actfap

House Edge Comparison

The house edge on every common casino bet, side by side, and what it costs you per hour.

This prices the base game at a stake and speed you choose, plus the side bets this site's own tools price alongside it: Three Card Poker's Pair Plus, 6 Card Bonus and Prime, blackjack's 21+3, Perfect Pairs, Lucky Ladies and Royal Match, and baccarat's Dragon 7. It does not price a side bet this site has no tool for, a progressive whose return depends on the reader's own meter rather than a fixed constant (see /tools/three-card-poker instead), or a promotional offer layered on top of a game, none of which appear in the table.

House edge, best to worst

% of each bet · $/hour at $10 a bet

What it costs you

Blackjack, basic strategy, 6-deck S17

Double on any two, double after split, split to four hands, no surrender, blackjack pays 3:2, dealer peeks. Derived from this site's own engine, not transcribed. The same shoe and the same chart with the dealer hitting soft 17 instead is the 0.63% row below.

Per hour

$3.28

at 80 decisions/hr, Blackjack

Average bet$10Decisions/hr80House edge0.41%Per hour$3.28

Decisions / hour

80/hrtypical for this game

Your session

Expected loss, 3h
$9.84
Money through the game
$2,400
Edge
0.41%
Loss per $100 wagered
$0.41

Same $2,400 wagered elsewhere

Blackjack (basic strategy)$9.84 (0.41%)

Roulette (European single-zero)$64.80 (2.7%)

Roulette (American double-zero)$126.24 (5.26%)

Edge against pace

Blackjack: 80/hr at 0.41%Every other bet in the list

Across is decisions an hour, up is the edge on each one. Cost an hour is the two multiplied by your bet, so every bet on one dashed line costs the same hour. Both scales are logarithmic: each gridline is about three times the one before.

These are long-run averages, not what happens tonight. Variance can leave you well up or well down after a few hours, the edge only shows up reliably over tens of thousands of bets.

Embed this chart on your site

Cost-per-hour drill

Best 0/10

Every pair below is a bet that loses money in expectation, at a fixed $10 a bet. The drill is about which one loses less per hour, not which one is good to play. Pairs are weighted toward the counter-intuitive case, about 50% of rounds, rather than drawn evenly from every possible matchup, so this is a teaching set and not a sample of how often edge and pace actually disagree.

Which costs more per hour at $10 a bet?

1/10

Correct 0

Written and reviewed by Alexander Korsager · last reviewed October 2, 2026 · corrections

The idea

What this tool shows

The house edge is the share of every bet the casino keeps on average over the long run. A 1% edge means that for every $100 you wager (not deposit, wager) you lose $1 on average. Because you rebet your winnings, total turnover in a session is many times your buy-in, so a small edge still adds up.

  1. The gap between games is enormous.Blackjack with basic strategy runs about 0.41%; European roulette is 2.7%; American roulette is 5.26%; a tight slot can be 10% or more; and the worst proposition bets in craps or on the big six wheel exceed 15%. Same casino, same hour, wildly different cost.

  2. What actually drains a bankroll is edge multiplied by turnover, and turnover is decisions per hour times average bet.A $10 blackjack player seeing 80 hands an hour puts $800 through the game and expects to lose about $3.28 of it. A $2-a-spin player at 600 spins an hour on a tight penny machine puts $1,200 through and expects to lose about $120, more than thirty times as much for an evening that looks much the same from the outside. The calculator turns your own numbers into an expected hourly figure.

FAQ

Questions players ask

Which casino game has the lowest house edge?

Per dollar wagered, craps: don't pass backed with 3–4–5× odds at 0.27%, then pass line with the same odds at 0.37%. Both assume the maximum odds are taken every time, which means several times your base bet on the table, so the edge is lower because the amount at risk is larger. Comparing base bets alone, blackjack with correct basic strategy (0.41%) and 9/6 Jacks or Better at optimal play (0.46%) come next, then the same blackjack game with the dealer hitting soft 17 at 0.63%, then baccarat banker at 1.06%.

Is a lower house edge always the better game?

For expected cost, yes, but speed and volatility matter too. A low-edge game played fast for high stakes can cost more per hour than a higher-edge game played slowly. Variance also decides whether you actually walk out ahead on any given night.

Does the house edge mean I lose that percentage of my buy-in?

No, it applies to total amount wagered. Over a session you rebet many times, so a $200 buy-in might generate $2,000 of turnover. A 2% edge on that turnover is an expected $40 loss, not $4.

Methodology

How this is calculated

Three separate things sit on this page. The expected cost is simple arithmetic: average bet, times decisions per hour, times the edge, so $10 a hand at 80 hands an hour against a 0.41% edge is an expected $3.28 an hour.

The three comparison lines beneath it take the turnover those same inputs produce and run it against three fixed edges, so they answer what this much money through a different game would cost rather than what a different evening would.

The table of 41 edges is the part that took the work, and the 41 figures are not all the same kind of number.

