Craps Strategy: We Tested 8 Popular Systems Over 100,000 Simulated Sessions

Key takeaways from 100,000 simulated sessions

• Every one of the 8 craps systems tested lost money on average.
• Pass Line-based play was the cheapest, at about $6 to $8 per 200-roll session, with or without odds.
• The Iron Cross was the most expensive at $94.62 a session, more than eleven times flat Pass Line.
• Odds bets lower the house edge per dollar but add swings, not savings.
• The Martingale ended in profit 48.0% of the time, yet its median session was minus $190.

Every craps strategy we tested lost money. All eight of them, over 100,000 simulated sessions each, dealt the exact same dice. That’s the honest headline and also the boring one. The interesting question isn’t which system wins, because none of them do. It’s how differently they lose.

Some cost about $8 a session. One cost nearly $95. One finished in profit almost half the time and still posted a median of minus $190. Below is the full results table, the reasoning behind the numbers, and why the system that feels best to play is the worst thing in our dataset.

How Did We Test 8 Craps Systems?

For this article we built a Monte Carlo model of a craps table and ran every system through identical dice, so the differences between them come from the bets rather than from luck. Every number below traces to these parameters.

Methodology

• Scale: 100,000 simulated sessions per system, 8 systems, custom simulation code, run on September 29, 2026.
• Bankroll: $500 to start, $10 table minimum.
• Session: 200 rolls of new betting. After roll 200, Place bets come down and are returned, line and Come bets resolve over up to 80 further rolls, and anything still pending is returned at face value.
• Common random numbers: all 8 systems see the same dice sequences, which cuts the noise in comparisons but doesn’t remove it. Seed 20260929, NumPy default_rng, two six-sided draws summed per roll.
• Field rule: 2 pays 2:1, 12 pays 3:1, the friendlier of the two common rules; 2:1/2:1 would look worse.
• Place and Come bets: Place bets work every roll including come-outs and are re-placed after every seven-out. Come-bet odds are off on the come-out, standard casino practice.
• “Could not afford next bet”: before roll 200 the bankroll could no longer cover the bet the system required. At that point the player stops betting and walks away with whatever is left, so it measures the plan breaking, not a zero balance.
• Empirical house edge: total net loss divided by total wagered, across all 100,000 sessions.
• Sanity check: the empirical edges land within simulation noise of the textbook values. Flat Pass Line came out at 1.37% against a theoretical 1.41%, Place 6 & 8 at 1.55% against 1.52%, and the Field at 2.79% against 2.78%.

Two limits up front. This is a computer model, not casino play, and the results carry sampling noise, which matters most for the three Pass Line variants: they land within about $2 of each other, and all three sit close to the roughly $8.50 that the math predicts for $600 of Pass Line action. Proposition and hardway bets were left out, so no craps betting strategy built on them appears here.

We didn’t test dice influence either. ThisGenGaming already covered controlled shooting and why it isn’t where a player’s edge comes from, so everything below assumes fair, random dice.

Which Craps Strategy Loses the Least? Results From 100,000 Sessions Each

The three Pass Line systems were the cheapest to play, at roughly $6 to $8 per 200-roll session, and for practical purposes they cost the same. The Iron Cross was the most expensive by a distance, at minus $94.62 a session, more than eleven times the cost of flat Pass Line play.

System	What it bets	Avg net	Median net	% sessions in profit	% sessions down $250+	% could not afford next bet	Avg total wagered	Empirical house edge
Pass Line only	$10 Pass Line every come-out	−$8.20	−$10	42.8%	0.1%	0.0%	$600	1.37%
Pass Line + 2x Odds	$10 Pass + $20 Odds on the point	−$7.59	−$10	48.2%	13.9%	2.1%	$1,394	0.54%
Pass Line + 3-4-5x Odds	$10 Pass + max 3x/4x/5x Odds	−$6.40	−$10	48.0%	29.1%	16.5%	$2,111	0.30%
3-Point Molly (2x Odds)	$10 Pass + up to two $10 Come bets, each with 2x Odds	−$18.11	−$30	47.0%	30.7%	19.8%	$3,099	0.58%
Place 6 & 8	$12 each on 6 and 8, re-placed after every seven	−$22.67	−$22	45.0%	9.8%	0.9%	$1,464	1.55%
Iron Cross	$10 Field + $10 Place 5 + $12 Place 6 + $12 Place 8	−$94.62	−$96	36.7%	30.6%	17.4%	$3,790	2.50%
Field only	$10 Field every roll	−$55.82	−$60	35.2%	12.0%	0.5%	$1,999	2.79%
Martingale on Pass Line	$10 Pass, double after every loss, reset after a win	−$16.19	−$190	48.0%	36.7%	50.9%	$1,163	1.39%

