A £5 maximum stake on UK online slots produced a 19% increase in average session length among players who hit the limit before midnight, according to operator-level data from a six-month trial covering roughly 2.4 million sessions. The finding runs against the assumption behind the £2 and £5 stake caps introduced for land-based venues and proposed for online: that lower limits shorten play and reduce harm. In this dataset, the opposite happened for a specific group of players — those who would otherwise have staked above £5 — and the effect was concentrated in the hours before midnight rather than across the whole day.

The data comes from a mid-sized operator running a voluntary £5 ceiling on slot spins across its UK-facing site between March and August 2024, compared against a matched control cohort on a sister brand without the cap. Sessions were defined as continuous play with no gap longer than 15 minutes. The 19% figure refers to mean session length in minutes for the capped group relative to control, weighted by stake band. It is not a claim about every player, and the operator is keen to stress that. But it is a large enough effect, across a large enough sample, to warrant a closer look at what stake limits actually do to behaviour.

What the £5 cap changed, and what it didn't

The headline number hides a split. Players whose typical spin was already below £5 — roughly 71% of the sample — showed no meaningful change in session length. The effect was almost entirely driven by the 29% who, in the control cohort, regularly staked between £5 and £20 per spin.

For that group, three things happened at once:

• Spin count rose. Mean spins per session went from 214 to 268, a 25% increase.
• Session length rose. Mean session length went from 41 minutes to 49 minutes.
• Gross loss per session fell. Mean net loss dropped 8%, from £34.20 to £31.50, despite the longer session.

That third point is the one operators and regulators should sit with. The cap did reduce the intensity of loss per minute of play. It did not reduce total time spent playing, and by one measure — session length — it increased it.

The midnight effect

The 19% figure is an average across the day. Broken down by hour, the increase was not uniform. Sessions starting between 21:00 and 23:59 showed a 26% increase in length; sessions starting between 09:00 and 17:00 showed a 6% increase. The late-evening window is where the cap bit hardest, and where the compensatory behaviour was strongest.

Why midnight specifically? The trial capped deposits as well as stakes for a subset of players — a £250 daily deposit limit resetting at midnight. Players who reached the stake ceiling earlier in the evening and then hit the deposit limit had a choice: stop, or wait for the reset. A meaningful share waited. The session didn't end; it paused, then resumed after 00:00. When the operator measured "session length" including that pause, the number inflated. When it measured "active play before midnight," the increase was closer to 9%.

This is a measurement artefact worth flagging. Session length is a blunt instrument. A player who plays for 40 minutes, waits 90 minutes for a deposit reset, and plays for another 40 minutes has a 170-minute session by the operator's definition and an 80-minute session by any sensible definition of play. The 19% headline depends on which definition you use. The operator's own analysts concede the 9% figure is the more honest one for pre-midnight active play.

Why lower stakes can lengthen play

There is a straightforward behavioural explanation, and it doesn't require assuming players are irrational. Lower stakes per spin mean the same bankroll buys more spins. A player with £100 and a 5% house edge expects to lose £5 per £100 wagered; at £5 a spin they can make 20 spins before their expected loss equals one spin's stake, at £1 a spin they can make 100. If the player's goal is time spent playing rather than money won — and for a large share of slot players it plainly is — a lower stake is a more efficient way to buy time.

The operator's data supports this. The capped group's average stake per spin fell from £7.80 to £4.60, a 41% reduction. Spins per session rose 25%. Time per session rose 19%. The player is, in effect, trading stake size for duration. The cap didn't stop them playing; it changed the exchange rate.

This is not a new observation. The 2020 review of fixed-odds betting terminal stakes found similar compensatory patterns in some venue data, though the evidence was mixed and the £2 cap was justified primarily on loss-per-session grounds rather than time-on-device. What the online data adds is scale and precision: 2.4 million sessions is enough to see the effect clearly, and the hour-by-hour breakdown shows where it concentrates.

The loss-per-session defence

None of this necessarily argues against stake caps. If the policy goal is reducing loss per session — and for the most exposed players, it should be — an 8% reduction in mean net loss is a real gain, achieved without reducing the number of players. The 19% increase in session length is a cost, but it may be a cost worth paying if the alternative is higher losses per minute.

The problem is that session length and loss are not independent. Longer sessions expose players to more variance, more opportunities to chase, and more time in the state where decisions get worse. A player who loses £31.50 over 49 minutes may be worse off in wellbeing terms than one who loses £34.20 over 41 minutes, even though the second player lost more money. Regulators have generally measured harm in money, not time. This data suggests that's an incomplete metric.

What this means for the £5 debate

The government's consultation on online slot stake limits closed in 2023, and the £5 figure has been the upper bound under discussion, with £2 and £15 also on the table. This trial doesn't settle that question, but it does complicate the argument that a lower cap is straightforwardly safer.

Three implications stand out:

1. Stake caps and deposit limits interact. The midnight effect was driven by a deposit limit reset, not the stake cap alone. A cap without a matching deposit limit may produce different behaviour. Operators designing harm-reduction packages should model the two together, not separately.
2. Session length needs a better definition. The 19% figure is real but fragile. Any regulator using session length as a harm indicator needs to specify whether pauses count, and how deposit resets are handled. Otherwise the metric is gameable in both directions.
3. The compensatory effect is concentrated, not universal. 71% of players showed no change. Policy should target the 29% who did, and within that group the late-evening players who waited for the reset. A blanket cap is a blunt tool for a behaviour that is specific and identifiable.

The open question

The trial ran for six months. That's long enough to see a behavioural shift but not long enough to know whether it persists. Do players who lengthen sessions under a £5 cap eventually adapt — either by reducing play or by finding ways around the cap — or does the longer session become the new normal? The operator hasn't published follow-up data beyond the trial window, and no independent replication exists.

That's the gap. A single operator's six-month trial, however large, is one data point. The 19% figure will be cited in policy debates for years; whether it holds up under replication, and whether the 9% active-play figure is the one that matters, is still unresolved. Anyone building a harm-reduction framework on this number should treat it as a prompt for further study, not a settled finding. And players reading it should note the obvious: a cap that lengthens your session by a fifth is not obviously doing you a favour, whatever it does to your loss per spin.