Sweet Bonanza: 398,892 Real Spins — The Largest Independent Dataset

398,892 Spins — What Real Data Looks Like

398,892 spins. Not a simulation. Not estimated. Real game engine output, captured directly from Pragmatic Play's game server API by SlotAI analyst Aleks N in 2026.

Most Sweet Bonanza analysis online starts from the certified 96.48% RTP figure and extrapolates. This dataset doesn't. It records every individual spin result — bet, win, cascade depth, free spin status, balance after — across 393,491 paid spins. That's the equivalent of playing 50 spins a day for just over 21 years without a break.

The headline result is 96.9% dataset RTP, converging to within 0.42 percentage points of the certified figure. But the headline number is almost the least interesting thing here.

The Mechanical Constant: 57.2% Zero-Win Rate

Before any analysis of RTP variance or bonus frequency, one number stands out for its stability across the full dataset.

In 398,892 total spins, exactly 57.2% returned zero wins.

This is not a sample artefact. Across every window of play examined in this dataset, the zero-win rate never deviated meaningfully from this figure. The range across different 10,000-spin windows sits within 57.2% to 57.8% — a span of 0.6 percentage points despite large swings in outcome.

The frequency of winning anything does not change based on how a session is performing. What changes is the size of wins when they occur — and critically, whether the bonus round triggers and how richly it pays.

The 57.2% zero-win rate is a structural property of Sweet Bonanza's base game math, not a temporary state. A player who believes they are "running hot" because they just won three consecutive spins is observing normal variation inside a system that returns zero on 57 in every 100 spins regardless of recent history.

RTP Convergence: How 96.9% Emerges from Variance

The full 393,491-spin result of 96.9% against the certified 96.48% is the most reliable external validation of Pragmatic Play's published math available from a non-certified source. The 0.42-point convergence gap falls within expected statistical noise at this sample size.

At realistic session lengths of 100–300 spins, the variance around the certified RTP figure is considerably wider than what any individual 10,000-spin window captures.

Bonus Frequency: The Biggest Driver of Session Outcome

If the zero-win rate is the mechanical constant, bonus frequency is the primary lever of session RTP variance. Every significant win in this dataset came from free spins. The base game's tumble mechanic produces small wins frequently, but the multiplier stacking in free spins is where large returns originate.

The long-run average across the full dataset: 899 bonus triggers across 393,491 paid spins — one every 437.7 spins.

Within the full dataset, bonus frequency varied significantly across different windows of play — some 10,000-spin segments received a trigger every 300 spins while others went 600+ spins between triggers. That variance, not base game behaviour, is what separates profitable sessions from losing ones.

The 1,193x Hit: What the Largest Win in 398k Spins Looks Like

Across 398,889 recorded spins, the single largest win was 1,192.5x the stake, recorded at spin 96,323 of the full dataset. At the £2 stake used throughout, this produced £2,385.

Wins of 100x or greater represent 0.1% of all spins across the complete dataset — roughly 1 in every 1,000 spins. Wins above 50x account for 0.2%. The theoretical maximum of 21,100x was not observed in this dataset.

The frequency figure has a practical translation: at 50 paid spins per session, a player would need approximately 700 sessions before statistically expecting to encounter a single 100x-or-greater spin.

What the Data Reveals

A player receiving bonuses at half the dataset average frequency will see their session RTP collapse toward 84% or lower. A player receiving bonuses at double the average frequency will approach or exceed 110%. Both players experience the same 57.2% zero-win base game floor. The difference accumulates entirely in how often and how richly the free-spin multiplier mechanic fires.

The longest losing streak in the full dataset — 21 consecutive paid spins with zero wins — occurred at spins 107,998 to 108,019.

Bankroll Math from Real Numbers

The dataset supports precise calculations based on recorded behaviour, not theoretical models.

Expected cost to first bonus at dataset average frequency: 437.7 spins × £2/spin = £875 expected cost

Range observed across the full dataset:

• In high-frequency bonus windows: ~£600 expected to first trigger
• In low-frequency bonus windows: ~£1,250 expected to first trigger
• Dataset average: £875

Expected loss per 500-spin session at dataset RTP (96.9%): 500 × £2 × (1 − 0.969) = £31.00 expected loss

Expected loss per 500-spin session at certified RTP (96.48%): 500 × £2 × (1 − 0.9648) = £35.20 expected loss

These are statistical expectations, not per-session guarantees. At 500 spins, the deviation range is wide.

Dataset Integrity

The complete dataset stats file SHA-256 hash, for independent verification:

c40219225fe0c103350def6417a0f02ee7d842cb35f33d39db9e7b87a5ba2e41

Data captured by Aleks N, SlotAI research lead. Method: direct Pragmatic Play game server API capture via authenticated WebSocket session. All 398,892 spins recorded in real-time at £2/spin. No spin results excluded or modified. Dataset trimmed at 398,889 spins to remove three trailing zero-balance logging artefacts. Six interior zero-balance recording artefacts forward-filled at spins 395,697–395,699 and 395,804–395,806.