Thirty-one are recomputed by this site's own test suite from the same models that render on our other tool pages, so this table cannot quietly disagree with /tools/craps-odds or /tools/roulette-payouts about the same bet.

Six are transcribed from a published source and stored next to the exact wording that source printed, with the date it was read.

The last four are slots, keno and the money wheel, where the real figure is set per machine or per house and nobody outside the operator knows it: those rows carry the published range they came from and pick a representative from inside it. None of the 41 is unsourced, and a test fails the build if one ever is.

Common misreadings5
  • The lowest edge is not the cheapest evening. At $10 a go, 9/6 Jacks or Better at 0.46% dealt 600 hands an hour costs about $27.60 an hour, while European roulette at 2.7% spun 40 times an hour costs about $10.80. Speed is doing more work than the edge in that comparison.
  • A lower edge does not mean you win more often. It is the average cost of a wager, not the share of wagers that come back, and the two move independently: a game can be cheap per bet and still return nothing on most of them.
  • Sorting the table best to worst sorts cost, not merit. Every row is money leaving at a different rate, and there is no figure on this page where the number turns in the player's favor.
  • The expected loss the calculator prints is not a budget you will land on. It is the center of a wide distribution, a losing session can be several times it, and a winning one is not evidence the edge stopped applying.
  • The three comparison lines are not three evenings. Select 9/6 Jacks or Better at $10 a hand for three hours and the strip is pricing $18,000, because that machine is dealt 600 hands an hour. The $946.80 beside American roulette is what $18,000 costs at 5.26%, not what three hours at a roulette table costs: a wheel turns about 40 times an hour, so the same three hours put $1,200 through it and cost $63.12. The strip compares edges at equal money wagered. The calculator above it compares evenings.
Assumptions7
  • Every edge is a share of the amount wagered, not of the amount you brought. Winnings get rebet, so a few hours of play puts many times your buy-in through the game.
  • Each row describes one rule set, and the figure moves when the rules do. That is why three games here get two rows rather than one. Blackjack is six decks, double on any two, double after split, no surrender and blackjack paying 3:2, listed once with the dealer standing on soft 17 (0.41%) and once with the dealer hitting it (0.63%). The baccarat tie is listed at both pays it is dealt at, 8 to 1 and 9 to 1. Jacks or Better is listed at the 9/6 and 8/5 schedules. A blackjack table paying 6:5 instead of 3:2 is a different game again, and a substantially more expensive one. Roulette gets five rows for the same reason: three wheels a US player can actually walk up to, at 2.70%, 5.26% and 7.69% for 37, 38 and 39 pockets, and the half-back rule some tables run on the first two of them.
  • Two rows in this table are true of part of their game rather than all of it, and both say so on the row itself. La partage at 1.35% and surrender at 2.63% are the six even-money bets only: red, black, odd, even, 1 to 18 and 19 to 36. Every other bet at those same tables still costs 2.70% or 5.26%, the American five-number bet's own row at 7.89% aside, because the rule refunds half of an even-money loss and reaches nothing else on the layout. Read them as a price on six bets at a table posting that rule, never as the price of roulette.
  • Rows labeled optimal assume optimal play throughout: correct basic strategy, the house way, the published cut-off for a Play bet. The blackjack figure is deliberately what a player following our published chart achieves, not the lower figure a hand-by-hand solver reaches, because quoting the solver would tell a reader their chart costs them less than it does.
  • The craps rows that mention odds assume maximum odds are taken every time. Pass line with 3–4–5× odds wagers about 3.8 units a hand rather than one, and the 0.37% is the same expected loss spread across that larger stake.
  • Typical speed is a representative pace for the game, not a measurement of your table. Machines deal several times faster than anything dealt by hand, which is most of the reason two low-edge games can cost very different amounts per hour. The speed field is editable for that reason.
  • The three comparison lines at the foot hold turnover constant rather than time, and the heading over them names the figure being held. Your stake, your speed and your hours decide how much money goes through the game, and the strip prices that same amount of money against blackjack and both roulette wheels. Selecting 9/6 Jacks or Better at $10 a hand for three hours puts $18,000 through the machine, so the strip reads $18,000 wagered and the $946.80 beside American roulette is what $18,000 costs at 5.26%. Holding the amount wagered still is how two edges are compared. Holding the hours still was only ever a proxy for it, and the proxy breaks whenever the two games are dealt at different speeds, which is the case the strip exists for.
Limits7
  • The two slots rows are still estimates, and the honest statement is that nobody outside the operator knows any one machine's answer. Returns are set per machine and mostly unpublished. Since #666 both are grounded on Nevada Gaming Control Board filings rather than a third-party summary alone: the state's own reports measure $1-and-above machines pooled together at 6.01% and penny machines at 8.79%, both dollar-weighted over the trailing twelve reports, so 6% stands for a higher-denomination machine just above that measured floor and 10% for a tight penny machine above the statewide penny mean. That is one state's floor over one twelve-month window, not a national figure, which is why both rows stay estimates rather than becoming derived ones.
  • Keno at 27% and the money wheel at 19% are single numbers standing in for spreads, 25% to 29% and 11.11% to 24.07% respectively, the latter varying by which symbol is backed. Read them as the neighborhood rather than the price.
  • Eight side bets are in this table now, priced at the pay table our own tool for that bet loads on: Three Card Poker's Pair Plus, 6 Card Bonus and Prime, blackjack's 21+3, Perfect Pairs, Lucky Ladies and Royal Match, and baccarat's Dragon 7. Each is here because what it costs is decided by the pay table rather than by the game, so the row states which schedule it describes and names the tool page where the reader can price a different one. Insurance beside a blackjack game is not here, for want of a tool: this site prices nothing for it. The Three Card Poker progressive is not here either, and not for want of a tool any more: /tools/three-card-poker prices it as a calculator, but no single constant can describe a bet whose return depends on the reader's own meter and posted prizes rather than on a fixed pay table, so it has no row on a table that ranks constants. A player taking any side bet is putting more through the game per hand than any base-game row here describes.
  • The triple-zero row at 7.69% is a wheel a US player meets on a casino floor rather than one they meet everywhere. It premiered in Las Vegas in 2016 and has spread since, but which floors spread it is an operator's decision that changes, and this table cannot tell you whether the wheel in front of you is one. Count the green pockets, which is printed on the wheel and on the layout and is the cheapest check on this page to actually make.
  • Which tables run a half-back rule is a fact about operators and jurisdictions rather than about the wheel, and it changes. The 1.35% and 2.63% rows say what the rule is worth where it is offered, not where that is, and a table's own posted rules are the only reliable answer to that. En prison is a further variant, works out slightly worse for the player than la partage, and is not priced anywhere here.
  • Comps, free play and rebates are not modeled in either direction, and neither is a tip.
  • An edge is an average over tens of thousands of decisions. Across one evening the outcome is decided by variance, and nothing on this page describes one evening.
Evidence

