outcomes, not engagement.

Two things need saying about that number. First, it is not a claim that players lost more money in absolute terms. Median session length in the cohort fell by 14 minutes, and total daily staked fell by 9%. Players who lost autoplay played less. The 22% is a rate, not a total.

Second, the mechanism is not fully settled. The most plausible explanation is that autoplay was doing quiet work that manual spinning doesn't: it kept spin cadence steady and, crucially, it kept players on titles long enough to ride out variance. Manual players in the control group were 31% more likely to end a session within 40 spins of a losing run — a behaviour that locks in a loss before the RTP has any chance to assert itself. Autoplay, whatever its harms, was a crude variance-smoothing device. Take it away at the stroke of midnight and you get more early exits, and early exits in a negative-EV game mean realised losses cluster closer to the worst end of the distribution.

The midnight detail is doing real work

The policy in question wasn't a blanket autoplay ban. It was a curfew — autoplay disabled between 00:00 and 06:00 — introduced by several brands in late 2024 as a voluntary measure ahead of expected Gambling Commission guidance. That timing is why the effect shows up in next-day returns rather than same-session returns. Late-night players who relied on autoplay were pushed into manual play at exactly the hour when decision quality is poorest, and the data shows it: manual sessions starting between midnight and 02:00 had a 19% higher rate of what the researchers call "rapid re-stake" — a new spin within two seconds of the previous result — than manual sessions earlier in the evening.

So the curfew may have produced a double effect. It removed the variance-smoothing of autoplay, and it replaced it with manual play at the worst possible hour for manual play. A blanket ban at all hours would likely show a smaller next-day effect, because daytime manual players are, on this data, less erratic.

Why the 22% figure should be handled carefully

The study is operator-side, which means it sees only what happens on those six brands. Players who migrated to other operators after the curfew are invisible, and migration is exactly the kind of behaviour you'd expect from heavy autoplay users. If the heaviest users left, the remaining cohort's figures are biased toward players who adapted rather than players who were displaced.

There's also a selection problem in the control group. Manual-spin players are not a random sample of slot players; they differ in stake patterns, session length, and probably in risk appetite. The researchers matched on observable variables, but "prefers manual spinning" is not fully observable, and it correlates with things that affect returns.

And the return metric itself deserves scrutiny. Net position as a percentage of staked is sensitive to session length in ways that can flatter or punish a cohort depending on how the session ends. A player who quits after 40 spins on a bad run shows a worse rate than a player who plays 400 spins through the same variance. The 22% may partly be measuring quitting behaviour rather than game outcomes — which is still interesting, but it's a different claim from "players got worse returns."

The researchers acknowledge this. Their robustness checks, which exclude sessions under 50 spins, cut the effect to 16%. That's the number I'd defend in a room full of sceptics. The 22% is the headline; 16% is the floor.

What this means for the autoplay debate

The Gambling Commission has been circling autoplay for years without landing a decisive rule, and this data explains part of the hesitation. Autoplay is a genuine harm vector for a minority — the players who set 500 spins and walk away are the ones the feature was designed for, and they're the ones a curfew protects. But the population that uses autoplay casually, as a cadence tool rather than a loss-chasing device, is larger than the policy conversation has assumed, and the curfew imposes a real cost on them.

That cost isn't obviously unacceptable. A 22% relative drop in returns for a cohort that was, by definition, using a feature associated with harm is a trade-off a regulator might take willingly. The question is whether the trade-off was measured before the policy was adopted or after. On this evidence, it was after.

There's a further wrinkle for the UK market specifically. The brands that introduced the curfew did so voluntarily, which means the playing field is uneven. A player who loses autoplay at midnight on one site can open another site where the feature still works. If the Commission moves to a blanket rule, the migration problem disappears but the return effect probably shrinks, because the daytime manual play that replaces autoplay is less erratic than the midnight version. If it doesn't move, the voluntary curfew is a competitive disadvantage for the operators who adopted it — and a natural experiment that will keep producing data.

The open question

The part of this that nobody has answered is whether the 22% is a permanent feature of manual play or a transition effect. Players who lost autoplay in October 2024 had years of autoplay habit behind them; their manual play in the following months may simply have been unpractised. If returns recover over a longer horizon — say, 12 to 18 months — then the curfew's cost is a one-off adjustment, not an ongoing drag, and the policy looks considerably better. If they don't recover, then autoplay was doing something to player outcomes that manual spinning can't replicate, and the harm-reduction case for restricting it has to be weighed against a persistent cost to the majority of users who never had a problem with it.

The data to answer that question is being generated right now, on the same six brands, and nobody has published it yet. That's the number to watch — not the 22%, but what it looks like a year from now.