How we test games: we play the free demo, log the mechanics spin by spin, and rely on published Pragmatic Play figures rather than affiliate marketing claims.
Demo first, spec second, affiliate copy never
Every review here starts the same way: we load the free demo and play it, then check what we saw against the published provider spec and the in-game info panels. We do not rate a game from a press release. When we reviewed Sweet Bonanza 2500, we spun the free demo for a few hundred rounds, logged how often something paid, and watched a handful of feature-buy sessions before we wrote a word. That order matters. The hands-on read tells us how a game feels; the spec tells us whether that feel matches the maths.
Our test is honest but small. A few hundred demo spins is not a certified sample, and we say so on every page that leans on it. You get the pattern we noticed, framed as illustrative, not a lab result. Base-game wins in Sweet Bonanza 2500 mostly landed in a low band in our sample, which lines up with the roughly 42.91% hit frequency and the rare bonus. We report the shape, then defer to the provider's published probabilities for anything a demo log cannot prove.
What we verify
We pin down the numbers that decide whether a game is worth playing: the RTP ladder and which build an operator can run, the max win, the bet range, the free-spins trigger, and how each feature actually behaves. For Sweet Bonanza 2500 that meant confirming the 96.52% / 95.53% / 94.54% RTP ladder, the 25,000x ceiling, the $0.20 to $240 bet range, and that the x2,500 candy bomb is the top multiplier rather than the win cap. We test the demo to confirm the feature triggers do what the spec claims.
What we flag as unpublished
Plenty of numbers are not published, and we refuse to invent them. Exact per-symbol paytable values, the average bonus multiplier, individual feature-buy tier RTPs, and precise bomb-landing frequencies are all unknown for this game, so we say so plainly instead of borrowing a figure from an affiliate page. When sources disagree, such as a stray 96.08% RTP against the confirmed 96.52%, we go with the provider and note the conflict. A hedge is more useful to you than a confident guess.
Turning testing into a 0 to 10 score
Our rating runs on a 0 to 10 scale, set by hand after the testing is done. Four things move it: the maths (RTP and the honesty of the volatility), the feature design and how readable it is in play, the value on offer relative to the risk, and the availability and cost of getting into the bonus. A game can post a strong ceiling and still lose points for a punishing ride or an expensive buy. Sweet Bonanza 2500 lands at 8.0 from us: a real 25,000x ceiling and a proven pay-anywhere tumble, marked down for high variance in play, a bonus at roughly 1 in 450 spins, and a max win that is a 1 in 66,722,269 long shot rather than a plan.
The score is one number from us and one only. Where other reviewers land near 8/10 we mention it unnamed, as corroboration, never as our own figure. We also state, on the pages where it matters, when we would not recommend a game to a particular player, because a rating that never says no is not worth much.
Which sources we trust, in order
Not all sources are equal, and we treat them in a strict order. At the top sits the provider itself: Pragmatic Play's own game page and, above even that, the in-game info screen and RTP panel captured in the live demo. Those are the primary record. When the demo's info panel called Sweet Bonanza 2500's volatility medium while third-party trackers called it high, we led with the panel and noted the tension in plain sight, because the studio's own rating is the source of record.
Below the provider come structured databases and trade press, useful for corroboration and release dates, but never a substitute for the spec. If a database and the provider disagree, the provider wins. At the bottom sits affiliate marketing copy, which we do not treat as fact at all. Precise-sounding numbers on a casino landing page, a "guaranteed" this or a "best odds" that, carry no weight here unless the provider confirms them. We would rather publish a gap than launder a marketing claim into a fact, and we never name rival review sites when we cite outside sentiment.