There is no such thing as the house edge of an online casino. The number differs by more than an order of magnitude between categories that sit side by side in the same account, and the way each category discloses it differs too.
When you look at an operator like the Hippodrome online casino, for example, its platform resolves into four separate products under that test, as most operators in this market do. A reel-based game arrives with a published theoretical return. A table game does not, because its return depends on rules the player selects at the table. The two are not the same kind of number, and they are routinely read as though they were. That leaves the customer to build the comparison unaided, which is a problem of disclosure and not of pricing. The arithmetic behind each category follows.
Low house edge: blackjack and European roulette
Blackjack rewards attention because the player makes decisions that change the math. Under liberal Vegas rules, six decks with the dealer standing on soft 17 and doubling permitted on any two cards and after splits, the theoretical edge with correct basic strategy sits at about 0.28 percent.
Rules move it hard, and none of them are advertised as a return figure. A dealer who hits soft 17 costs about 0.22 percentage points. Removing doubling entirely costs 1.48. Deck count is the quieter variable: under one fixed rule set, a single-deck game runs about 0.48 percentage points better for the player than an eight-deck version of the same game.
European roulette is simpler and less forgiving. One zero across 37 pockets produces a house edge of 1/37, or 2.70 percent, and that figure holds for every bet type on the wheel. A straight-up number and a red/black wager carry identical expectation, and only variance differs. The double-zero layout runs 2/38, or 5.26 percent. French tables carrying la partage return half the stake on even-money bets when zero lands, reducing those bets to 1.35 percent.
Volatility and expanding paylines
Return to player is the one figure that is published, and it is widely misread as a description of a session. It is a specification. The Gambling Commission’s guidance on live RTP monitoring sets out the tolerance band around a theoretical figure at various sample sizes: plus or minus 4.91 percent at 50,000 plays, still plus or minus 1.10 percent at one million. A published RTP is a design target with a confidence interval attached, and the interval stays wide for longer than most people assume.
Volatility describes how fast a game converges on that target. It is not a standardized metric. The low/medium/high labels attached to games are set by the studio, against no external definition, so they are not comparable across suppliers.
Megaways is the best-known expression of high variance. Big Time Gaming introduced the mechanic in 2016 by its own account, replacing fixed paylines with six reels carrying a variable symbol count per spin. Seven symbols on each of six reels gives 7 to the power of 6, or 117,649 combinations. Evolution acquired Big Time Gaming in 2021 and said at the time that the mechanic was “incorporated in over 200 games from multiple providers.”
Live dealer studios
The live category is a broadcast operation, and the disclosure regime there is different again. New Jersey’s regulations require a simulcast control server to “provide the patron with real time visual access to the live game being played,” to “record dealer-verified game results before posting” them to the platform, and to carry a mechanism capable of voiding a result where necessary.
None of that produces a customer-facing number. A live blackjack table’s expected cost is still a function of its printed rules, which the player reads at the table in the same way they would in a room.
Instant multipliers
Crash-style games invert the usual structure, in that the player sets the exit rather than the wager type. At a stated return of around 97 percent, the probability of reaching a multiplier of w is approximately 0.97 divided by w. A 3x target is roughly a 32 percent proposition and a 10x target under 10 percent, and the edge does not improve at any cash-out point.
Every percentage cited above is a long-run property of a game’s design, measured across sample sizes the regulator specifies in the hundreds of thousands. The same discipline applies to any quarterly result, which is that a figure has to be read against the sample it was drawn from. The four categories above are not on a common scale and were never intended to be.
