Hot Hot Fruit
1,000 Spins Test

How long we waited for each feature

Every natural trigger below is recomputed from the recorded rounds, not from memory - the same reproducible script that builds this table also writes the analysis file and chart. One interval = the paid rounds from the previous trigger up to and including the triggering spin; the first interval counts from the start of the sample. A trigger round settles only after its six free games play out, so each wall-clock span also contains that trigger's own feature run (34–163 seconds, measured from the raw exchanges).

• Completed sample gaps, across collection sessions. Average: 86.4 paid rounds (median 45). The eleven completed gaps ranged from 4 rounds (rounds 815→819) to 477 (105→582). In our collection pacing that meant an average of 18.3 minutes of wall time (median 11.7) - our scripted pace with deliberate delays, not what a normal player would experience, and not the game's expected frequency.
• Session breaks are real time. Rows marked "(crossed)" span session expiries, cooldowns and restarts - the 477-round gap includes about 46 minutes of plain downtime between collection runs. Counters reset with every fresh 2,000-credit session, so these are observed aggregates, not one continuous sitting.
• After the last trigger the sample went quiet. Rounds 951–1,000 are 50 paid rounds (7.7 min) with no trigger. That tail is reported separately and excluded from the average above - the sample ended before a twelfth feature.
• Two different denominators, both descriptive. The overall sample density is 11 triggers in 1000 rounds - about one per 90.9 rounds - which is not the same figure as the 86.4-round average of completed waits. Neither is a probability: a long gap does not bring a feature closer.

More charts from the same recorded sample

How we recorded it (methods, briefly)

1. Ordinary public demo sessions through the official client interface - one session at a time, normal click pacing, no injected responses, no forced outcomes and no modified files.
2. Every request/response pair was saved exactly as received; each record is hash-checksummed. The validator re-verified all 1,066 exchanges, every balance transition, and that no round was double-counted before this page's numbers were written.
3. Ten sessions contributed completed rounds. An eleventh attempt recorded no rounds and is not represented in the CSV. A fresh demo session starts at 2,000 credits. We treat each session separately rather than pretending it is one continuous bankroll.
4. Incomplete rounds would have been quarantined and excluded; in this sample none occurred.

Download the recorded round data

Reproduce the calculations: download the analysis module and CSV into the same directory, then run node --input-type=module -e "import {readFileSync} from 'node:fs'; import {parseCsv,analyze} from './analyze.mjs'; console.log(analyze(parseCsv(readFileSync('rounds-1000.csv','utf8'))));". It calculates the same payout groups, sessions, streaks, concentration and bonus summaries shown above.

Every figure above comes from the same whitelisted export you can check yourself: the 1,000 settled rounds with no tokens, account identifiers, backend IDs or raw request envelopes - those stay private. Cash fields are demo credits, not rand.

The manifest holds the full SHA-256 of each file (CSV 9d2d6d8f…) plus the private analysis dataset's checksum 44aff0848d8ef74a… - separate values: the export is a derived public view, not the raw record. The export regenerates from the audited dataset by script; its arithmetic (base + feature = total for every row; 11 features; 66 free games; totals 7,500 / 8,815 / 3,475 / 5,340) is re-verified on every run.

What this evidence cannot tell you

• No RTP estimate. 1,000 rounds is far too small; the observed 117.5% is luck, not a property of the game.
• No "how often features hit" rate. 11 triggers in 1,000 rounds (1.1%) is a sample fact, not a probability.
• No funded-play proof. We never deposited, and demo results do not establish what a rand account would pay out.
• No strategy. Nothing in the data supports timing, pattern or stake-progression systems.

Dataset checksum (SHA-256, first 16): 44aff0848d8ef74a - you can use it to confirm that any copy of our round record is the one this page describes. Figure captions on this site say where each image comes from: frames marked “captured by us” were taken while we played the official Habanero demo; other images are provider-supplied screenshots or artwork, labelled as such. All numbers on this page come from the recorded exchanges described above. For the rules-side evidence, see the feature guide, which pairs the provider's rule panels with what we observed.