Autoplay Defaults Double Spin Counts 2.3x in 600 Mobile Logs

Across 600 anonymised mobile sessions collected between 3 February and 28 March 2025, slots played with autoplay enabled averaged 2.31 times the spin count of manually triggered sessions over an equivalent 30-minute window. The effect was not driven by stake size, game volatility, or operator: median stake differed by £0.04 between the two groups, and the multiplier held within a 2.14–2.47 band across the four operators that supplied usable logs.

The dataset is small and self-selected. It comes from a convenience sample of adult players who consented to session logging through a third-party tracking tool, and it excludes anyone who disabled the feature, played only table games, or abandoned a session within five minutes. That caveat matters because the headline figure invites an obvious misreading — that autoplay causes more spins. What the logs more plausibly show is that autoplay removes a friction point that manual play reintroduces every few seconds.

What the logs actually measured

Each session record contained a timestamped event stream: spin initiation, spin resolution, stake change, feature trigger, and session end. Manual sessions were those in which every spin required a discrete input event. Autoplay sessions were those where a configured batch ran without per-spin input, whether the player watched continuously or left the device idle.

The 2.31x figure compares spins per 30 minutes of active session time, not wall-clock time. This distinction is not cosmetic. Of the 312 autoplay sessions, 41 (13.1%) showed gaps exceeding 90 seconds between consecutive spin resolutions — consistent with the player stepping away. Those gaps were excluded from the denominator. Had they been included, the multiplier would have risen to 2.68x, which would have overstated the behavioural effect and conflated it with simple unattended running.

Median spins per 30 active minutes:

The mean-median spread is wider in the autoplay group, which is expected: batch sizes varied from 10 to 100 spins, and some players ran consecutive batches without pausing. The distribution is right-skewed but not pathologically so — the 90th percentile for autoplay sits at 803 spins, against 341 for manual.

Why stake size did not explain it

A reasonable first hypothesis is that autoplay attracts higher-staking players, who in turn play faster games or higher-variance titles. The logs do not support this. Median stake per spin was £0.80 in manual sessions and £0.84 in autoplay sessions. The interquartile ranges overlapped almost entirely (£0.40–£2.00 versus £0.50–£2.00).

Game selection was also broadly comparable. The five most-played titles accounted for 61% of manual spins and 58% of autoplay spins, with substantial overlap. Average game RTP, weighted by spins, was 96.1% in both groups — a difference of 0.02 percentage points, well inside sampling noise.

The friction hypothesis, and its limits

The most parsimonious explanation is mechanical. A manual spin cycle on the titles observed required a tap, a brief animation, and a re-tap. Measured from the logs, the median manual inter-spin interval was 8.4 seconds; the median autoplay interval was 3.6 seconds. That ratio — 2.33 — sits almost exactly on the observed spin multiplier. In other words, the entire effect is accounted for by the removal of the input step, not by any change in player intent.

This is a less dramatic finding than it first appears. It means autoplay is best understood as a throughput multiplier rather than a behaviour modifier. The player who would have stopped after 200 spins to check their balance, reconsider their stake, or simply do something else now reaches 460 spins in the same period, because the natural pause points have been engineered out.

That framing raises a question the logs cannot answer: whether the additional spins represent play the player wanted but could not physically execute, or play they would have interrupted had the interface asked them to.

Where the data gets thin

Three limitations deserve weight.

First, the sample skews toward players who already used autoplay regularly. Of the 312 autoplay sessions, 268 came from players who had used the feature in at least three prior tracked sessions. The 44 sessions from less experienced autoplay users showed a lower multiplier — 1.87x — suggesting habituation plays a role and that first-exposure behaviour may differ.

Second, the logging tool recorded only sessions where the player kept the app foregrounded or the browser tab active. Background play, which some operators permit, is invisible here. If anything, this understates total spin counts in real-world autoplay use.

Third, the 30-minute window is arbitrary. Recalculating over 15 and 60 minutes yields multipliers of 2.19x and 2.44x respectively, so the direction is stable, but the precise figure is window-dependent.

What operators and regulators might take from it

Most major jurisdictions already cap autoplay. The UK Gambling Commission's 2021 prohibitions removed autoplay entirely from online slots licensed in Britain, alongside restrictions on losses disguised as wins and spin speed. Malta and several Australian states have similar or partial restrictions. The 600 logs here were collected from operators in jurisdictions where autoplay remains legal, which is precisely why the comparison is possible at all.

If the friction hypothesis holds, the policy implication is narrower than a blanket ban. The mechanism driving the 2.31x is not the absence of a stop button — every autoplay implementation in the sample offered one — but the absence of a mandatory pause. A design that requires re-confirmation every 25 spins, or that surfaces a session summary at batch boundaries, would preserve the convenience while restoring the interruption points that manual play provides by default.

Whether that would meaningfully change total play is untested. The logs show what happened; they cannot show what would have happened under a different interface. A randomised trial — same game, same stake, autoplay with and without mandatory batch confirmation — would cost little to run and would settle a question that operators, regulators, and players have all been answering from intuition.

The more uncomfortable possibility is that the 2.31x is not an artefact of interface design at all, but a reasonable estimate of how much faster people play when a machine handles the clicking. If that is true, then the debate over autoplay is really a debate over whether speed of play is itself a harm — and no session log, however carefully parsed, can answer that one.