Forbidden Alchemy: transformed Wilds and verified RTP evidence

Forbidden Alchemy is PG Soft ID 1981965, code forbid-alchemy. Its English PDF is version 1.0.r3, updated 29 January 2026. The five-reel, three-row game uses 20 lines and can randomly convert selected high-paying symbols into Wilds while building additive multipliers.

There is a first-party discrepancy: the versioned PDF states 96.76% RTP, while the API catalogue checked on 8 October 2026 exposes 96.68% as maximum RTP. Neither value should silently replace the other; live help must identify the active profile.

Evidence table

Field	Value	Source scope
Identity	1981965 · forbid-alchemy	Catalogue/API
Rules version	1.0.r3 · 29 January 2026	English PDF
Layout	5 reels × 3 rows · 20 lines	Main game
PDF RTP	96.76%	Versioned rules
API maximum RTP	96.68%	Catalogue on 8 October
PDF split	68.38% main · 28.38% Free Spins	Simulation summary
Overall hit rate	32.26%	PDF summary
Volatility	High	Supplier label
Maximum win	8,000× bet	Formal cap
Euro defaults	€0.20–€100	Reference only

The conflict may reflect a later catalogue profile, but that is an inference rather than a confirmed change log. Record the exact in-game percentage and build instead of averaging 96.76 and 96.68.

Wild Transformation Feature

The feature can trigger randomly during any spin. It selects one or more high-paying types from Coral, Mandrake, Claw and Skull at the top of the reels. Every matching selected symbol visible on the board transforms into a Wild.

Each transformed instance increases the corresponding top multiplier by one. Before wins are paid, the game adds all displayed top multipliers and applies that sum to qualifying awards.

The arithmetic is additive across selected types. If two counters display 3 and 4, the relevant factor is 7×, not 12×. A transformed symbol both substitutes as Wild and contributes to its type’s counter.

Free Spins tiers

Three Scatters award 12 Free Spins, four award 15 and five award 20. During the feature, increases created by Wild Transformation remain until Free Spins end.

Two Scatters during Free Spins add three more spins, and the feature can retrigger. Persistent counters mean an early transformation may influence later qualifying wins, but no transformation or particular multiplier is guaranteed.

The distinction between selection and persistence matters: only selected high-symbol families transform on that event, while already increased top counters retain their values through the feature.

RTP, hit rate and cap

The PDF’s 96.76% model assigns 68.38 percentage points to the main game and 28.38 to Free Spins. Its overall hit rate is 32.26%, split into 31.84% main and 0.41% feature.

High volatility and an 8,000× maximum do not predict session results. If the accumulated award reaches the cap, the spin ends even when remaining Free Spins or persistent counters are visible.

The API’s 96.68% field is a maximum catalogue claim, not proof that every operator uses that profile. The only defensible account statement comes from live help.

Feature Buy boundary

Version 1.0.r3 documents a menu that buys Free Spins at the displayed price. It does not establish legality, availability or price in New Zealand. Never infer the option from a global PDF alone.

If enabled, record the full purchase cost, purchased feature RTP if separately disclosed and bonus restrictions. A buy button changes entry, not the underlying rule that outcomes remain random.

Mobile audit

The four top counters must remain visible above the five reels on a phone. Verify that selection highlights identify Coral, Mandrake, Claw and Skull without colour alone and that the summed multiplier appears in history.

Enlarge text, open paytable and confirm all 20 line diagrams remain accessible. During Free Spins, test whether persistent counters and remaining spins survive an app switch.

Reconstructing transformation arithmetic

Before the event, list every high symbol by type and position. After selection, count converted symbols for each family, update its counter and add the displayed values.

Copy the raw line award and applied sum from history. Do not multiply one type after another. Repeat the worksheet across Free Spins so retained counters are distinguished from new transformations.

Unconfirmed deployment facts

The studio sources do not prove NZ access, 22Bet availability, NZD staking, a local demo or bonus contribution. They also do not resolve which RTP value an operator has loaded.

The correct check matches the name, ID, code, version and help percentage. If the deployed game lacks a visible RTP disclosure, the conflict remains unresolved and should be stated plainly.