A pattern falls out of the table. The three Pass Line variants cluster at the cheap end within about $2 of each other. The odds-heavy versions look slightly cheaper partly because the sessions that went broke stopped betting early and made fewer Pass Line bets, so treat the three as one group rather than a ranking. Place 6 & 8 sits mid-pack on cost but is far calmer than the odds-heavy systems, with only 9.8% of sessions down $250 or more. Iron Cross and Field only are the two clear losers, and that distance is real: both sit well outside the noise around the Pass Line group. The Martingale’s average of minus $16.19 looks survivable. Its median of minus $190 does not.

What “average net” does and doesn’t tell you

An average is a summary, not a forecast. For Place 6 & 8 the average is minus $22.67 and the median minus $22, and they agree because the distribution is tight. For the Martingale the average is minus $16.19 and the median minus $190. Nothing about a typical Martingale session resembles its average, which is propped up by the 48% of sessions that finish around plus $300 before the doubling catches up with them. When average and median disagree that far, distrust the average.

What Are the Best and Worst Craps Bets?

The odds bet is the only wager on a craps table with no house edge at all, which is why it dominates every serious discussion of the best craps bets. It makes your average edge look better without making you lose less money. The Pass Line systems were cheapest because the line bet carries a low 1.41% edge and puts few dollars at risk.

What the odds bet really does

Watch the empirical edge fall as odds go on: 1.37% flat, 0.54% at 2x odds, 0.30% at 3-4-5x. That’s a real improvement in cost per dollar wagered, for a simple reason. A craps odds bet pays true odds, so its expected value is zero, and adding a zero-edge bet alongside a 1.41% bet dilutes the blended percentage. It doesn’t beat anything. Sean Kent’s craps analysis on the University of Wisconsin–Madison statistics server walks through the same derivation and lands on the same 1.41% for the Pass Line.

Where the 6 and 8 actually sit

Place 6 and 8 pay 7:6 and carry a 1.52% house edge. The system finished in profit 45.0% of the time, was down $250 or more in only 9.8% of sessions, and broke its own plan in just 0.9%. So the popular advice to build a craps strategy around the 6 and the 8 holds up as a low-variance choice, not a low-cost one. It cost minus $22.67 a session, nearly three times flat Pass Line, because $12 on each number keeps feeding action into the table. Working out what these bets return on a given stake is easier with a craps payout calculator than with a memorized table.

The bets to leave alone

A Field bet in craps looks generous because it covers seven numbers, yet it posted a 2.79% empirical edge and the worst profit rate in the test at 35.2%. For reference, the standard craps house edge on the bets that matter: Pass Line 1.41%, Don’t Pass 1.36% (not part of our test), taking odds 0%, Place 6 or 8 at 7:6 is 1.52%, Place 5 or 9 at 7:5 is 4.00%, Place 4 or 10 at 9:5 is 6.67%, and the Field is 2.78% when 12 pays 3:1. Proposition and hardway bets sit above all of these. We didn’t simulate them, so we won’t put a number on them.

Which Systems Lost the Most? Iron Cross and Field Only

The Iron Cross lost $94.62 per session on average, more than eleven times flat Pass Line play, and finished in profit in only 36.7% of sessions. It was the single worst result in the test, and the confidence interval around it doesn’t come close to overlapping anything else.

