RTP recalculates per bet size; the demo mode never says so

Most online slots disclose a single RTP figure, usually somewhere between 94% and 97%, and players treat it as a fixed property of the game. It isn't. On a growing number of titles, the number that governs your expected return moves with your stake, and the demo mode you tested before depositing shows you the version nobody plays for real money. The gap can run to several percentage points — enough to turn a 96.2% game into a 92.1% one at the wrong bet size.

That's not a bug or a rounding quirk. It's a designed feature, and it's disclosed in places most players never look.

How stake-dependent RTP actually works

The mechanism is straightforward once you see it. A slot's payout table is fixed; what changes is which maths model the game loads when you spin. Operators and suppliers configure these models per bet level, usually through a small number of discrete "bet profiles" rather than a smooth curve. Spin at 0.20 a go and you might be on the 96.2% model; spin at 5.00 and you might be on a 94.4% model; spin at the maximum and some titles drop to 92% or below.

A handful of suppliers are openly associated with this. Pragmatic Play's Bet & Get and Bet & Win style mechanics, various Reel Kingdom releases, and several Relax Gaming and Push Gaming titles have shipped with two, three, or four RTP configurations selectable by the operator. The published figure on the game's help screen is often the highest of those configs — the one attached to the lowest bet — while the config actually live on your account is the operator's choice.

The scale is bigger than most players assume. In 2023, testing published by several affiliate labs found variants of the same Pragmatic title ranging from 96.5% down to 88.1% depending on configuration. That's an 8.4-point spread on identical reels, identical volatility, identical graphics. Nothing in the spin animation tells you which one you're on.

Where the disclosure lives

Regulators have pushed back, unevenly. The UK Gambling Commission requires RTP to be displayed within the game itself, and suppliers now typically list all available configurations in the paytable or help file rather than a single headline number. Malta and Gibraltar-licensed operators follow similar rules. Elsewhere — and this covers a large share of the market — the only place the figure appears is a PDF on the supplier's website that you'd have to know existed.

Even where multiple RTPs are listed in the help screen, the game doesn't tell you which one is active for your bet. You get a menu of possibilities and no pointer.

The demo mode problem

Free-play versions of slots exist for one reason: to let you feel the volatility and see the features. They're useful for that. They are not a test of the game you'll be playing for money, and the RTP is the clearest reason why.

Demo builds are typically hard-coded to a single configuration — usually the highest RTP variant, because that's the most flattering version and because demo mode has no operator to configure. When you spin 500 times in a demo at a 0.50 stake and get a feel for how often the bonus lands, you're learning the behaviour of a model that may not be the one your real-money account runs. If the live operator has selected the 92% config and the demo ran at 96.2%, the bonus frequency can be similar while the paytable values behind it are quietly trimmed.

This is the part that catches experienced players, not beginners. Someone who knows to check RTP will check the help screen, see 96.2%, and feel reassured. The number is real. It's just not necessarily theirs.

A concrete example

Take a title with two configs: 96.2% at stakes up to 1.00, 93.5% at stakes above. A player deposits, plays at 0.50, and does fine. They then raise their stake to 2.00 to chase a bonus round, and every spin from that point is on the lower model. Their expected loss per 100 in wagered turnover rises from 3.80 to 6.50. Over a 5,000 turnover session — roughly 2,500 spins at 2.00 — that's a difference in expected loss of 135. Nothing on screen changed except the stake.

Volatility often shifts with the config too, which compounds the effect. The lower-RTP model frequently comes with a slightly different hit frequency, so the session feels different in ways players attribute to "cold" or "hot" streaks rather than a configuration switch.

What this does to bonus terms

Wagering requirements are calculated on turnover, and turnover is multiplied by the RTP of whichever model is running. A 35x wagering requirement on a 100 bonus means 3,500 in spins. At 96.2% RTP, expected loss during clearing is about 133. At 92.1%, it's about 276. Same bonus, same terms, twice the expected cost — purely because the operator's chosen config sits at the low end.

This is where stake-dependent RTP stops being an abstract fairness argument and becomes a bankroll question. If you're clearing a bonus, the sensible move is to find the highest-RTP config available to you and stay at a stake that qualifies for it. On many titles that means staying at or below the first bet tier, which is often 1.00 or even 0.50. Max-bet bonus hunting on a stake-dependent game is close to the worst combination available: top volatility, bottom RTP, and a wagering clock running.

Check the paytable before you clear anything. If it lists multiple RTPs, assume the lowest applies to your stake until you can confirm otherwise, and check whether the operator publishes its active configuration — a few now do, mostly in response to UK scrutiny.

The regulatory direction

The trend is toward disclosure, slowly. The UK's approach — RTP inside the game, all configs listed — is becoming the reference standard, and suppliers who want UK market access have rebuilt their help screens accordingly. The Netherlands has taken a harder line on maximum RTP caps for online slots, which indirectly limits how low a config can go. Other jurisdictions haven't moved.

The unresolved question is whether listing every possible RTP counts as disclosure at all. A player looking at four numbers with no indication of which is live has been told something true and nothing useful. The more meaningful fix — showing the active RTP in the game's UI, updating when the stake changes — is technically trivial and commercially unpopular, because it makes the stake-RTP trade-off visible at the exact moment a player is deciding to raise their bet.

Until that happens, the practical rule is unglamorous: treat the headline RTP as a ceiling, not a fact, and treat the demo as a preview of the reels rather than the maths. If a title lists more than one RTP figure, the interesting question isn't which is highest — it's which one you're on right now, and whether the stake you're about to raise is the thing that changes it.