Corrections

What went wrong. Under “Is the official figure true?”, the label correctly said “Does not match”, but the sentence below it was fixed text that always said the official figure sat inside our measured range. On Fortune Tiger it read: we measure 144.5% (between 140.8% and 148.3%), the official figure sits inside that range, next to an official 96.81%.

What it affected. The wording on 35 of the 84 game pages that can be checked against a single official figure. The numbers, the range and the label were right; only that sentence was wrong.

What we did. The sentence now follows the result in all three languages, and an automatic build check fails if it ever goes back to fixed text.

Fortune Tiger

What went wrong. Games whose measured return falls more than four of our margins of error below the published RTP are flagged automatically. The check only compared against a single published figure, so games whose provider publishes a range (Habanero, Hacksaw: for example 92–98%) were never flagged, however far off they were. The warning text on flagged pages also said “no free-spin round recorded” and “every other game sits within two margins”, which was written for one game and was not true for every game the rule later flagged.

What it affected. Wild Trucks measured 74.7% against a published 92–98% without being flagged (see the next entry). The warning on Chinese New Year 2 said no free spins had been recorded, when six had.

What we did. The check now compares with the lowest published figure, the most lenient comparison for us. The warning is split: the free-spin sentence appears only on games that have never triggered one, and the “within two” sentence is gone.

Wild Trucks

What went wrong. Wild Trucks has a meter on each reel that builds up over spins and is kept per game session. To share time between 14 games, our collector opened a new session every 30 spins, so the meter always restarted from zero. The return rose with the position inside a session: 62% for spins 1–10, 75% for 11–20, 88% for 21–30. Two games without such a meter stayed flat.

What it affected. The Wild Trucks page showed 74.7% over 206,686 spins against a published 92–98%, and its session ranges appeared in the RTP guide. The other 11 Habanero games keep no state between spins and were unaffected. JILI also switches session every 30 spins; the same test showed no rise there.

What we did. Wild Trucks now runs on one continuous session. The 206,686 earlier spins are kept in an archive but removed from every statistic, and counting restarted from spin 1. Its page returns once it passes 20,000 spins.

Wild Trucks

What went wrong. Collection for these three providers paused on 8 October, but their game cards kept saying the spins were queued and still rising.

What it affected. Twelve days of game cards for those three providers. The spin counts themselves were right; they just were not moving.

What we did. The cards now say plainly which lines are paused or running slowly, and how many spins a day to expect.

What went wrong. JILI sends a base spin and its whole free-spin round in a single response. Our code looked for a separate free-spin step, so in about 75,000 spins it never recognised one.

What it affected. Free-spin frequency and “spins since the last free spins” on JILI games. Return, hit rate and the biggest win were unaffected: the winnings were always counted, only not labelled as free spins.

What we did. The rounds inside each response are now read one by one. Free-spin figures for JILI are counted only from the moment of the fix (7 October 2026, 10:50 UTC); earlier spins do not enter them.

What went wrong. After 49,840 spins Joker King measured 57.0% ± 3.9 against a published 96.8%, and it had never triggered a free-spin round, while ten other NextSpin games trigger one every 71 to 242 spins. The likeliest explanation is that we are not capturing its free-spin rounds, which usually make up a large share of the return.

What it affected. It was briefly the homepage headline card for “the biggest gap between measured and official”.

What we did. It is flagged automatically, kept out of every measured-versus-official comparison and out of llms.txt, and its page carries a warning not to quote the return. With over a million spins it still measures about 57%. We have not found the cause.

Joker King

What went wrong. Games that had never triggered free spins, because they have no such feature or because we had not captured it, ranked at the top. Golden Fa showed “67,860 spins without free spins”, which was simply its total spin count.

What it affected. The dry-streak rankings and the examples that quoted them.

What we did. Games with zero triggers are kept out of every “longest without” ranking.

What went wrong. The methodology said a game gets its own page after 20,000 spins and 384 free-spin triggers. In practice the page is published at 20,000 spins; the free-spin frequency is printed only once there are 384 triggers.

What it affected. At the time 25 of the 28 published game pages did not meet the rule as written, on the page a sceptical reader checks first.

What we did. The methodology text now describes the rule as it actually works.

What went wrong. One family of Pragmatic Play slots marks “the cascade isn’t finished” with a field our code did not know, so each cascade step was treated as the end of a spin and its winnings were counted again. One game briefly measured a 141% return.

What it affected. The return on the affected games while the error was live.

What we did. The balance check after every spin caught it: balance change must equal win minus bet, and it did not. The field is now read, so each spin’s winnings are counted once.