The mechanic explains both the appeal and the damage. Iron Cross craps covers every number except the 7, so almost every roll pays something back, which is genuinely what players like about it. But 6 of the 36 two-dice combinations total 7, making it the most frequent single number on the table, and when it lands it takes the Field bet and all three Place bets at once. The system pushes $3,790 through the table at a blended 2.50% edge. Multiply those and you get about $95 a session, the bill for all that coverage.

Field only is cheaper but not better. It cost minus $55.82 a session at a 2.79% edge and holds that worst-in-test 35.2% profit rate. It rarely broke a bankroll, though: only 0.5% of sessions hit an unaffordable bet, because a flat $10 field bet in craps never escalates. Losing steadily and losing suddenly are different failure modes, and this is the steady one.

Why “covering the table” is not coverage

Covering more numbers raises how often you win and how much is at risk when you lose, and those two effects aren’t symmetrical. Frequency of winning is a feeling. Total money wagered against a house edge is the bill. A system that pays small amounts constantly and takes a large amount occasionally isn’t protecting you, it’s spacing out the invoice.

Does a Lower House Edge Mean a Smaller Loss?

The 3-Point Molly ran a 0.58% empirical house edge, better than flat Pass Line’s 1.37%, and still lost more than twice as much per session: minus $18.11 against minus $8.20. The reason is exposure. It pushed $3,099 through the table where flat Pass Line pushed $600.

The arithmetic is one line: what you lose is the house edge multiplied by the total amount you wager. For the Iron Cross that’s 2.50% of $3,790, about $95. For the Molly it’s 0.58% of $3,099, about $18. A small edge on a lot of action can cost more than a bigger edge on a little, which is why house edge alone can’t tell you what a session will cost.

Now the variance half. Pass Line plus 3-4-5x odds had the best edge in the test at 0.30%, and it was down $250 or more in 29.1% of sessions against 0.1% for flat Pass Line. It ran out of affordable bets in 16.5% of sessions against 0.0%. Same underlying game, radically different ride.

So if you’re after an optimal craps strategy, “optimal” splits into two decisions that pull against each other. Lowering your edge per dollar and lowering your total risk aren’t the same goal, and odds bets improve the first while worsening the second. All three Pass Line variants cost roughly the same $6 to $8 a session, because the odds bet adds no expected loss either way. What it adds is swing.

Does the Martingale Work? What 100,000 Sessions Say About How to Win at Craps

The Martingale finished 100,000 sessions with a median result of minus $190 and hit a bet it could not afford in 50.9% of them. Half the time, the plan simply ran out of bankroll before roll 200.

And yet 48.0% of Martingale sessions ended in profit, among the highest rates in the test, and 47.4% finished up $100 or more. That combination is why the system feels like it works. Each winning cycle recovers the earlier losses plus $10, and those cycles add up to sessions that finish near plus $300. The average of minus $16.19 and the median of minus $190 are both true, and the gap between them is where the catastrophic sessions live.

The doubling arithmetic is unforgiving with a $10 base and a $500 bankroll. Five straight losses run $10, $20, $40, $80, $160, which is $310 gone and $190 left. The sixth bet in the sequence is $320, and it can’t be covered. Five consecutive Pass Line losses are not rare over 200 rolls, which is how you get to 50.9%.

The most telling number is the one that didn’t move. Martingale’s empirical edge came in at 1.39%, essentially identical to flat Pass Line’s 1.37%, because it’s the same bet in a different order. That’s the general lesson about every progression in martingale craps and beyond: a staking plan rearranges the shape of your outcomes, not the cost of playing. The best Martingale session in 100,000 was plus $540, and the worst sessions lost the entire $500 bankroll. Because a broken Martingale stops betting and keeps what’s left, its average of minus $16.19 is, if anything, a flattering number.

Why Practice a System Before You Bet On It?

A system’s shape is the part these tables can only describe. How often it swings, how fast it escalates, how a seven-out feels when four bets come down at once, that’s what a player needs to experience before money is involved, and it’s why craps strategy for beginners should start away from a live table.

