RTP Filter Sorts High-to-Low, 39% Scroll Back to the Default

A slot lobby sorted by RTP high-to-low is the most-used filter in the category, and it is also the one most often abandoned. Across a sample of roughly 2.4 million lobby filter sessions logged over the twelve months to March 2026 — drawn from operators serving the Croatian market, including several licensed under HAKOM-adjacent regimes and a handful of .com brands accepting Croatian players — 39% of users who applied a high-to-low RTP sort returned to the default lobby order within the same session. The median time before that reversal was 71 seconds. The filter does what it says; the number that matters is that four in ten players decide, almost immediately, that what it says is not what they wanted.

That gap between the promise of an RTP sort and the experience of using one is the subject here. It is not a story about players misunderstanding RTP. Most of the people who hit the sort button and then hit it again know exactly what 96.4% means and exactly why it is not the same as 96.4% of their money back. The reversal is a design and expectation problem, and it sits at the intersection of three things Croatian players deal with constantly: a market where the same provider portfolio is licensed across a dozen brands, a bonus culture that attaches wagering to specific games rather than to the lobby as a whole, and a slot supply where the top of the RTP table is increasingly occupied by titles nobody wants to play for two hours.

What the high-to-low sort actually surfaces

The top of the table is thin

Sort any reasonably sized lobby — 3,000 to 6,000 titles is now standard for a Croatian-facing operator carrying Pragmatic, Play'n GO, Hacksaw, Nolimit, Push and the usual aggregator long tail — and the first screen is predictable. You will see a cluster of titles in the 97.0%–97.8% range, and the cluster is smaller than players expect. In a 4,200-title lobby we examined, only 61 games cleared 97.0% RTP, and 34 of those were the same handful of "RTP-boosted" configurations of games that also ship in a lower-RTP variant elsewhere on the same site. The genuinely high-return catalogue is not a long tail. It is a short head, and it is dominated by low-volatility, low-max-win titles that most players sort past on their second scroll.

This is the first mechanism behind the 39%. The sort promises a ranked list; what it delivers is a ranked list where positions 1 through 40 are games the player has either never heard of or has already decided against. The interesting part of the lobby — the new Nolimit release, the Pragmatic sequel, the Hacksaw title everyone in the Discord is talking about — sits at 96.1% or 96.5% and therefore lands somewhere around position 400. The player sorted to find the best games and found the least exciting ones.

RTP is not stable across the same title

A second problem is that the sort treats RTP as a fixed attribute of a game. In practice it is an attribute of a configuration. The same slot can be deployed at 94.0%, 96.0% and, for a handful of titles, 96.5% or higher, and which version a Croatian player gets depends on the operator, the aggregator contract, and sometimes the jurisdiction. A lobby that sorts by RTP is sorting by whatever number the operator has chosen to display, and operators are not consistent about whether they display the theoretical maximum, the deployed configuration, or a provider-supplied figure that no longer matches what is running.

For a Croatian player comparing two brands — say a domestically licensed operator against a .com site accepting the same deposit methods — the RTP sort produces two different rankings for what looks like the same catalogue. That is disorienting in a way that a volatility filter or a provider filter is not, because those attributes don't silently change value between sites. When the sort's central promise is "these are the highest-return games," and the number behind the sort is contested, the sort loses authority fast. The 71-second median reversal is consistent with players noticing that the top of the list doesn't match what they expected from the same game on another site.

The bonus conflict

The third mechanism is the one operators tend to underweight. Most Croatian-facing welcome offers and reloads carry wagering that is restricted to a subset of games, and the restricted subset is rarely the high-RTP head of the table. A player who activates a bonus, sorts by RTP, and starts playing the 97.6% low-volatility title at the top of the list may be contributing reduced or zero wagering progress. The sort has, in effect, walked them away from the games that clear the bonus.

This is a design failure, not a player failure. The lobby knows which games are bonus-eligible; the RTP filter does not. A player who has an active bonus and applies a pure RTP sort is being shown a ranking that is, for their current session, actively wrong. The fact that 39% of them figure this out within a minute and a half is a testament to how quickly players learn the rules of a specific lobby — and an indictment of a filter that doesn't account for the state the player is in when they use it.

Why the number is 39% and not 60%

The reversal rate is not uniform across the sample, and the variation is more informative than the headline. Three patterns stand out.

Bonus-active sessions reverse far more often. Among sessions where a bonus was active at the time of the sort, the reversal rate was 58%. Among sessions with no active bonus, it was 27%. The bonus conflict described above is doing most of the work. Players without a bonus are sorting out of curiosity or to find a specific game; players with a bonus are sorting to optimise, discover the optimisation is broken, and revert.

Mobile reverses more than desktop. Mobile sessions reversed at 44%, desktop at 31%. The likely explanation is screen real estate: on a phone, the first screen of a high-to-low sort shows perhaps six titles, and if none of them are recognisable, the cost of scrolling to find something playable is higher than the cost of tapping back to default. On desktop, a player can see 24 tiles at once and is more likely to spot a familiar title further down the ranked list and stay with the sort.

**Experienced players reverse less — but they