Slide to pick a number 1–100. Bet OVER or UNDER. Multiplier scales by win-chance — pick 50% for 2× payout, pick 1% for 99× payout. 1% house edge.
| Roll over 50 (49% chance) | 2.02× payout |
| Roll over 75 (24% chance) | 4.13× payout |
| Roll over 90 (9% chance) | 11× payout |
| Roll over 95 (4% chance) | 24.75× payout |
| Roll over 99 (0.01% chance) | 99× payout |
| Expected return is 99% regardless of which target you pick — house edge is fixed at 1%. Same math as Limbo, different UI. | |
"Dice" in the modern online sense is a single-die simulation where you set a target number, choose to win if the roll is over or under that target, and the payout is set so the house has a small (1-2%) edge. Originally a Bitcoin casino staple, it's now a common provably-fair option. Like crash, dice is essentially a continuous lottery with a small skim — every bet has the same expected return regardless of how you configure it.
Pick a target number between 1 and 99 (the underlying roll is 1.0000 to 99.9999 with four decimals). Choose "roll under target" or "roll over target." The payout multiplier is calculated as (99 / win_chance%) × (1 − house_edge). For example, betting "under 50" gives you a 50% win chance; the payout is roughly 1.98x your bet. Betting "under 10" gives 10% win chance and ~9.9x payout. Press roll; the random number is generated server-side (with a client seed for provable fairness), and you win or lose.
Like crash, the expected value of every dice configuration is the same — house edge is built into the payout multiplier. The only meaningful decision is variance preference: low-target/high-chance for frequent small wins; high-target/low-chance for rare big payouts. Martingale-style progressions are popular and dangerous; they always hit a bankroll wall eventually. The "perfect" strategy is flat-bet a fraction of your roll (1-2% per spin) at whatever target matches your variance tolerance, and quit at a session win goal or loss limit.
For house edge h and win chance p (as a decimal), the payout multiplier m = (1 − h) / p, giving EV per dollar = p × m − 1 = (1 − h) − 1 = -h. So with h = 0.015, the expected loss per bet is 1.5% regardless of target choice. Variance, however, scales as approximately m / p — so very low-probability targets have wildly higher variance than 50/50 targets.
Provably-fair dice emerged with Bitcoin-era casinos around 2013 (Bitsler, Primedice). The "client seed + server seed + nonce → hash" verification scheme became a standard pattern for online casinos that wanted to differentiate from traditional RNG-only games.
| Bet | Payout | Notes |
|---|---|---|
| Under 50 / Over 50 | ~1.98x | 49.5% win — lowest variance |
| Under 25 / Over 75 | ~3.96x | 24.75% win |
| Under 10 / Over 90 | ~9.9x | 9.9% win |
| Under 2 / Over 98 | ~49.5x | 1.98% win — highest variance |
Every game on placebets.ai uses virtual chips that reset whenever you clear browser data. There is no signup, no deposit, no withdrawal mechanism, and no monetary value attached to the chips shown on screen. Use the practice environment to drill dice roll's math and strategy without risk. Decide for yourself whether you ever want to play for real money — we'd statistically rather you didn't.