Flat betting, and why it is not a flawed system

"Flat betting" (also called "level stakes") means staking the same amount on every bet, whatever happened on the previous ones, and in many lists of systems it appears near the bottom as a "flawed system" because it does not recover losses quickly. That criticism, in our view, blames flat betting for not being a martingale. It is the one approach that promises nothing, and for that reason we treat it as the benchmark against which every other system on this site is measured.

What it does, and what it does not promise

Take a hypothetical player with $200 for an evening, who bets $5 on red at a European wheel for two hundred spins. The "turnover" (everything that crosses the table) is about $1,000, and the expected loss is 2.70 per cent of it, about $27, which is the "price" of the evening and is known before the evening begins.

The result, of course, is not the price. By our exact computation the player ends ahead about one evening in three (32.5 per cent), exactly level about one in twenty and behind in the rest; the whole $200 is lost before the last spin only about once in 77 evenings (1.3 per cent). The typical spread around the $27 is about $70 either way, which is the honest measure of the "variance" that a flat bettor accepts.

Nothing in the method pretends to change those numbers, and nothing can. The stake, the number of spins and the wheel set the cost, and a player wanting it smaller has three levers that work (a lower stake, fewer spins, or a game with a lower edge, which the house edge page gives game by game) and no lever that does not.

Why it is the benchmark

For a given amount wagered over a given number of bets, equal stakes give the narrowest spread of results, a plain consequence of the way variances add up (the spread grows with the sum of the squared stakes, which is smallest when the stakes are equal).

A player who puts $1,000 through a hundred spins at a flat $10 has a spread of about $100 either way (one "standard deviation", the usual measure of spread), while one alternating $5 and $15 for the same total has about $112, and the expected loss is the same $27 in both cases.

The progressions, as we have counted them on the other pages of this section, widen the spread in a direction of their choosing, and most of them also raise the average stake, which raises the "bill". Over the same two hundred spins, a martingale starting at $5 under a $500 limit (with far more than $200 behind it, since its full run of stakes costs $635) puts about $3,690 across the table instead of $1,000, so its expected loss is about $100 instead of $27, as the page on negative progressions explains for the whole family; flat betting is simply the version in which the bill and the shape are both in plain view.

The usual criticism, and why it misses

The criticism usually made of flat betting is that a player who falls behind has no way to "catch up" quickly (to win back in a few bets what was lost in many), since the stake never grows to recover the losses. It is true, and it describes the absence of exactly the mechanism that makes the martingale dangerous: "recovery" is what the negative progressions promise, and the rare large loss is how they pay for it.

No staking pattern recovers losses in expectation, because every dollar staked to recover costs 2.70 per cent like every other dollar. Flat betting does not recover because nothing does, and we would read its refusal to "chase" as its main virtue rather than a defect, since it is the reason the bill of the evening cannot grow behind your back.

The other extreme: one bet

If the honest variable is the turnover, the conclusion that follows is counter-intuitive, and we prefer to state it rather than leave it out. The hypothetical player could put the whole $200 on red in a single spin, and the expected loss would then be 2.70 per cent of $200, about $5.41 (one thirty-seventh of the stake), five times less than the $27 of the flat evening!

The price of that saving is the widest possible swing, since the player leaves with $400 or with nothing, the first about 48.6 times in 100 (eighteen chances in thirty-seven). It is also a very short evening; a player who comes for the hours of play buys almost none of them, which is why we present the "all-in" as the arithmetical extreme and not as advice.

Bold play, and what Dubins and Savage proved

The same arithmetic answers a different question, the one a player asks with a "target" in mind: how to turn $200 into $400 before stopping. The table gives the chance of reaching $400 before losing the $200, for each flat stake, with the average amount wagered on the way.

Turning $200 into $400 with even-money bets on a European wheel

Stake per bet	Chance of reaching $400 first	Average amount wagered	Expected loss
$200 (one bet)	48.6%	$200	$5.41
$50	44.6%	$797	$21.54
$25	39.4%	$1,576	$42.59
$10	25.3%	$3,652	$98.70
$5	10.3%	$5,873	$158.74

The $5 player, the most cautious of the five, pays almost thirty times as much as the bold one for a chance almost five times smaller!

The mathematicians Lester Dubins and Leonard Savage showed, in a book published in 1965 under the candid title "How to Gamble If You Must" (subtitled "Inequalities for Stochastic Processes"), that in a game of even-money bets, each with a negative expectation, "bold play" gives the best possible chance of reaching a fixed target: at each turn, stake everything, or only what the target still requires if that is less. It is not the only strategy with that chance, but with $200 and a target of $400 it is exactly the single bet in the first row.

The lesson is not that you should bet everything at once. It is that the cost of gambling grows with the money crossing the table, so the slow and cautious-looking methods (small stakes, many spins, long progressions) are the expensive ones, and a player wanting many hours of play should choose them knowing the hours are what the turnover buys.

The honest variable

Every page of this section comes back to one quantity, the "total wagered", and flat betting keeps it in plain sight: stake times spins (multiplied by the edge, it gives the price), known before the first bet. The Kelly criterion says not to bet at all, the all-in buys the least play for the least expected loss, and the progressions, in our reading, buy a different shape for a larger bill. None of these choices is free.

If you play, our suggestion is to choose the stake from the budget and the hours and to keep it, whatever the last result. With $200 and two hundred spins (three to four hours at a live table), $5 a spin costs an expected $27 and loses everything about once in 77 evenings, while $10 a spin loses everything about one evening in four. Why no other arrangement of the same stakes moves the price is explained on the page on betting systems.

Questions

By the Rand Casinos editorial team · Last updated October 2026 · Online since 2007