Cashout Button Placement Cuts Mid-Round Exits 26% in 600 Sessions

Moving the cashout control from a corner menu into the persistent in-play panel reduced mid-round exits by 26% across a 600-session observational study, according to figures circulated by the operator group that ran the test between 3 February and 28 March 2025. The finding matters less for the headline number than for what it suggests about the relationship between interface friction and abandonment: players who could see the exit did not, on the whole, take it more often. They took it at different moments, and in smaller stakes.

The study covered 600 sessions drawn from a pool of roughly 41,000 eligible accounts, stratified by tenure and average stake band, with a control cohort using the legacy layout. "Mid-round exit" was defined as a cashout executed after the first settlement event but before the natural conclusion of the round — a definition that excludes pre-match and between-round withdrawals. The 26% figure is a relative reduction against control, not an absolute one; the control cohort exited mid-round in 18.4% of eligible rounds, the treatment cohort in 13.6%. That distinction matters, because a 26% relative drop on a low base is a smaller behavioural shift than the phrasing implies.

What the interface actually changed

The legacy layout placed cashout behind a two-tap path: a menu icon in the upper-right, then a confirmation sheet. The treatment layout promoted it to a fixed button in the bottom third of the viewport, visible throughout live play, with the same confirmation step retained. Nothing about the cashout mathematics changed — the offer engine, the margin applied to the cashout quote, and the settlement timing were identical across cohorts. The only variable was placement and the number of taps required to reach the confirmation sheet.

That isolation is the study's main methodological strength, and it is worth being precise about why. Cashout uptake is unusually sensitive to presentation. The quoted value moves with the underlying market, so a player who reaches the confirmation sheet three seconds later may see a materially different number. By holding the engine constant and varying only the path length, the operator could attribute the delta to friction rather than to price.

The counterintuitive part

Conventional product thinking holds that making an exit more accessible increases exits. The data ran the other way, and the mechanism the researchers propose is plausible: a visible button reduces the anxiety of feeling trapped. Players who cannot see the exit may abandon a position early through a different route — closing the app, navigating away, or simply letting a round lapse without engaging — because the perceived cost of finding the control exceeds the value of managing the position. When the control is visible, the decision to stay becomes an active one rather than a default.

The study reports a 31% drop in "app backgrounding during live round," which supports this reading. Backgrounding is a crude proxy for disengagement, but it moved in the same direction and roughly the same magnitude as the primary metric.

Where the reduction concentrated

The aggregate 26% masks a wide spread across stake bands, and the spread is the more interesting result.

The effect is strongest among low-stake players and weakest among high-stake players — which is what you would expect if the mechanism is friction-driven rather than price-driven. High-stake players already know where the control is; they have used it enough times that the two-tap path costs them nothing cognitively. Low-stake and newer players bear the full cost of search, and it is their behaviour that moved.

There is a second reading, less flattering to the operator. Low-stake players are also the cohort most sensitive to small changes in quoted cashout value, and a persistent button makes the quote continuously visible. If the visible quote is systematically worse than the player expects — because it reprices against a margin the player has not internalised — a persistent display could function as a deterrent rather than a convenience. The study did not measure whether treatment-cohort players accepted cashout offers at a different rate once they reached the confirmation sheet. That gap is significant.

The retention question the study did not answer

The commercial logic behind the test is straightforward: mid-round exits are, from the operator's perspective, a form of churn. A player who cashes out mid-round has ended their engagement with that round and may not return to the next one. Reducing mid-round exits by 26% is presented as a retention win.

But the study reports no session-length data beyond the round level, and no next-day return rate. A player who stays in a round because the exit is visible is not necessarily a player who places more rounds. It is equally consistent with the data that the treatment cohort simply held positions longer within the same number of rounds, with no change in overall engagement. Without a 30-day return metric — the standard window for retention attribution in this sector — the 26% figure describes a within-session behaviour, not a retention outcome. The operator's own summary acknowledges this obliquely, noting that "downstream retention effects were outside the scope of this phase."

A note on the sample

Six hundred sessions is a modest base for a four-cell stake-band breakdown. The over-€50 cell, at roughly 90 sessions after stratification, carries a confidence interval wide enough that the −12.1% figure should be treated as directional rather than measured. The under-€2 cell is the most robust, and it is also the cell where the effect is largest, which is convenient for the study's thesis but not evidence for it.

The observational design also cannot fully exclude selection effects. Sessions were assigned to cohorts, but the assignment method is not described in the material circulated, and if assignment correlated with account tenure — a plausible risk in stratified sampling — the treatment cohort may have contained a different mix of experienced players than the control. The reported balance checks cover stake band and tenure but not prior cashout usage, which is the variable most likely to confound the result.

There is also the question of what "mid-round exit" excludes. The definition counts only cashouts after the first settlement event. A player who abandons a round before that event — by closing the app, by letting the bet run to settlement without engagement — is not counted as a mid-round exit under either cohort. If the treatment layout reduced pre-settlement abandonment while increasing post-settlement cashouts, the headline metric would understate the total behavioural change and misattribute its direction.

What the number is actually good for

The 26% is best read as a lower bound on the effect of interface friction, not an estimate of its full size. The cleanest finding is the stake-band gradient: the more experienced the player, the smaller the placement effect, which is exactly the signature of a search-cost mechanism. That gradient is more portable across operators than the headline figure, because it does not depend on the specific layout, market mix, or player pool of the original test.

The open question is whether visibility and accessibility are the same variable. The treatment made the cashout both easier to reach and continuously visible; the design cannot separate the two. An operator testing a persistent but non-interactive display — a visible quote with the two-tap path retained — would learn whether the effect comes from knowing the exit exists or from being able to reach it in one tap. If the former, the product implication is a display change. If the latter, it is a control change, and the responsible-gambling implications diverge sharply between the two. A one-tap exit that reduces abandonment by making players feel less trapped is a different instrument from a one-tap exit that simply makes leaving easier, and the study, as published, does not tell us which one it built.