What a 2.5% House Edge Actually Costs You Per Hour, Not Per Session
A house edge of 2.5% or 0.5% sounds small enough to round to nothing, and that reading is where the misconception lives. The number is small because it is a percentage of every pound wagered, not a percentage of the money sitting in front of a player, and the missing variable that turns a small percentage into a meaningful hourly cost is simply how many bets get placed. Two games with identical edges can cost wildly different amounts per hour, and the same game can cost ten times more or less depending on nothing except pace.
What "2.5% House Edge" Actually Means
House edge is defined as the casino's expected return as a proportion of total amount wagered, sometimes called handle, across many bets. A single-zero roulette wheel carries a 2.70% house edge because, averaged over the 37 possible outcomes, the payout structure returns 97.30% of every pound bet on it in the long run. That figure says nothing on its own about how much a specific player loses in a specific hour, because it only measures loss per pound wagered, and a player who wagers ten times more in an hour wagers ten times more pounds into that same 2.70% calculation.
The intuitive misreading treats the percentage as if it applied to a bankroll: "I've brought £100, so a 2.5% edge means I should expect to lose about £2.50 tonight." That statement is only true if exactly £100 total gets wagered across the whole session, one single pound-for-pound bet. It is not true the moment a player rebets their winnings and their stake repeatedly across dozens or hundreds of individual wagers, which is how casino games are actually played.
The Missing Variable: Bets Per Hour
Expected hourly loss follows a simple derivation: house edge multiplied by average bet size multiplied by the number of bets placed per hour. Each term is independently verifiable, and the last one is the term most players never think to estimate.
Published figures for bets per hour vary by format and by how many players share a table. Casino operations research by Kilby, Fox and Lucas, cited by the gambling mathematics reference Wizard of Odds, records blackjack hands dealt per hour ranging from 52 at a full seven player table up to 209 hands per hour heads up against a dealer with no other players. Industry-facing estimates from casino operators themselves tend to cluster lower, commonly citing 60 to 70 hands per hour as a typical multi-player average, with faster or slower play depending on table occupancy. For roulette, the same operations research records a range from 35 spins per hour at a six player table up to 112 spins per hour played solo.
Online RNG play removes the physical dealing, shuffling and other-player pacing that sets those live table limits. Speed of play data compiled by gambling software analysts puts online blackjack software anywhere from roughly 250 hands per hour on slower platforms up to approximately 2,300 hands per hour on fast autoplay settings, a range with no live table equivalent because nothing physical constrains how quickly a random number generator can resolve the next hand.
Worked Case: Blackjack
House edge held constant at 0.5%, average bet held constant at £10
| Setting | Hands per hour | Expected hourly loss |
|---|---|---|
| Full table, multiple players | 60 | £3.00 |
| Casino operations research, full 7 player table | 52 | £2.60 |
| Casino operations research, heads up | 209 | £10.45 |
| Online RNG, slower software | 250 | £12.50 |
| Online RNG, fast autoplay | 2,300 | £115.00 |
The house edge never changes across this table. The average bet never changes. The expected hourly loss changes by a factor of over 40 between the slowest and fastest settings, entirely because of how many times that 0.5% gets applied within an hour.
Worked Case: Roulette
House edge held constant at 2.70% (single zero), average bet held constant at £10
| Setting | Spins per hour | Expected hourly loss |
|---|---|---|
| Six player table | 35 | £9.45 |
| Casino operations research, solo play | 112 | £30.24 |
Roulette carries a higher edge than blackjack played with correct basic strategy, which is why its hourly figures sit higher even at comparable pace. But the same pattern holds inside the game itself: playing alone against the wheel, with no other players setting the pace, more than triples the expected hourly cost relative to a full table, without the edge itself moving at all.
What Intuition Gets Wrong, Set Against the Maths
Illustrative comparison, not a specific operator's figures
| What intuition suggests | What the maths actually shows |
|---|---|
| "A 2.5% edge means I'll lose about 2.5% of what I brought" | The edge applies per pound wagered across every bet placed, not once against a starting bankroll |
| "A small edge means a small hourly cost" | Hourly cost depends on edge multiplied by bet size multiplied by bets per hour; a small edge at high speed can outcost a larger edge at low speed |
| "Online play is basically the same as playing at a table, just on a screen" | Removing the physical pacing of a live table can raise the number of bets per hour by a factor of ten or more for an identical edge and bet size |
| "Playing solo against a dealer or wheel doesn't change my odds" | It doesn't change the edge on any single bet, but it raises the rate at which that edge gets applied, which raises the expected hourly cost |
What This Looks Like Across a Longer Session
The per hour figures above compound in the most literal sense: multiply by however many hours the session runs, because expected loss under a fixed edge and fixed pace is linear in time. There is no point at which the maths changes direction, no session length at which the expected value stops accumulating in the casino's favour, because each individual bet carries the same fixed edge regardless of how many bets came before it.
Expected loss, not a guaranteed outcome; any individual session can land above or below this figure
| Format and pace | 1 hour | 3 hours | 6 hours |
|---|---|---|---|
| Blackjack, full table (60 hands/hr) | £3.00 | £9.00 | £18.00 |
| Blackjack, online autoplay (2,300 hands/hr) | £115.00 | £345.00 | £690.00 |
| Roulette, six player table (35 spins/hr) | £9.45 | £28.35 | £56.70 |
| Roulette, solo play (112 spins/hr) | £30.24 | £90.72 | £181.44 |
One hour of fast online autoplay blackjack, at £115.00 expected loss, costs more than six full hours of full-table blackjack at £18.00, despite the shorter session running for a sixth of the time. Session length alone tells a reader very little without also knowing the pace at which bets were placed within it, which is the same missing variable from the hourly figures above, just extended across a longer stretch of time.
Why Speed Is the Real Lever
None of this changes the probability of winning or losing any individual hand or spin. The edge on a single £10 blackjack bet at 0.5% is identical whether it is the first hand of a slow evening at a crowded table or the two thousandth hand of an hour spent on autoplay. What changes is exposure: the number of times that fixed, unfavourable edge gets to act on money wagered within the same stretch of time. A player who wants to reduce expected hourly cost without changing the game, the bet size or the strategy has exactly one lever available that this derivation identifies: the pace of play itself, which is also the one variable most players never think to measure, let alone set a limit on.
A 2.5% house edge costs 2.5% of everything wagered through it, not 2.5% of a bankroll, and at the table that difference is decided almost entirely by how fast the bets keep coming.
Responsible Gambling
None of the maths above predicts what happens in any specific session; it describes an average across many repetitions, and any individual outcome can sit well above or below it. Anyone offering these games to UK players should hold a valid UK Gambling Commission licence, and a time and deposit limit set before play starts is the only lever in this piece that a player actually controls, since it caps the pace variable directly rather than hoping a session ends before the expected value catches up.