Why dice is a “low-edge” sandbox
Many crypto dice games run on a roughly 1% house edge, which corresponds to about 99% RTP. Primedice and Stake’s dice-style originals are widely documented examples in the provably-fair space, and third-party reviews frequently cite the 1% edge. Lower edge means smaller expected loss per unit wagered compared with typical slots.
Provably-fair implementations show how each roll is generated from a server seed, your client seed, and a nonce via HMAC hashing; both Stake and Primedice publish their RNG details so you can verify results independently.
The core math: payout vs win chance vs edge
In a standard dice UI you choose a win chance p and get a payout that is slightly reduced by the house edge. For a game with RTP r (for example r = 0.99), offering “odds” b (for example 2-for-1), the required win probability is roughly r/b. Wizard of Odds demonstrates this directly for crypto dice: with 99% RTP and a 2-for-1 payoff, the platform sets win probability to 0.99/2 = 49.5%. That general proportionality is how the 1% edge is applied across all targets.
Another way to see the edge is at 50% win chance: a “fair” game would pay 2×, but 1%-edge dice typically pays 1.98×, implying a 99% theoretical return on that bet. Community explanations lay out this equivalence.
Expected loss and variance: what changes between short and long runs
House edge is the casino’s long-run average profit per bet (percentage of stake). Your session’s expected loss is roughly edge × total handle. That’s independent of your chosen target multiplier because the game compensates win chance and payout to keep the same RTP.
Variance is what you actually feel. For n independent bets with win probability p, the standard deviation of wins scales with √(n·p·[1−p]); averages stabilize with more trials by the law of large numbers, but the total swing can still be substantial. Don’t confuse convergence of averages with guaranteed profit; the average approaches the expected value as n grows, not the running sum.
Bankroll safety rails you should not skip
The Kelly criterion gives an optimal fraction only when you have a genuine edge; with zero or negative edge, Kelly prescribes a bet size of zero. In dice you do not have an edge against a 1% house, so treat Kelly as a sizing warning, not a reason to scale up.
Gambler’s ruin shows that with a negative expectation, a persistent player with finite bankroll will eventually go broke against the house, regardless of betting system. No progression scheme fixes the math. Design sessions to cap losses and time, not to “chase back.”
Short-run plan (about 20–150 rolls): stable, low-swing sessions
- Use a small fixed unit size relative to bankroll to keep drawdowns tolerable. A flat bet is easier to audit than progressions.
- Prefer moderate win chances (for example, 40–60%) to keep hit rate steady; your EV stays the same, but variance per bet drops compared to high-multiplier shots. The platform sets win chance and payout using the RTP proportion rule, so you’re mainly choosing volatility.
- Pre-set a loss stop and a time limit, and exit when either hits. This contains the practical impact of variance even when averages are converging only slowly.
- If you’re testing a site, rotate provably-fair seeds and verify a few results with the site tool before upping stake size. Stake and Primedice publish the steps.
Long-run plan (500–5,000+ rolls): disciplined volume with known costs
- Budget the whole session’s theoretical cost: expected loss ≈ 1% of total handle on a 1%-edge dice. If you plan 2,000 rolls at 0.001 BTC each, total handle is 2 BTC-equivalents in units of minimum wager; the 1% expectation gives a clear cost benchmark.
- Keep unit size small so that n is large; the standard deviation of the average result shrinks with √n, smoothing outcomes, though you can still have deep drawdowns.
- Track session P/L versus the theoretical curve so an anomalous swing doesn’t tempt staking changes. Avoid escalation systems; they increase ruin probability in negative-edge games.
- Consider breaking a long run into equal “blocks” with mini stop-losses and cool-offs to interrupt tilting behavior. That risk discipline, not a new bet formula, is what protects your bankroll.
Volatility tuning: when to pick high vs low targets
Higher multipliers deliver fewer but larger hits; lower multipliers hit more often with smaller wins. Because crypto dice adjusts p and payout to the same RTP, your choice is primarily a volatility dial. For steady bankroll arcs, favor mid-probability bets; for “lottery-style” shots, expect long miss streaks and plan smaller units.
Provably-fair checklist before you start
Open the provably-fair page and confirm the inputs (server seed, client seed, nonce) and the HMAC algorithm. Note the hashed server seed is shown before play and revealed later; use the site’s verifier to reproduce outcomes from the seeds. Stake and Primedice both document these steps.
Worked micro-examples you can adapt
Example A: steady short session. Suppose you run 100 rolls at a 50% win chance with fixed units. With a 1% edge, expected loss is about 1% of total handle. Variance of wins around the 50-win mean scales with √(100·0.5·0.5) = √25 = 5 wins; that’s why short sessions can easily finish up or down despite the small edge.
Example B: long session budgeting. Plan 2,000 rolls with the same unit size. The expected loss scales with handle, but the standard deviation of the average result drops by √(2000/100) ≈ 4.47× versus the short session, making outcomes tighter around expectation, not profitable.
Common pitfalls to avoid
Chasing losses with progressions accelerates ruin in a negative-edge game. The mathematics of gambler’s ruin doesn’t care about your last result or your bet-doubling scheme; with finite bankroll, persistence increases the chance of going broke.
Treat Kelly fraction claims with skepticism unless you truly have an edge; in casino dice you do not. With zero/negative edge, Kelly suggests betting nothing.
Quick start checklist
- Pick a 1%-edge, provably-fair dice game and learn its verifier.
- Fix a unit size and number of rolls; compute expected loss as edge × handle.
- Choose volatility by setting a moderate win chance for steadier sessions.
- Set hard stops for loss and time; avoid progressions.
- Rotate seeds and spot-verify results occasionally.
FAQ
Does a higher win chance improve RTP?
No. The platform adjusts payout so expected return stays near the game’s RTP (for example, 99% at 1% edge). You trade off variance, not EV. Wizard of Odds shows this proportion for crypto dice.
Isn’t the law of large numbers supposed to “guarantee” profit eventually?
It only says the average result converges to the expected value with many trials, not that cumulative profit turns positive. With negative expectation, long play merely makes losses more predictable.
How do I verify a roll?
Use the provably-fair page to enter the revealed server seed, your client seed, and the nonce. The HMAC process reproduces the roll so you can confirm fairness. Stake and Primedice publish implementation details.
Can Kelly sizing help me here?
Kelly only applies when you have a positive edge estimate. For house games with negative edge, the Kelly fraction is zero.
