The house doesn't need you to lose

Every bet you place has a fee attached. It is not printed on the slip, it is not on the receipt, and nobody will ever tell you what it was.

This is what that fee is in each game, why it gets more certain the more you play, and why it is still not the thing that cleans most people out.

The number that isn't on your slip

Take the most ordinary bet in sports betting: a point spread at -110. Both sides look like a coin flip, and they are priced like one.

You risk $110 to win $100. For that bet to be worth placing, you have to be right more than 52.38% of the time. Not 50%. The extra 2.38 points is the fee.

Now look at it from the other side of the counter. Two bettors, one on each team, $110 each. The book is holding $220 and pays the winner $210. It keeps $10 no matter who wins, which is 4.5% of everything wagered.

That is the house edge. It is not a trick and nothing is rigged. It is a price. The only unusual thing about it is that it is the one price in your life that is never written down anywhere.

What it costs, game by game

Roughly, with standard rules, as a share of what you wager:

• Blackjack, played with basic strategy: about 0.5%
• Baccarat, banker bet: 1.06%
• Craps, pass line: 1.41%
• Roulette, single zero: 2.70%
• Sports, straight bet at -110: about 4.5%
• Roulette, double zero: 5.26%
• Slot machines: 7.15%, the actual statewide average across Nevada in 2025
• Sports, parlays: around 20%
• Keno: 25% to 35%

Two things are worth noticing in that list.

The first is that the honest games are at the top and nobody advertises them. Nothing on a betting app is designed to move you toward the pass line.

The second is the direction everything has been travelling. Nevada's slot hold was 6.55% in 2004 and 7.15% in 2025. American sportsbooks kept 6.9% of everything wagered in 2019 and 10.2% in 2025, and the reason is parlays. Over one recent twelve month stretch parlays were 32% of everything wagered in New Jersey and produced 65% of the revenue. In Maryland it was 36% of the handle and 63% of the revenue.

The product did not drift in that direction by accident. Same game parlays, boosts, and live in-play markets are all the same move: replace a 4.5% product with a 20% product, and make the 20% one the fun one.

Why time is the enemy

Take single zero roulette, $10 a spin. The edge is 2.70%, so on average you lose 27 cents a spin. Twenty seven cents. That is the whole disadvantage, and it is small enough that you will never feel it.

The reason that tiny number always wins in the end is that it accumulates and your luck does not.

After a number of spins, your expected loss is 27 cents multiplied by that number. Your swing, meaning how far luck can move you in either direction, grows by the square root of it instead. Their side grows with the count of bets. Your side grows with the square root of the count of bets. Square root loses. Always, given enough bets.

What that looks like at $10 a spin:

• After 100 spins: expected loss $27, typical swing $100. You have about a 39% chance of being ahead.
• After 1,000 spins: expected loss $270, typical swing $316. About 20%.
• After 10,000 spins: expected loss $2,700, typical swing $1,000. About 1 in 300.
• After 100,000 spins: expected loss $27,000, typical swing $3,162. It does not happen.

Nobody sits at a wheel for 100,000 spins. That is exactly why the last twenty years of this industry have been spent raising the count. A slot machine takes hundreds of spins an hour. An app in your pocket has no closing time and no drive home. In-play betting turned one bet per game into one bet per drive.

None of that changes the edge by a single decimal. It does not have to. Volume is the product.

The edge is not what cleans you out

Here is where most explanations stop, and where they quietly get it wrong.

If the edge were the whole story, losing everything would take a very long time. At $10 a spin with 27 cents against you, a $1,000 bankroll survives a very long night on average. People do not lose their savings that way. They lose it in a weekend.

So the edge is not the mechanism. It is the tilt in the floor. Something else does the pushing.

Start with what the tilt alone actually does, because this part gets exaggerated. Money in an index fund grows exponentially: each year's return lands on a balance that already grew, so it accelerates. Losing at a fixed bet size does not work like that in reverse. Bet $10 a spin forever and your expected balance falls in a straight line, 27 cents at a time. Straight lines are slow.