A free craps simulator is built for exactly that. It’s manual play with a $1,000 virtual bankroll, and it supports Pass and Don’t Pass, Come and Don’t Come, odds, the Field, Place 4 through 10, hardways and proposition bets. Run the Iron Cross by hand for fifty rolls and the 6-in-36 arithmetic stops being abstract.

One clarification on provenance: the 100,000-session results above came from a separate batch script, not from that tool. A manual craps strategy simulator is the right instrument for feeling out a system’s rhythm. For distributions, you need a batch run.

Craps Strategy FAQ

What is the best craps strategy?

None of the eight systems we tested was profitable, so the best craps strategy is the one that costs least for the fun you get. Across our 100,000 simulated sessions, Pass Line-based play was cheapest at about $6 to $8 per 200-roll session. Adding odds kept that cost and added bigger swings.

What is the 3/4/5 rule in craps?

The 3-4-5x odds rule caps your odds bet at 3x the line bet on 4 and 10, 4x on 5 and 9, and 5x on 6 and 8. Those multiples are set so the payout is a flat multiple of the line bet whichever point is on. In our test this was the lowest-edge configuration of all eight, at 0.30% (the textbook figure is about 0.37%; ours runs lower because sessions that went broke stopped betting).

What is the most profitable craps strategy?

None of the eight systems was profitable on average across 100,000 simulated sessions. The cheapest to play were the three Pass Line variants, at about $6 to $8 per 200-roll session, close enough to count as the same cost. Adding odds did not change that cost, only the size of the swings around it.

How likely are you to roll a 7 in craps?

Six of the 36 possible two-dice combinations total 7, so the probability on any given roll is 6/36, or 1 in 6, as Richland Community College’s craps probability lesson lays out sum by sum. That makes 7 the most frequent single total in craps, which is why it appears in almost every rule on the layout, as a winner on the come-out and a loser once a point is set.

Why bet $6 on 6 and 8 in craps?

Place 6 and Place 8 pay 7:6, so a $6 bet divides cleanly and wins $7 with nothing rounded down. Any stake that isn’t a multiple of $6 loses value to rounding. Our simulation used $12 on each number, which pays $14 on a win and keeps the same clean ratio.

What is the riskiest bet in craps?

Judged by our data rather than opinion, the Martingale broke its own bankroll most often, hitting an unaffordable bet in 50.9% of sessions, while the Iron Cross lost the most money per session at minus $94.62. Proposition and hardway bets carry higher edges than anything we tested and sat outside the test.

What not to do at a craps table?

Three things our data argues against. Don’t chase losses with a doubling progression, since it broke half our Martingale sessions. Don’t cover more numbers than you can re-place after a seven-out. And don’t run the Field as your primary system: it finished in profit in only 35.2% of sessions, the worst rate in the test.

What 100,000 Sessions Actually Taught Us

Four takeaways:

• The odds bet is the only real edge reducer in craps, because it’s the only bet on the layout with a 0% house edge.
• House edge and total loss are different quantities. Loss tracks edge times total wagered, which is how the 3-Point Molly ran a better edge than flat Pass Line and still cost twice as much.
• Progressions change variance, not cost. Martingale’s 1.39% empirical edge was indistinguishable from flat Pass Line’s 1.37%.
• Every system we tested lost money, from about $6 to nearly $95 a session.

Craps is a negative-expectation game by construction, and this test put a price on that. The only bankroll decision fully inside a player’s control is how much they’re willing to lose before they stop. If gambling has stopped feeling like a choice, the National Council on Problem Gambling runs the National Problem Gambling Helpline at 1-800-MY-RESET, by call or text.

The dice don’t remember, the edge doesn’t move, and all any craps strategy changes is how quickly and how loudly you pay it.

Sources

• Monte Carlo simulation of 8 craps systems run for this article, 100,000 sessions each, seed 20260929, run September 29, 2026.
• Sean Kent, “All Bets Are Off: Re-learning the Pass Line Bet in Craps,” University of Wisconsin–Madison statistics server, February 28, 2019 (retrieved September 29, 2026).
• Richland Community College, “Craps” probability lesson (retrieved September 29, 2026).
• National Council on Problem Gambling, National Problem Gambling Helpline (retrieved September 29, 2026).