How we test

Where the facts come from

• The game catalogue: release date, reels, bet range and features, collected from the studios.
• Hacksaw's own game page: the RTP versions, max win, max win odds, hit frequency and the feature descriptions.
• The game's rules and buy menu: how each feature and bonus works, and what each buy costs.
• Our recordings: everything that happened when we played the game, described below.

Figures from the studio's page are labelled as studio figures on the fact sheet. We publish them as the studio states them; we don't verify them ourselves. Where an editor has also checked a figure in the game's own info screen, the page says so with the date.

How we record a game

An automated browser plays the official free demo with play money, in turbo, at the game's default stake (2.00 in our current recordings). It records every session as a video with sound. For each game we record a test kit:

• 100 base-game spins;
• 100 BonusHunt FeatureSpins, when the game has them;
• 10 spins of each other FeatureSpins mode;
• each bonus buy three times, one recording per buy.

Bonuses play out without optional gambles. Every result is read from the game server's responses, not from the screen, and the videos on the page are those recordings.

What we measure

• For each way of playing: how much was paid back, the share of spins that won (hit rate), the bonuses we saw, how much we staked per bonus, the biggest win and the longest losing streak.
• For each buy: every result as a multiple of the bet, against the price of the buy.

What small samples can't tell you

A few hundred spins and three buys per bonus show how a game plays: how often it pays, how the bonuses look and how far the balance swings. They say nothing reliable about its RTP, which only shows over millions of rounds. One lucky or unlucky bonus can move a result a long way, so every result on the site shows its sample size, and we never present our numbers as the game's return.

Previews

Some games get a page before release, recorded and written from the studio's pre-release demo. These pages say Preview and First look, and their score is labelled as a pre-release score, because the game can still change before release. On release day we record the game again, rewrite the preview into a full review and score it again.

Games we can't record

When a game's demo can't be recorded, it gets a review only if the studio's game page or the game's rules describe it, and we write it from those alone. Those pages make no claims about tests and get no score.

How reviews are written

Each review is drafted by an AI model from an evidence pack: the numbers from our recordings, the studio's text, the game's rules and frames from the videos. Automated checks then go through every draft:

• every number in the text must be in the evidence;
• the RTP isn't repeated through the text;
• no stock phrases;
• the tense fits the release date;
• the overlap with our other reviews stays low.

An editor reads and approves every review before it goes live. See the editorial policy for how we use AI.

Rating

Reviewed games get a score out of 10 from one formula, the same for every game. The current version is ss-1. Each part is scored from 0 to 10 and weighted:

• Small samples count for less. A part from our tests is pulled toward 5 by n/(n+k), with k = 15 paid rounds (a buy is one round, and so is each FeatureSpins spin), 400 spins or 100 base spins. Six buys and 10 FeatureSpins rounds count for 52%, 200 spins for 33%.
• When a part has no data, its weight is shared out over the other parts.
• There is no score without an official RTP and at least one test result.
• Previews are scored from the pre-release demo, labelled as a pre-release score, and scored again on release day.
• Games we can't record get no score.

Each review shows its parts, their inputs and how much each counted, next to the score.

Version log

• ss-1, 3 November 2026: first version.

Corrections

A page updates as soon as its data or text changes, and its "Updated" date moves with it.