Advertised: 5% house edge.
Actual: up to 40.13%.

The benchmark uses the advertised edge and the tile odds. We use exact fractions 2/3, 1/2 and 1/3, rather than treating the help page’s shortened decimals as exact probabilities.

The client retains a three-decimal multiplier table and uses shorter labels on screen. That table is compared with a fixed 5% edge above. These labels are not proof of the precise amount credited: the historical records show payouts calculated beyond three decimal places.

1.9 × 1.9 = 3.61. A fixed 5% would instead pay 1.9, 3.8, 7.6 and so on to 972.8×. Benchmark multipliers are rounded to three decimals here; calculations use their full value.

The three-decimal table adds small truncation differences. For example, Medium row 5 gives 22.625%, rounded to 22.63%; the untruncated formula gives 22.62%. The graphs use the untruncated formula, which gives the same curve across all three difficulties.

The retained October client source has a towersHouseEdge: .05 default. Its Towers component then places 1 − houseEdge inside the per-row multiplier. A November 2025 archived client contains the same payout construction.

// Client-store default in the captured October 2026 source
towersHouseEdge: .05

// Towers payout construction, condensed
roundMultiplier = choices / (choices − bombs) × (1 − houseEdge)
clientMultiplier = floor(1000 × roundMultiplier^row) / 1000

// Medium: 1.9 is raised to the row number
row 1: 1.900×    row 2: 3.610×    row 10: 613.106×

The 5% reduction is inside the number being multiplied again. A fixed 5% would be applied once after calculating the fair winning multiplier. Here it is applied at each step. The client code establishes the construction; the test balances and historical records establish the payout evidence.

The house edges and payouts in this report are calculated from the payout formula. Settlement rounds down to the operator’s smallest unit and single wins are capped at $200,000; both can only lower a payout further. Below the cap, every cash-out past row 1 pays less than a fixed 5% would.

Table notes: payouts include the original stake. November contains 27 selected operator-hosted wins dated 8 and 13 November 2024, retrieved 28 October 2026. October includes all 13 cash-outs in the 9 October test; A5’s displayed payout was not retained and is left blank. The table compares recorded payout displays with fixed-5% calculations rounded to cents; multiplier labels are not treated as exact settlement amounts. The P2-3 display and cent-rounded balance observations differ by one cent.

October 2026 check

The October example uses the cash-out display and balance checkpoints at cent precision; exact sub-cent settlement was not captured. The November record displays a 13 November 2024 event date and was retrieved from the operator on 28 October 2026.

The $100 Easy row-2 record is particularly clear. The operator displays $203.06. A literal 2.030× label gives $203.00, while the full formula gives $203.0625. A fixed 5% gives $213.75. Thirteen of the 27 records distinguish these calculations. The pages were retrieved on 28 October 2026; their displayed 2024 dates are not independent 2024 capture timestamps.

The public feed adds 416 distinct Towers records: 330 positive payouts and 86 zero payouts. All 330 positive raw amounts match the full formula floored to the feed’s integer unit. Modes and rows are inferred from those amounts. The feed is a sample, so it isn’t used for total losses or total wagering.

Stake’s game rules give Master mode a 1-in-4 chance per row; its official guide confirms 3.92× first and 256,901.12× ninth multipliers. Shuffle’s official Hard table runs from 1.96× to 501.76× over nine rows, with 1-in-2 row odds. These are published-model comparisons, not nine witnessed settlements.

// A fixed edge cancels out when top multiplier is divided by first
Shuffle Hard   501.76 ÷ 1.96 = 256 = 2^8 across 8 further rows
500 Medium     613.106 ÷ 1.90 ≈ 322.69
// Fixed 5% across its 9 further rows would give 2^9 = 512.

Extra expected loss per dollar

5–15¢

In the scenarios below; assumed effective edge of 10–20%.

Modelled Towers revenue above the benchmark

50–75%

Above the fixed-5% benchmark under those assumptions.

Default scenario

$10.11m

Over three equal-scale years; not a measured historical loss.

On 21 October 2026, FairGambling’s 365-day deposit figure showed $333.5m and Tanzanite’s Year figure showed $327m. We use a rounded $330m annual deposit input for the central scenarios. These are reported tracked crypto deposits, not Towers wagers or losses; the underlying transfers and coverage have not been independently audited. Neither view establishes annual volumes for each historical year.

To estimate scale, we assume how often deposited money is wagered, the casino’s overall hold, and the shares of gross gaming revenue attributable to Originals and Towers. Gross gaming revenue means stakes less payouts. Those game shares are revenue shares, not wagering shares: dividing modelled Towers revenue by its assumed effective edge gives Towers wagering. The table keeps the same annual scale for three years; it does not claim that this volume or game mix actually persisted.

The default uses 17.5% of Originals revenue for Towers, between the two central rows: $3,368,750 per year, or $10,106,250 over three equal years. Every wagering, hold, game-share and effective-edge input is an assumption. The lower row is not a minimum loss, the range is not a confidence interval, and none of these figures is an established refund amount. Actual losses could fall outside this range. Rounding, payout caps, bonuses, rebates and changes in play are excluded. The sampled public bet feed does not establish complete Towers wagering or the target-row mix.

// Scenario calculation. Shares below are shares of gross gaming revenue.
Towers revenue = deposits × wagering/deposit × casino hold
                 × Originals share × Towers share
Towers wagering = Towers revenue ÷ assumed Towers edge
extra expected loss = Towers wagering × (Towers edge − 0.05)
three-year scenario = annual extra expected loss × 3

// A real winning cash-out: $20 Easy bet, row 3, November 2024
fixed-5% benchmark = $20 × 0.95 × 1.5^3 = $64.125 → $64.13
operator’s displayed win = $57.87
rounded benchmark difference = $64.13 − $57.87 = $6.26

The final example is a comparison for one successful cash-out, not an average loss on every $20 attempt. The operator’s displayed win matches the compounded formula: $20 × 1.425^3 = $57.8728125, rounded to $57.87. An account-level review is needed to establish compensation, including applicable settlement rules and any other adjustments.