Managing a Craps Bankroll

Bankroll management can’t lower the house edge. The pass line still costs 1.41% per bet decided, however you size or order your chips. What a plan controls is how long your money lasts and how much a bad night can take, so I make one before every session. Bring at least 20 line bets for a flat game, far more with odds behind them.

Total bankroll and session bankroll

Your total bankroll is the money set aside for gambling over a trip or a month, cash you could lose without touching rent or savings. A session bankroll is the slice of it you carry to one table for one sitting. Make that split at home, before the first chip changes hands.

Splitting a Vegas trip into sessions

Say you’ve budgeted $1,200 for three days in Las Vegas and plan two sessions a day. I’d cut it into six sessions of $200. By the math further down, $200 fits a flat $10 pass line for an assumed three hours, and a disastrous first night costs you one envelope instead of the whole trip.

The same $1,200 looks much thinner once odds go behind the line. A $10 line with double odds wants about $550 for those same three hours, so the trip covers two of those sessions with $100 left over. Neither plan is wrong. You only need to know which trip you’re taking before the plane lands.

I split the cash into labeled envelopes and leave every envelope but tonight’s in the room safe. Chips I never bought can’t be lost at the rail.

Splitting a month of play

A monthly budget works the same way. Take $320 for four Saturday sessions, and each Saturday gets $80. That fits a flat $5 pass line for about two hours at my assumed pace of 30 line decisions an hour. A month with five Saturdays doesn’t earn a fifth envelope. The $320 is the wall.

What an hour of craps costs on average

Average cost per hour is the bets decided in that hour, times the bet, times the house edge. A $10 pass line, at an assumed 30 decisions an hour, costs 30 × $10 × 1.414%, or $4.24. Odds behind the line add nothing to that figure. Extra come bets and place bets add their own share.

The pace I assume

Every hourly figure depends on how fast the table runs, and tables vary. I assume a live game throws about 100 rolls an hour. A pass line bet needs about 3.4 rolls on average to settle, so my assumed 100 rolls produce roughly 30 line decisions. About 70 of those rolls come with a point on, the stretch when place bets work. Time your own table for ten minutes and scale my figures to what you count.

Pass line edge

1.414% for each bet decided (1.41% rounded), counted from the 36 dice combinations, or roughly 14 cents per decision on $10

Four common patterns priced

The formula holds for any spread of bets. Count how many of each bet get decided in an hour, multiply by the size and by that bet’s edge, and add the results.

Pattern, $10 tableAssumed decisions an hourAverage cost an hour
$10 pass line30$4.24
$10 pass line with $20 in odds30 line decisions, odds at 0%$4.24
$10 line plus two $10 come bets70 (30 line, 40 come)$9.90
Place 6 and 8 at $12 apieceabout 43$7.78
Average cost per hour at my assumed 100 rolls an hour

The come bet row assumes you keep two come bets going whenever a point is on. By my rough count that adds about 40 come decisions to an assumed hour. Each one costs the same 14 cents as a line decision, so 70 decisions × $10 × 1.414% comes to $9.90.

Place bets take one more step. With a point on, a place 6 is decided on 11 of the 36 combinations, five that pay it and six 7s that take it. Over my assumed 70 point-on rolls an hour, that’s 70 × 11 ÷ 36, about 21.4 decisions for the 6 and the same for the 8. Multiply 42.8 decisions by $12 and by 1.515% and the hour costs $7.78.

What a three-hour session costs

Three hours at my assumed pace is about 90 line decisions. A flat $10 line costs $12.73 on average over that stretch, and so does the same line with double odds. Your actual result will land a long way from $12.73. How far is the next question.

How far a session can swing

The swing is measured by standard deviation, the typical distance between one night’s result and the long-run average. For flat even-money bets it’s close to the bet times the square root of the decisions. A $10 line over 90 decisions swings about $95, against an average loss under $13.

Flat line bets

A pass line bet ends each decision $10 up or $10 down. Its variance per decision is $10 squared, 100, less a sliver for the edge (0.14 squared, or 0.02). Variances add across independent decisions, so 90 decisions give 8,998, and the square root of that, $94.86, is the standard deviation. Thirty decisions give $10 × √30, about $54.77.

