A 17% drop in slot session exits during the 21:00 hour sounds like the kind of number a marketing team would invent, so it's worth being precise about what was actually measured. In a controlled test across three UK-licensed operators running the same 14 slot titles, displaying a live RTP figure next to the reels reduced the share of players ending a session within 60 seconds of a losing spin by 17% between 21:00 and 22:00, compared with a control group seeing the standard paytable link. The effect was measured over 41 days in February and March 2025, across 268,000 sessions, and it did not hold at anything like the same strength outside that evening window.

What the test actually changed

The intervention was deliberately small. Players in the treatment group saw a persistent line beneath the reels reading something like "This game's RTP: 94.2% — average return per £100 staked over millions of spins." No spin count, no loss tracker, no nudge to take a break. The control group saw the game exactly as it ships today, with RTP reachable via the information icon and the paytable.

That matters because the UK already mandates accessible game information. Licence condition 3.2.1 and the associated RTS requirements mean RTP has to be available before play and within the game client. What the test changed was salience, not availability. The figure moved from a document you could open to a line you could not avoid.

Three findings stood out beyond the headline:

• The exit-rate reduction was concentrated in players aged 25–44, who accounted for 61% of the treatment group's effect.
• Average session length rose by just under two minutes, but average stake per spin fell 4.3% — players did not spend more, they churned less and bet slightly smaller.
• The effect decayed over the test period, from a 22% reduction in week one to 11% by week six.

That last point is the one worth arguing about. A 17% average is really a curve, and the curve points downward.

Why 9pm specifically

The 21:00–22:00 window was pre-registered, not cherry-picked after the fact, and the researchers' reasoning was straightforward. It is the peak hour for UK online slot play on a weekday, it sits after the post-dinner lull and before the late-night cohort arrives, and it is the hour in which session-ending decisions are most likely to follow a single painful spin rather than a planned stop.

Outside that hour the reduction was 6% — real, but within the range you'd expect from noise and minor interface changes. The 17% figure should be read as an evening-peak number, not a general one.

Why an RTP line might work at all

The standard objection is that RTP is the wrong number to put in front of a player. It describes a long-run theoretical return across millions of spins, not what any individual session will do. A player on a 96.1% game can lose 40% of their bankroll in twenty minutes and the RTP figure will not have been violated in any meaningful sense.

And yet the behavioural literature on salience of cost points the other way. People consistently underestimate the cost of repeated small transactions — the classic example being the price of a payday loan expressed as an annual percentage rate versus a per-£100 fee. Slots have the same structure: each spin is cheap, the aggregate is not, and the aggregate is invisible in the moment. An RTP line is a crude cost signal, but it is a cost signal where previously there was none.

There's a second mechanism worth naming. An RTP figure implicitly tells a player that the house edge is a fixed property of the game rather than a run of bad luck that will correct itself. That undercuts the "due a win" reasoning that sustains a lot of extended sessions. The test didn't measure belief change directly, which is a gap, but the stake-per-spin reduction is consistent with it.

The counter-argument: 94.2% is not a warning

The strongest objection is that an RTP line may read as reassurance rather than caution. A player who sees 94.2% may conclude the game is "nearly fair" and that £100 in yields £94.20 back — which is true only in the sense that a coin flipped a million times lands heads half the time. In the short run, variance dominates completely, and on a high-volatility title with a 94.2% RTP the realistic outcome distribution over 200 spins spans everything from a small loss to a wipeout.

A 2023 study of Australian machine players found that displaying return percentages had no significant effect on expenditure and, in one arm, slightly increased persistence among regular players who read the figure as evidence the game was regulated and therefore safe. The UK test did not replicate that, but it also ran for six weeks against a novelty effect that was visibly fading.

What this means for operators and the regulator

The Gambling Commission has spent the last three years pushing toward what it calls "meaningful" customer interaction — the shift from passive information to active intervention under the remote customer interaction requirements that came into force in 2023. A persistent RTP line sits awkwardly between the two. It is information, not intervention, but it is information placed where it cannot be ignored.

For operators, the commercial logic is not obvious. A 17% reduction in session exits means more sessions continuing past a losing spin, which in a 94% RTP game is not automatically good for the player or the operator. The 4.3% drop in stake per spin probably offsets the extra time. Nobody in the test published a net gaming revenue figure, which is the number the industry will actually want before adopting anything.

There is also a design question the test leaves open. RTP is one number among several that could sit beside the reels — session loss to date, time on site, spins since last deposit. The evidence here says that a number in that position changes behaviour. It says nothing about which number does the most good, or whether stacking them produces a warning or a wall of text players learn to ignore entirely.

The 9pm question

Which brings back the hour. If the effect is genuinely concentrated between 21:00 and 22:00, that suggests something specific about decision-making at that time — fatigue, alcohol, the last hour before sleep, the point at which a losing session becomes a decision about tomorrow. If it is instead an artefact of when traffic peaks and therefore when the sample is thickest, the whole finding shrinks.

The test design doesn't separate those two explanations. A follow-up running the same intervention at 07:00 and at 01:00 would. Until then, the honest reading is that a visible RTP line reduced evening session exits by 17% over six weeks in one sample, that the effect halved within that period, and that nobody has yet shown whether the players who stayed were better or worse off for it.