Casino
House Edge and the Martingale Illusion
Lowering your stake, chasing losses, changing tables — none of it removes a house edge. The Casino Lab shows why, transparently and without a cent at risk.
Key takeaways
- House edge lives in the payout table; no staking pattern can alter a payout table.
- Expected value is linear in the stakes, so any sequence of negative-EV bets has negative EV.
- Martingale converts a large chance of a small win into a small chance of a total loss.
- A 1,023-unit bankroll buys only ten doublings, and ten consecutive losses is not a rare event.
- Progressions reshape the distribution of outcomes; they never move its mean.
The edge is in the payout, not the strategy
A house edge is the gap between the fair payout and the actual payout. European roulette pays 35:1 on a single number that hits 1 in 37 — a 1/37, roughly 2.70%, edge on every spin. No betting pattern changes the payout table, so no betting pattern changes the edge.
The Casino Lab shows the fair probability, payout, house edge and player EV for each bet, and lets you watch realized results converge toward that edge over many rounds.
There is a one-line proof that no staking system can help, and it is worth internalising. Expected value is linear: the expected value of a sum of bets is the sum of their expected values, regardless of how the stakes were chosen or how they depend on earlier results. If every individual bet has negative expected value, then every finite sequence of them does too. Staking systems choose the sizes; they cannot change the signs.
Why Martingale feels like it works
Doubling after a loss recovers prior losses plus one unit — until a long streak or a table limit or a finite bankroll stops you, and the single catastrophic loss erases many small wins. Martingale trades a high chance of a small win for a small chance of a huge loss. Its expected value is still negative, because the edge never left.
It feels like it works because it usually does. That is not a contradiction — it is the design. A session that ends in a small profit is the common case, and the rare case that pays for all of them is invisible until it arrives.
Run flat staking against Martingale on the same seed and compare the ruin rate. The progression does not beat the house; it just reshapes when the house collects.
Walk the sequence to the wall
Start with one unit on an even-money roulette bet, which wins 18 times in 37. Lose and stake 2, then 4, 8, 16, 32, 64, 128, 256, 512. Those ten stakes total 1,023 units. So a bankroll of 1,023 units — a thousand times your opening bet — buys exactly ten doublings before the eleventh loss has nowhere to go.
The probability of losing ten in a row at 19/37 per loss is about 1 in 1,650. Play a hundred rounds of this and your chance of hitting the wall at least once is around 6%. Play a thousand and it is better than 45%. When you do hit it, you lose 1,023 units at once, having accumulated at most a few hundred units of one-unit wins on the way.
Multiply it out and the edge reappears exactly where it always was. Every unit staked, at every level of the progression, is still losing 2.70% in expectation. The progression stakes more units, so it loses more in absolute terms — which is why the honest summary is that Martingale increases your expected loss while making it feel less likely.
What progressions actually change
Progressions are not useless to study; they are just useless as strategy. What they genuinely do is reshape the distribution of session outcomes, and that reshaping is worth seeing clearly.
Flat staking gives a roughly symmetric outcome distribution centred a little below zero. Martingale gives a highly skewed one: a tall spike of small wins and a thin, distant tail of catastrophic losses, with the same mean sitting underneath both. Reverse progressions do the opposite — many small losses and a rare large win. The mean is invariant across all of them.
That invariance is the most transferable idea in the whole Casino Lab, and it applies well beyond casinos. Whenever someone shows you a strategy that wins most of the time, the correct next question is not 'how often does it win' but 'what does the losing case cost, and how does the product compare'.
Frequently asked
Can any staking system beat a house edge?
No. Expected value is linear in the stakes, so the expected value of any finite sequence of bets is the sum of the individual expected values. If each bet is negative, every sequence of them is negative regardless of how the sizes were chosen.
How much bankroll does Martingale need?
Doubling ten times from one unit requires 1,023 units in total. That buys ten consecutive losses and no more — and ten consecutive losses on an even-money roulette bet happens roughly once in 1,650 attempts, which is far from rare across a long session.
Why does Martingale win so often in short runs?
Because it is designed to convert a large probability of a small gain into a small probability of a total loss. Most sessions end with a modest profit; the rare session that hits the bankroll or table limit repays all of them at once.
What is the house edge on European roulette?
About 2.70% on every bet. A single number pays 35:1 but wins 1 time in 37, so the fair payout would be 36:1. That one-slot gap is the entire edge, and it applies identically to even-money bets through the zero.
Try it yourself
Everything in this article is something you can run and tweak in the lab — with your own settings and a reproducible seed.
Open the lab →