How to document the RTP conflict

Start with a screenshot of the live rules header showing game name, version and percentage together. A percentage without identity could belong to another profile or title. Add the operator domain, account territory and UTC check time to the internal record.

If the screen displays 96.76%, it aligns with PDF 1.0.r3. If it displays 96.68%, it aligns with the catalogue maximum observed on 8 October. A third value should be reported as a separate deployment, not treated as a typo.

Do not call the higher value “default” merely because it appears in a PDF. The catalogue and versioned manual are both first-party sources, but neither can override the authenticated help for a specific account.

Worked transformation record

Imagine the feature selects Coral and Claw. Count every visible Coral transformed into Wild and add that number to its top counter; repeat for Claw. Leave Mandrake and Skull unchanged because they were not selected.

Next add the resulting top counters to obtain the factor applied before payment. Store the raw line award and final multiplied credit. If the animation shows one counter rising several steps, the before-and-after screenshots should reconcile with the number of transformed instances.

The example explains procedure, not a promised outcome. The selected families, number of matching symbols and resulting wins remain random.

Persistent counters during Free Spins

Create a four-column ledger for Coral, Mandrake, Claw and Skull. Write the starting value at feature entry, then add new increments spin by spin. Do not reset the ledger after a non-winning spin because version 1.0.r3 says increases remain until the feature ends.

Two Scatters add three spins but do not themselves increase a top counter. Record the retrigger in a separate remaining-spins column. This prevents two independent events from being merged into the multiplier arithmetic.

At settlement, compare the final counters with the last winning calculation. Visible accumulated values have no standalone cash worth when no line win occurs.

Twenty-line result review

The base game uses fixed lines, not ways or cluster pays. Identify every winning line with its path and symbol family before applying the feature factor. Scatter handling and Wild substitution must follow their separate rules.

On mobile, export or screenshot the complete line diagram if support allows. A cropped centre reel cannot establish whether the first or fifth reel completed a path.

Responsible handling of Feature Buy

A purchased feature can end below its cost. Persistent multipliers and 20 starting spins from five Scatters do not guarantee recovery. Set a cash limit before opening the menu and never increase the next purchase because the previous transformation was small.

The page deliberately does not publish a universal price because the operator screen is the only relevant amount. If that price changes, record the new figure with currency and date rather than replacing historical evidence without explanation.

Session-history evidence pack

For a disputed transformation, save five items: the game/version header, complete 15-position board before selection, selected families above the reels, counter values after conversion and final round calculation. Include the parent round ID and total wager.

This pack lets support distinguish a missed visual increment from a settlement problem. A final wallet screenshot alone cannot show which symbols were selected or how the additive factor was built.

If the session disconnects during Free Spins, wait for server history to settle before repeating any action. Record remaining spins and all four persistent counters on reconnection. Do not launch another feature to “test” the discrepancy, because that creates an unrelated round and additional financial exposure.

Source hierarchy for future updates

The active in-game help is strongest for one operator account. The versioned PDF is strongest for the documented global build, and the API is strongest for current catalogue identity and maximum fields. A third-party review may help discover a change but should not supply the final numerical claim.

When sources conflict, publish scope and date beside each value. That approach is more accurate than selecting the most recent-looking page when its deployment context is unknown.

Final assessment

Forbidden Alchemy’s distinctive mechanism is a random symbol-family conversion whose counters add before settlement and persist through Free Spins. The provider sources document that order well.

The unresolved 96.76% versus 96.68% difference is material, not cosmetic. This page preserves both first-party records and directs the reader to the only source that can settle an account-specific value: the live rules.

Official and related sources

• PG Soft Forbidden Alchemy game information PDF
• PG Soft official catalogue
• PG Soft catalogue guide
• Responsible gambling guide
• Corrections policy

How to verify the version you can actually open

A provider page proves that a title exists and describes the supplier's intended rules. It does not prove that a casino accepts players in New Zealand, that the title is enabled for a particular account, or that the highest published RTP is the version in the lobby. Treat availability, operator eligibility and product configuration as three separate checks.

Start with identity, then configuration

Open the game only through the authenticated casino lobby. The information or help panel should identify PG Soft, the exact title and the active mathematical return. Compare the reel or grid layout, paylines or cluster rules, named features and symbol behaviour with the official product material. Similar artwork or a matching title is not enough when the help file is missing or inconsistent.

