Loss Chasing Persists 31% Longer Under Cashback Than Flat Rebates

A 2024 behavioural study of 4,180 tracked accounts across three licensed European operators found that players receiving cashback on net losses continued betting for a median of 31% longer after a losing session than players on flat, turnover-based rebates of equivalent monetary value. The gap widened with stake size: among accounts wagering above €5 per spin, the differential reached 44%. The finding complicates the industry's assumption that cashback and flat rebates are interchangeable retention tools distinguished only by accounting mechanics.

Two refund structures, two behavioural signals

Cashback and flat rebates are often treated as variants of the same instrument. Both return value to the player; both are typically calculated weekly or monthly; both appear in the same promotional slot on an operator's site. The distinction that matters is what triggers the payout.

A flat rebate is turnover-linked. A player wagering €10,000 on slots at a 0.2% rebate receives €20 regardless of whether the session ended up or down. The rebate is a volume discount, functionally closer to a loyalty scheme than to insurance.

Cashback is net-loss-linked. The same player, down €800 after that €10,000 turnover, receives a percentage of the loss — commonly 5% to 15% at the tiers examined in the study, so €40 to €120. If the player finishes up, the cashback is zero.

That asymmetry is where the behavioural divergence originates. A turnover rebate is earned by playing. A loss rebate is earned by losing. The second structure creates a contingent relationship between a negative outcome and a future reward, and the study's account-level data suggest players learn that contingency faster than operators assume.

The measurement problem

Self-reported gambling data is unreliable, particularly for loss-chasing, which carries stigma. The study avoided surveys entirely. It used session-level wagering logs matched to cashback credit events, defining "continuation" as any further real-money wager placed within 72 hours of a session that closed at a net loss exceeding 20% of the player's trailing 30-day average session result. Median continuation time was calculated from session close to final wager before a 14-day dormancy threshold.

The 31% figure is a median across the full sample. It is not a mean, and it is not adjusted for the handful of extreme accounts that bet for weeks without pause. The distribution is right-skewed, and the researchers report the interquartile range separately: cashback players continued for 4.1 to 19.6 hours; flat-rebate players for 2.8 to 15.2 hours.

Why the 31% figure survives scrutiny

Three alternative explanations were tested and rejected.

First, selection. If cashback attracts a different player type — higher-risk, more loss-tolerant — the effect would be compositional rather than causal. The study addressed this with a within-player design: 1,240 accounts experienced both structures during the observation window, as operators rotated promotional calendars. Within that subgroup, the same account continued 27% longer under cashback than under flat rebates. The effect shrank but did not disappear.

Second, stake size. Cashback tiers in the sample were correlated with higher average stakes. But the effect held when the sample was stratified into four stake bands, with the smallest differential (18%) in the lowest band (under €0.50 per spin) and the largest (44%) in the highest.

Third, credit timing. Cashback was credited weekly in two of the three operators and monthly in the third. Flat rebates were credited on the same schedules. Timing did not predict continuation; structure did.

The mental accounting behind the gap

The mechanism the authors propose is straightforward and consistent with prior work on mental accounting. A flat rebate arrives as a separate, unconditional credit. Players describe it, in the qualitative follow-up interviews (n=62), as "a bonus" or "a discount on my play." It is not mentally linked to the preceding loss.

Cashback arrives as a partial refund of a specific loss. Players in the interview subset described it as "getting some of it back" and, critically, as "a reason to keep going." The credit does not restore the loss; it reframes the loss as recoverable, which is the precise cognitive move that characterises loss-chasing.

This is not a novel insight in gambling research. What is new is the quantified differential in a real-money, account-level dataset rather than a laboratory task. The 31% figure gives operators and regulators a concrete number to work with when evaluating whether a promotional structure is a retention tool or a risk amplifier.

The regulatory blind spot

Most licensing regimes regulate bonus mechanics — wagering requirements, maximum cashout, expiry — without distinguishing between refund structures. A 10x wagering requirement on a €50 cashback credit and a 10x requirement on a €50 turnover rebate are treated identically on paper. The behavioural data suggest they are not identical in effect.

Malta and Sweden have both moved toward stricter bonus rules in the past three years, but neither jurisdiction's guidance addresses the loss-contingency of the credit itself. The UK Gambling Commission's 2023 consultation on bonus terms mentioned "loss-making" promotions only in passing. If the 31% differential replicates, that omission looks less like an oversight and more like a gap in the evidence base regulators were working from.

What operators do with the number

The commercially awkward implication is that cashback may be doing two things at once: retaining players and extending losing sessions. Operators optimising for gross gaming revenue have little incentive to separate the two. Operators optimising for player lifetime value net of regulatory and reputational risk have a harder calculation.

A few possibilities are already visible in the market. Some operators have shifted from percentage-of-loss cashback to fixed-credit structures that trigger on session count rather than loss magnitude, severing the contingency. Others have capped cashback at lower tiers and moved high-value players onto turnover rebates. Neither approach has been formally evaluated, and the study's authors are explicit that they measured continuation, not harm — a longer session is not automatically a worse one.

The open question is whether the 31% differential reflects a structural property of loss-contingent rewards or a transient effect that decays as players learn the payout schedule. The study ran 14 months; the effect was stable across the final nine. That is long enough to be interesting and short enough to be inconclusive. A replication across a full promotional cycle, ideally with operators disclosing the outcome data their regulators already hold, would settle it. Until then, the number stands as a caution rather than a verdict — and as a reminder that how a bonus is framed may matter more than how much it is worth.