But almost nobody bets a fixed amount. Bet sizes move with the balance, up after a good run and up after a bad one. The moment they do, the maths turns multiplicative, and multiplicative maths is savage in a way that is genuinely hard to feel.

Here is the cleanest demonstration of it, and it uses a perfectly fair game with no house edge at all.

Bet half of everything you have on a coin flip paying even money. Win and you have 1.5 times what you started with. Lose and you have half. Do one of each, in either order, and you are at 0.75. You are down 25% having won exactly half your bets, on a fair coin, against nobody's edge.

The loss came entirely from sizing the bets as a fraction of a moving balance. There is a name for the gap between what the average says you should get and what you actually end up holding, and it is volatility drag. It is why the same swings that make a huge night possible make a flat year impossible.

So: does it compound against you the way an index fund compounds for you? Yes, and that is a real effect, not a figure of speech. But it is not coming from the house edge, which is linear and slow. It is coming from the bet sizing. The house supplies the tilt. The sizing supplies the exponent.

Winning lets you keep playing. Losing ends it.

Now add the thing that makes it final.

You have a floor and the house does not. If you lose everything you stop. If the house loses a hand it deals the next one. Those two positions are not symmetrical, and probability has a name for the difference: an absorbing state, a place you can enter and never leave.

This has been understood for roughly 350 years. It is the gambler's ruin problem, and the result is harsher than most people expect. Play a perfectly fair game against an opponent with much more money than you, keep playing, and your probability of eventually going broke approaches 1. Not likely. Certain. With no house edge involved at all. The only requirement is that they can absorb a bad run and you cannot.

The all in version is that same fact at high speed. Double or nothing on a coin flip is a fair bet every single time you take it. Taking it repeatedly is not a fair strategy, because a loss ends the sequence and a win only earns you the right to take it again.

Ten straight coin flips leaves you standing about 1 time in 1,000. Even at 90% to win each time, which nobody has ever had, ten in a row leaves you standing 35% of the time and twenty leaves you at 12%.

That is why the story always sounds the same. It is never "I lost it slowly". It is "I was way up, and then I put it all on one thing". The wins never end the sequence. Only the loss does.

Why the last win is never enough

There is still a gap. All of the above explains why you cannot beat the edge. It does not explain why people keep going well past the point where they can see what is happening, and it does not explain the particular feeling that follows a big win, which is not satisfaction.

Three findings cover most of it.

Your baseline moves to wherever you last were. In 1978, three researchers interviewed 22 major lottery winners and compared them with their own neighbours. The winners were not measurably happier. They also rated ordinary daily pleasures as significantly less enjoyable than the control group did. The theory behind it is adaptation level: a peak experience does not simply fade, it raises the bar for everything that comes after.

On a betting account that is the mechanic you already recognise. Once you have been up $4,000, a $200 win is not a $200 win. It gets measured against $4,000 and it registers as nothing. To feel what you felt, you now have to risk what would have sounded insane a month ago. Not because you got reckless. Because the scale moved.

Losses make people actively want the riskier option. Thaler and Johnson, 1990, with real money on the table. After a loss, people strongly prefer bets that offer a route back to even, and will take worse odds to get one. They called it the break even effect. It is not a collapse of willpower, it is a predictable shift in what looks appealing, and it shows up exactly when you can least afford it. It is also the engine underneath chasing losses.

And it is visible in the account data. Researchers looking at real online betting accounts found that people who eventually closed their account over gambling problems had been increasing their stake per bet as that closure got closer. A separate study following 32,262 online sports bettors found that escalation in bet size and frequency during the first eight months predicted later self-exclusion on its own, over and above simply betting a lot.

