How these numbers are produced, and where they are wrong

The closed form, and why it is not enough

A wagering requirement means staking the bonus some multiple of times. Multiply that turnover by the house edge and you get the cost of clearing it:

turnover = multiplier × base ÷ game weighting
cost     = turnover × (1 − RTP)
EV       = bonus − cost

This is the calculation nobody publishes, and for most advertised bonuses it produces a large negative number. It is also, on its own,wrong — sometimes in the wrong direction entirely.

It assumes you complete the requirement. Frequently you cannot afford to. On a $500 deposit with a $1,000 bonus and $40,000 of turnover at 96% RTP, the expected cost is $1,600 against a $1,500 bankroll. You go broke first — and going broke caps your loss at the deposit while the bonus absorbs the remainder. Downside bounded, upside open.

So we simulate the bankroll

Each run stakes bet after bet, crediting weighted turnover and resolving each round against a payout model, until either the requirement clears or the balance cannot fund the next bet. That captures the three things the closed form cannot see: the chance of busting first, the bonus forfeited when you do, and the winnings a cashout cap removes.

The result is that the wagering multiplier is rarely the binding term. The cashout cap and the maximum bet are. A cap removes the upper tail that made the gamble worthwhile; a low max bet forces you to grind the full turnover rather than gamble through it. Both are printed far less prominently than the headline multiple.

Two estimators

Averaging simulated outcomes is noisy when payouts are skewed — the mean sits in a tail that rare runs dominate. So the same expected value is also computed by Wald’s identity, which replaces the random payouts with their expectation given the number of rounds played:

EV = bonus − E[rounds] × bet × edge
        − E[forfeited] − E[clipped by cap]

Both are unbiased for the same quantity. With no cashout cap every term above is bounded, so this form is the one published. With a cap, the clipped term becomes heavy-tailed while the direct average is bounded by the cap itself — so the average is published instead. Each figure states which was used, and a 95% interval when it is the noisy one.

How the grade is derived

The grade in the listings table is a pure function of computed expected value as a fraction of the deposit. It is not a rating, an editorial opinion, or anything an operator can influence — every cutoff is published here so you can check any row yourself.

Offers whose terms make a single honest EV impossible — free spins, rakeback, cashback — are ungraded rather than guessed at.

Where this model is wrong

• Payout shape. Each round is modelled as a single win-or-nothing outcome calibrated to the game’s RTP and a volatility setting. Real slots have fatter tails, which understates bust risk somewhat at the same setting.
• Pooled balance. The deposit and bonus are treated as one balance. Operators that segregate them, or release the bonus in increments, will behave differently.
• Cashout interpretation.A cap is applied to winnings above the deposit. Some operators cap total withdrawal instead, which is harsher than modelled here.
• Flat betting. One bet size throughout, no strategy adjustment as the balance moves.
• Expiry ignored. Time limits are listed but not modelled; a requirement that cannot be ground out in the window is effectively worse than shown.

On “zero house edge”

A genuine 100% RTP game clears a wagering requirement at no expected cost, which makes it the strongest possible term an offer can carry. It does not make it risk-free: a fair random walk still hits zero sometimes, and the bust probability shown for such games is real.

Zero-edge claims are only published here alongside their stated conditions — daily wager allowances, maximum bets, and what the RTP becomes once those are exceeded. An unconditioned zero-edge claim is an advertisement, and the data model rejects it.