What is house edge? RTP and house edge, explained properly.
Two terms, one number. Here's the exact formula behind both, why even a "small" edge guarantees the house wins in the long run, and which common house-edge claims are actually myths.
Published August 21, 2026 · RTP Lab
RTP and house edge are the same fact, said two ways
Return to player (RTP) is the long-run share of every dollar wagered that a game pays back — a property of the game's payout table, not a promise about your next spin or roll. House edge is just the other side of that same coin:house edge = 100% − RTP. A game with 99% RTP has a 1% house edge. There's no second formula to learn — it's the identical number, framed from the house's side instead of the player's.
A 1% edge still wins, every time, given enough bets
A single bet at a 1% edge can obviously go either way — that's the whole appeal. But house edge isn't describing a single bet, it's describing the average across thousands of them. Wager $100 at a 1% edge and the house's expected take is exactly $1.00, whether that shows up as one $1 loss, a $50 win followed by a $51 loss, or any other path — the average across enough repetitions converges on that number by simple law-of-large-numbers arithmetic, not because any individual round is rigged toward it.
This is why "the house always wins" is a statement about volume, not about any one bet being predetermined — see how provably fair actually works for why individual rounds can't be rigged toward that average even though the average is guaranteed.
Crash and Dice run one of the tighter edges around
House edge varies a lot by game, and it's worth knowing where a game sits before you play it. These are widely-cited industry figures, not independently re-derived the way this site's own Crash and Dice formulas are:
What house edge doesn't do
"Hot" and "cold" don't exist
Each round is independent. A machine or table that just paid out big carries no memory into the next round — the edge is identical either way.
Bet size doesn't change the edge
Betting bigger or smaller changes how much is at stake, not the percentage the house keeps on average. Doubling your bet doubles the risk, not your odds — see how this plays out for the Martingale system specifically.
Board state doesn't leak information
On a game like Mines, which tiles you've already revealed changes nothing about where the remaining mines are — they were placed and hashed before the round started.
A "safer" target isn't a better edge
Choosing a lower target or narrower threshold changes how bumpy your results are — the underlying edge stays fixed. That's a variance question, not an edge question — covered fully in RTP vs. volatility.
Quick answers
What's considered a "good" house edge?
Under 1–2% is tight by casino standards — Crash and Dice here both run at a cited 1%. Blackjack with correct basic strategy can go lower (around 0.5%). Slots typically sit much higher, often 4–15%.
Does RTP mean I'll get that percentage back?
No — RTP is a long-run average across many thousands of rounds, not a guarantee for any single session. A short session can land well above or below it purely from normal variance.
Is a higher RTP always better?
All else equal, yes — a smaller edge is mathematically better for the player. But RTP alone doesn't capture volatility, which is a separate axis entirely; see RTP vs. volatility for that distinction.
Can a platform change its RTP after I've started playing?
That would undermine the entire premise of a published edge, and provably-fair verification exists specifically so a per-round result can't be quietly adjusted — see how the verification actually works.
More from the blog
Is provably fair real?
The actual three-step mechanism, a real verification walkthrough, and what it deliberately doesn't prove.
Does the Martingale strategy work?
The proof, a real bankroll-explosion table, and a 1,000-session live simulator to test it yourself.
RTP vs. volatility
A wider target doesn't touch the house edge — it changes variance, a completely separate number.
See this exact edge, at any target, computed live on the full toolkit.
See the full toolkit