Put those three together and the shape of a bad ending stops being mysterious. The bar rises after every high, the appetite for risk rises after every loss, and the bet size rises with both. The edge does not have to be big. It just has to be there while all of that runs downhill.

They know exactly who pays

None of this is a secret to the people selling it.

In 2024, researchers analysed the accounts of 39,995 customers of Norway's state gambling operator. For sports betting, the top 1% of accounts produced 30.8% of all losses. The top 5% produced 66.1%. The top 10% produced 78.9%.

That last number is worth sitting with, because it changes what the business actually is. This is not a small edge spread thinly across a lot of casual customers. Roughly eight in every ten dollars comes from one customer in ten.

Which means the edge is the advertised product and the concentration is the real one. A company in that position does not need more customers. It needs the ones it has to keep going, and it needs the heaviest ones most of all.

That is not an inference. In September 2025 the Massachusetts Gaming Commission published an analysis of the sportsbooks licensed in the state and found that operators had restricted customers who win regularly, while extending VIP status and rewards primarily to those who often lose. The commission spent the following month debating guardrails, including a minimum age of 25 for VIP programmes and affordability checks before anyone can be enrolled in one.

Free tickets, a personal host, a reload bonus after a rough week. Those are not rewards for being a good customer. On the regulator's own reading, they are what a losing streak buys you.

What to actually do with this

• Know the fee before the bet, not after. If you cannot say roughly what the house's cut is on what you are about to place, you do not know the price. Parlays are the ones to check.
• Watch the size, not the result. Escalating stakes is the marker that shows up in the research, well before anything else does. If this month's normal bet is bigger than last month's and you cannot say why in one sentence, that is the signal.
• Never bet a fraction of a moving balance. Doubling up to get back to even is the exact structure that produces ruin, and it does it even in games with no edge at all.
• Treat "one more, all in" as the end of the sequence. Because it is. Tonight or in six months, that sequence ends the same way, and the wins are what keep it running.
• Count from the start, not from the peak. Your real number is what you have put in against what you hold now. The peak is a number you never actually had.

If the hard part is not the maths but the hour after a bad loss, that is a different problem and it needs different tools: a plan written down before you need it, a person you can call, and something that puts a gap between the urge and the app. Ungamble is built for that gap, with a panic button and Rux, an AI friend trained on the science of quitting who will happily run any of these numbers with you at 1am.

The one thing worth remembering

The house edge is real, it is small, and it is not what takes people apart.

What takes people apart is a floor at zero that the house does not have, bets that grow with the balance, a bar that climbs after every win, and an appetite for risk that spikes after every loss. The edge just makes sure all of it runs one direction.

You cannot out-play a fee charged on volume. You can stop supplying the volume. That is the entire move, and it is the only bet in the building with a guaranteed positive return.

Sources: house edge figures are standard for common rule sets. Nevada slot hold: analysis of Nevada Gaming Control Board revenue reports by the UNLV Center for Gaming Research. National sportsbook hold and the parlay share of handle and revenue in New Jersey and Maryland: state gaming regulator filings as reported by Legal Sports Report and the Washington Post, 2025. Break even and house money effects: Thaler and Johnson, "Gambling with the House Money and Trying to Break Even", Management Science 36(6), 1990. Adaptation to major wins: Brickman, Coates and Janoff-Bulman, "Lottery Winners and Accident Victims: Is Happiness Relative?", Journal of Personality and Social Psychology 36(8), 1978. Escalating stakes before account closure: Xuan and Shaffer, Journal of Gambling Studies, 2009. Escalation predicting self-exclusion among 32,262 bettors: "Overtime: Long-Term Betting Trajectories Among Highly-Involved and Less-Involved Online Sports Bettors", Journal of Gambling Studies, 2024. Concentration of losses: Rossow, Kesaite, Pallesen and Wardle, "Concentration of gambling spending by product type", Addiction Research & Theory, 2024. VIP programmes: Massachusetts Gaming Commission analysis, September 2025.