The gate, in exact terms
- Three reviewers, thirds of a star. Each reviewer gives 0 to 3 stars in steps of one third. The three scores add up to a total out of 9.
- 6 points to be published. That is a 2-star average. The star shown is rounded, but publication is decided on the exact points: a game can display 2 stars and still fall short.
- Three sections are closed to 2 stars. New Releases, Slots and Burst Games list 3-star games only. A 2-star game is published without them, and reaches Stake Exclusives only if players carry it there.
- No feedback. Reviewers do not explain a rating. Engine sends creators to its Discord to ask other creators.
- Three days locked. Below the threshold, the approval thread is closed for 3 days before the game can be submitted again.
The math is a gate of its own
The submission form blocks a game before any reviewer opens it if the math fails one of these checks:
- Every mode has an RTP between 96.0% and 96.7%. Older games can sit higher: the most-wagered game in the 14 November ranking shows 98.00%.
- The cheapest mode costs 1.0× the bet, and its standard deviation is at least 0.6.
- No mode pays more than 1,000,000× or costs more than 50,000×.
Two more points are judged by a person. The maximum win has to be realistically obtainable, which Engine puts at more often than about 1 round in 10,000,000. And a game where 90,000 rounds out of 100,000 pay nothing can be rejected for that alone.
A second set of risk checks does not block the game but shrinks it: each failed class lowers the largest bet the game may accept, from $500,000 down to $10,000. A very high maximum win is paid for in bet size.
Why games get sent back
Engine names four reasons. Two are about art, two are about game design.
| Engine's reason | What Engine describes | What it changes in production |
|---|---|---|
| Shallow gameplay | Players place one or two bets and leave. | A feature the base game builds towards, visible from the first spin. |
| Generic AI-generated assets | Standard fonts, gradients, emoji icons, border effects. | Lettering, symbols and frames drawn for the game, in place of default fonts and effects. |
| Inconsistent or weak visuals | Mismatched art styles, poor animation. | One art direction signed off on a few pieces before the full set is drawn. |
| No engaging features | No bonus mode, no extra mechanic. | Bonus modes are specified in the math model before the art starts. |
What reviewers actually test
Reviewers play every submission. Engine publishes what they try:
- The game with bet levels rewritten by the server, with a round interrupted halfway, with a zero balance and with a broken server address.
- Four screen sizes, and replay in a small pop-out window.
- Space bar to bet, no page scrolling, double-tap zoom disabled.
- A confirmation before autoplay and before any mode that costs more than 2× the bet.
- Replay of a past round with no wallet call, showing the real cost of that round.
- Game rules that state RTP and maximum win per mode, and match the math files.
- Social mode for sweepstakes markets, where restricted words have to disappear from everything a player can see, text drawn into the art included.
Before a game is submitted, you receive this list filled in: each test, the result, and what was changed. Engine is clear that passing it is not an approval: originality, quality and how the game feels are judged by people.
Where the third star is decided
Engine's words for a 3-star game are "studio-quality games showing exceptional creativity, uniqueness and attention to detail". Behind that sits a short list: correct rendering and clean sound on a range of devices, a bundle that loads fast, and cohesive, polished art and animation.
A studio with two 3-star games on Engine describes the gap between two and three stars as "almost always polish": win presentation, anticipation, load times. Our production rules follow from that.
- Celebrate in proportion. No celebration under 2× the bet, and the full sequence starts around 20×. Engine's math SDK sorts every win into ten levels by default; the animations are mapped to those levels.
- Anticipation comes from the math. On Engine, each reveal event carries an anticipation value per reel. The front end turns it into slower, later reel stops once the bonus symbols start landing.
- Symbols read while the reel moves. Low symbols in cool colours, high symbols in warm ones, each with a silhouette that holds at phone size.
- Animation has a budget. Every animated symbol costs frames on a mid-range phone and bytes at load, and Engine lists load time among its 3-star criteria. What moves is decided before it is drawn.
- Sound that does not tire. A player hears the spin loop hundreds of times. It has to stay out of the way, and the mute has to mute everything.
After approval: what the public ranking shows
Engine publishes a live ranking of every game. From its figures on 14 November 2026, covering the previous 30 days, we counted:
| Measure | Value |
|---|---|
| Games ranked | About 7,800, from more than 2,000 studios |
| New games in 30 days | 2,479 |
| Share of all wagers taken by the top 10 games | 43% |
| Share taken by the top 100 games | 73% |
| Median game | 84 players and about $5,600 wagered in 30 days |
| Top 100 games released less than 30 days earlier | 33 |
| Median maximum win among the top 20 games | About 47,900× |
Wagered money is not revenue: the creator's share is a fraction of what the game keeps. Half of the games had fewer than 84 players in the month, while a third of the top 100 were new releases.
Inside New Releases, games are ordered by the ranking of the team that published them over the last 30 days, re-arranged every Friday. A team with no ranking yet starts low in that list.
Engine's own advice is to plan promotion alongside the game: partnerships with streamers that include a revenue share, and followers on Stake, where it reports launch emails opened by about 30% of a studio's followers. One choice has to be made before the build: games released on the wider Engine network are not eligible for Stake's promotional opportunities.
What we do not promise
A rating is the judgement of three people who give no reasons. Nobody can guarantee 3 stars. The math gate, the test list and the four reasons above can be checked before submission; the rest is the reviewers' call.
We have not released a game on Engine yet. Our first one is in production, and we will publish its rating here, whatever it is.
Questions creators ask
Can a 2-star game be published?
Yes, with 6 points out of 9 or more. It goes live but is not listed in New Releases, Slots or Burst Games.
How long does the review take?
Engine's documentation gives no estimate. Reviews are prioritised by star rating, and a game below the threshold waits 3 days before it can be submitted again.
Can the math be changed after release?
No. The math is locked once the game is live, which is why it is checked first.
Sources: Engine documentation (game quality rankings, math verification, AI review, promote your game) and the Engine Insights public ranking, read on 13 and 14 November 2026; Engine's public math and web SDK repositories for win levels and anticipation. Engine can change its rules at any time.
Building for Stake Engine?
Tell us what you have and what is missing. We answer in writing, with a quote.
Get a quote