When the supplier publishes several RTP values, report all of them as configurations. Do not assume the highest option is deployed. RTP is a theoretical long-run measure across a very large number of rounds; it is not a refund rate for one player and cannot predict the next result. Volatility is different: it describes how unevenly results are distributed on the provider's scale.

Read stake controls as money, not interface decoration

Provider sheets often quote a reference range in euros. That range does not establish New Zealand-dollar support or the limits in a specific account. Before a paid round, identify the total amount that will leave the balance. If the interface separates denomination, coins and lines, verify the combined total. If Extra Bet is active, calculate its full cost before pressing spin.

Feature-buy prices are multiples of the current bet, not promised returns. A 100x purchase at a $1 stake costs $100; at a $2 stake it costs $200. Availability can vary by market and some mathematical configurations may not support the control. If the live help does not explain the price, delivered feature and active RTP, leave the option unused.

Keep simulation figures in their proper scope

A game sheet may publish a maximum result observed in a large simulation. That is not automatically the contractual maximum payout and never describes a likely session. Preserve the provider's wording, sample size and configuration. Check the live rules for a binding cap and the operator terms for any separate payout limit. Do not turn an extreme result into a promotional headline.

Audit a disputed round

If an animation freezes or a connection drops, do not repeat the wager immediately. Reopen the game through the authenticated lobby and inspect round history. Record the time, total stake, balance movement and round identifier. The result can be settled by the gaming system even when the device never displays the final animation. Send evidence through verified operator support; never disclose a password, recovery phrase or one-time code.

Read the paytable as a calculation document

The paytable should explain what creates a win, which symbols substitute, where special symbols can land and how multiple awards combine. For line games, confirm direction and active paylines. For cluster games, confirm adjacency, minimum group size and the order of collection. For respins, identify what remains locked, what resets the counter and what ends the feature. A graphic effect can illustrate those rules but cannot replace them.

Separate addition from multiplication. Several prize values may be added before a multiplier applies, while reel-specific multipliers can sometimes multiply together only when each reel participates in the same win. Reconstruct one low-stake result from game history. If the arithmetic cannot be matched to the help file, stop and ask the operator rather than guessing.

Do not confuse hit frequency with a profitable round

Provider sheets can publish hit frequency or category probabilities. A hit simply means a round returned a qualifying amount; it can be smaller than the stake. “Big”, “Mega” or “Epic” thresholds are provider reporting categories, not promises. They are useful for comparing mathematical versions only when definitions and samples are the same.

Random outcomes do not become due after a long gap. A feature that appeared repeatedly in a demo or marketing clip has no obligation to repeat in a funded session. Previous rounds do not create a debt from the next random result.

Confirm market and device behaviour independently

A title can exist in the supplier catalogue while an operator disables it for a territory, currency, device or bonus. Language localisation is not evidence of legal availability. Confirm the operator's identity, accepted countries and current lobby separately from the provider review.

On mobile, verify that balance, total bet, feature state and help remain visible without overlap. On desktop, check that browser zoom does not hide a purchase control or change the apparent denomination. Mathematical configuration should not change merely because the screen changes; if the help shows another RTP, treat it as a different deployment and record it.

Keep a compact evidence record

For any fact used in a review, save the title, provider, official URL, document date when available, access date and exact territorial scope. For a live session, add the operator domain, Game ID, RTP, currency and a non-sensitive screenshot of the help panel. This makes later corrections possible when a supplier updates a sheet or an operator changes configuration.

Missing information should remain missing. Do not estimate an RTP from results, convert a euro reference range into an operator's NZD limit, or treat another casino's configuration as evidence. A transparent “not confirmed” is more useful than a precise-looking number without the correct source.

Use limits that survive the feature design

Collection meters, expanding grids, sticky symbols and near misses can make a session feel unfinished. Unless the rules explicitly preserve progress, the visible state has no value after the round or feature ends. Set a money limit and time limit before launch, count feature buys immediately, and stop at the limit even when a meter appears close to an upgrade.

No random game is an income plan. Do not borrow, use essential funds or raise the stake to recover a loss. If play stops feeling recreational, use timeout, deposit-limit or self-exclusion tools and consult the site's responsible gambling guide.