edges.test.ts is the check, and it is worth knowing what it does and does not prove. It asserts that the table and the citation file describe exactly the same 41 rows, so no row can be added without saying where it came from, and it fails if any row is left unsourced.

The thirty-one derived rows are recomputed from this repo's own models, the code that renders /tools/craps-odds, /tools/roulette-payouts, /tools/baccarat-odds, /tools/video-poker, /tools/three-card-poker, /tools/blackjack-strategy and /tools/crazy-time, and each has to match to within half of the last digit shown here.

That check exists because it caught something real: this table published the craps Any 7 bet at 16.9% while our own craps page printed 16.67%, which is exactly one sixth, and both were live for as long as both pages existed.

The six transcribed rows and the four range estimates cite the Wizard of Odds, retrieved 2026-09-07, and hold the figure as that source words it so the number and the citation cannot be edited apart.

The two half-back roulette rows added on 2026-09-10 are derived like the rest, from the same function /tools/roulette-payouts prints, and also carry a citation, because the arithmetic and the premise are different claims: that half a stake back on a zero halves the edge is arithmetic, while the rule existing at all, being called surrender and la partage, and covering six bets and no others, is a fact about how casinos operate.

Those come from wizardofodds.com's roulette basics page and casinonewsdaily.com's surrender-rule guide, both retrieved 2026-09-10, and the test checks each source prints the figure our derivation lands on.

The triple-zero row added on 2026-09-11 splits the same way and carries two citations rather than one, because no single source states every premise it rests on. That 36 units returned across 39 pockets is a loss of 1 in 13 is arithmetic, and it is recomputed here like the other wheels.

That such a wheel exists, that it holds 39 pockets and that it still pays a straight-up number 35 to 1 are not: wizardofodds.com's roulette basics page gives the 39-number wheel and the 1/13 edge but leaves the payout to be read off its statement that the wheel carries the same numbers and colors as a double-zero one, so it is cross-checked against americancasinoguidebook.com's triple-zero guide, which states the 39 spaces, the 35-to-1 payoff and the same 7.69% together.

Both re-read and re-quoted 2026-09-12, when #456 item 2 made the wheel selectable on /tools/roulette-payouts, and the Wizard of Odds quote here now carries the any-green payout this row prices in prose as well as the pocket count. The test holds both sources to the same bar.

One caveat on the word derived, since it is easy to read as more than it is: it means the figure is recomputed from a model in this repo and so cannot drift from the rest of the site, not that we originated it. Three Card Poker's 3.37% is a published figure our own code holds.

And a passing test means the code matches the model, not that the model matches the table in front of you.

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.

Read up

Go deeper

Related tools

Try next

We’d like to use privacy-friendly analytics (Google Analytics, Vercel Analytics and Speed Insights) to see which tools and pages are actually useful, so we can improve them. Nothing is sold, and nothing loads unless you say yes. See our privacy policy for details.