House Edge, RTP and What Casino Games Actually Cost You

Where these figures come from

This page keeps two kinds of figure apart. One kind is read at build time from this site's own House Edge tool table, edges.ts, the same game models that render Craps Odds, Blackjack Strategy, Baccarat Odds and Roulette Payouts, so this guide cannot quietly disagree with those pages about the same bet; a test re-derives each one independently and fails the build on drift. Which of these figures belong to that group is stated as data below rather than in this paragraph, precisely because restating it in prose is what went wrong the last two times. The other kind is transcribed from a named external citation rather than computed here at all, and stays that way on purpose: presenting a published range as a mechanically ranked point figure would be false precision, not a fix. Which of the page's figures fall into this group and what each is cited to is likewise stated as data below. Two things about that split are worth saying that the data cannot say for itself. First, a figure being derived does not mean every fact near it is: the triple-zero wheel's edge is computed, but that a 39-pocket wheel paying a number 35 to 1 actually exists on casino floors is not something the arithmetic establishes, so it carries a second citation (americancasinoguidebook.com's triple-zero guide, retrieved 2026-09-12) on top of the tool page. The same is true of the half-back rule itself: the surrendered double-zero wheel's edge is computed like any other roulette figure on this page, but that either wheel actually offers a half-back rule is a fact about what a casino floor runs, not something this page's arithmetic can establish on its own. Second, the 'what an hour actually costs' figure in 'Edge times turnover' is a separate derivation from the same source: it uses sessionCost, the identical function house-edge.tsx calls for its own 'Per hour' column, at that tool's own default $10 average bet, so the guide's figure and the tool's panel cannot disagree about what a row costs. The pace each row assumes (blackjack 80/hr, European roulette 40/hr, keno 8/hr, dollar slots 600/hr) is edges.ts's own ROWS.speed field, a representative pace stated as an assumption there, not a measurement. The two craps-odds rows are excluded from that figure because their edge is only that low when several times the average bet is also riding on the odds behind it, which a flat per-row average bet cannot represent honestly. Eleven of the figures above are read directly from edges.ts and re-derived independently by house-edge-ranking-data.test.ts, which fails the build if any of them drift. One of the figures above is read directly from edges.ts and re-derived independently by edge-rtp-comparison-data.test.ts, which fails the build if any of them drift. Five further figures are transcribed rather than derived, each reproducing the figure or range its cited source publishes: Big six wheel (often 11.11% to 24.07%), cited to Wizard of Odds (retrieved 2026-09-07); Keno (often 25% to 29%), cited to Wizard of Odds (retrieved 2026-09-07); La partage, single-zero half-back rule (1.35%), cited to wizardofodds.com and casinonewsdaily.com's surrender-rule guide (retrieved 2026-09-10); Pragmatic Play published return-to-player figure for a European-wheel roulette title (97.3%), cited to Pragmatic Play's own published return figure (retrieved 2026-09-14); Pragmatic Play published return-to-player figure for Baccarat (98.76%), cited to Pragmatic Play's own published return figure (retrieved 2026-09-29). One further figure is composed here rather than reproduced directly from the cited source: Slots (typically 4% to 10%, depending on denomination and machine), cited to Nevada Gaming Control Board revenue filings (retrieved 2026-09-17): the 4% low end is the lowest individual denomination NGCB measures statewide ($5 machines, 4.37%), rounded down to a whole percent; it sits below the slots-dollar row's own 6.01% dollar-weighted 'typical $1+ machine' figure because this bracket describes the spread across denominations and machine types, not a single typical machine, the same distinction slots-penny draws from the other end. The 10% top end is this site's own figure for a tight machine, chosen above the slots-penny row's measured 8.79% statewide mean; NGCB does not publish a top figure for a tight machine, and this page does not claim it does.

Verification

Checked against this site's own /tools/house-edge comparison table (edges.ts, provenance.ts) and, one level further back, Wizard of Odds, americancasinoguidebook.com and casinonewsdaily.com for the rows house-edge itself transcribes rather than derives: agrees on every figure in this guide's cheapest/most-expensive list, the prose and FAQ answers restating it, the misuse note, the triple-zero and half-back roulette figures, the four per-hour figures in the 'what an hour actually costs' figure, and the two rows of the 'computed edge vs published RTP' figure (European roulette, baccarat player, each against its own matching provider citation; the figure's third row, baccarat banker against Play'n GO's Mini Baccarat, was removed on #1451 when Play'n GO withdrew that citation) (verified by edges.test.ts (house-edge tool), src/guides/figures/house-edge-ranking-data.test.ts (this guide's cheapest/most-expensive figures), src/guides/figures/cost-per-hour-data.test.ts (this guide's own per-hour figure) and src/guides/figures/edge-rtp-comparison-data.test.ts (the computed-vs-published RTP figure, #1255: re-derives every row's computed figure from edges.ts and published figure from RTP_RECORDS independently, including an assertion that fails the build if a row's providers ever disagree with each other); the transcribed big six, keno, tight-slots and la partage figures have no automated cross-check and were checked by hand against edges.ts and the cited sources on 2026-09-29, since no automated cross-check between a guide's prose and an external citation exists yet).

Know the risk

• Comparing craps at 0.27% against blackjack at 0.41% as if they cost the same to access. The 0.27% figure assumes maximum 3-4-5x odds taken on every point, several times your base bet at risk each roll; a player who does not take full odds every time is paying a materially higher effective edge than the headline number.
• Reading a 96% RTP slot as meaning you get 96 back for every 100 played in a session. RTP is a long-run average measured over tens of thousands of spins or more; a single session is dominated by variance and can easily land far from it in either direction.
• Treating hold and house edge as the same number. Hold, what a casino actually keeps of the money players bring to the table, is inflated by players rebetting their winnings and is usually much higher than the house edge; using hold to estimate your own expected loss overstates it substantially.

Gambling can cause serious harm. The National Problem Gambling Helpline is free, confidential and open 24/7, call or text 1-800-MY-RESET (1-800-697-3738), or chat at ncpgambling.org/chat.