On a bell curve, and a few dozen decisions already come close to one, about two nights in three land within one standard deviation of the average. About 19 in 20 land within two. For 90 decisions of a flat $10 line, that puts two nights in three between roughly −$108 and +$82, and 19 in 20 between −$202 and +$177.

Line plus odds, counted from the 36 combinations

Odds make each decision lumpier, so the shortcut stops working and the swing has to be counted. Over 990 average rounds, the come-out’s 36 combinations settle 220 rounds as naturals and 110 as craps. The 4 and 10 set 165 points and a third of them get made, 55. The 5 and 9 set 220 and 40% get made, 88. The 6 and 8 set 275, and 5 in 11 of those, 125, get made.

Round ends withRounds in 990Net resultNet squared
Natural (7 or 11)220+$10100
Craps (2, 3 or 12)110−$10100
Point of 4 or 10 made55+$502,500
Point of 5 or 9 made88+$401,600
Point of 6 or 8 made125+$341,156
Point lost to a 7392−$30900
Every round result for a $10 line with double odds, over 990 rounds

Weight the last column by the rounds and you get 220 × 100 + 110 × 100 + 55 × 2,500 + 88 × 1,600 + 125 × 1,156 + 392 × 900 = 808,600. Divide by 990 for 816.77, take off the squared average loss of 0.02, and the variance per decision is 816.75. Its square root is $28.58, almost three times the flat bet’s $10.

Check the money column while you’re there. The rounds net to −$140 over 990, or 14 cents a round, exactly the flat line’s cost. Double odds nearly triple the swing and leave the price alone.

Over 90 decisions the variance grows to 816.75 × 90 = 73,508, and its square root is $271.12. A double-odds player and a flat bettor at the same rail lose the same $12.73 on average over that session. The odds player’s typical night just lands nearly three times further from it.

How many units a session needs

I size a session to cover the average loss plus two standard deviations. For 90 decisions of a flat $10 line, that’s $12.73 + 2 × $94.86, about $200, or 20 units. Double odds push it to about $550, and the place 6 and 8 at $12 apiece to about $380.

Pattern, $10 tableMost on one rollSwing over 3 hoursSession stakeIn $10 units
$10 pass line$10$94.86$20020
$10 line, $20 in odds$30$271.12$55055
Place 6 and 8 at $12 apiece$24$177.74$38038
Session stakes for an assumed three hours (90 line decisions or 210 point-on rolls)

A unit here is your flat line bet, and at most tables I play it’s the table minimum. The pass line lesson covers choosing that flat bet. My concern is how many of them the rack has to hold.

Place 6 and 8 in dollars

Per point-on roll, the pair wins $14 on 10 of the 36 combinations and loses $24 on the six 7s, after which I rebuild both bets. The variance per roll is (10 × 196 + 6 × 576) ÷ 36 = 150.44, less a sliver for the average loss. Over an assumed 210 point-on rolls in three hours, the variance grows to about 150.43 × 210 = 31,591, and its square root is $177.74. Two swings plus the $23.33 average cost come to about $380.

Line plus two come bets

With two come bets and double odds on everything, one seven-out can take $90. All three bets fall to the same 7, so their swings move together instead of evening out. My habit for that spread is eight full sweeps, $720. When the rack drops under half of that, I go back to one come bet. What each point pays on the odds, and the multiple a table allows, belongs to the free odds bet.

How much to bring to a $10 table

For a flat line over an assumed three hours, $200. With double odds, $550. For a shorter sitting, scale the swing by the square root of the decisions. Ninety minutes at my assumed pace is about 45 decisions, and 2 × $10 × √45 plus the $6.36 average cost comes to about $140.

How long a bankroll lasts

Measured by the edge alone, a flat bankroll lasts a long time. A $200 stake at $10 a decision would take about 1,414 decisions to drain on average. Swings end it far sooner on a bad night, so the useful figure is the chance of losing the whole stake within one session.

The average lifetime

Divide the units by the edge. Twenty units ÷ 0.014141 = 1,414 decisions, or about 47 hours at my assumed pace. Half of all $200 stakes are gone within about 534 decisions, and the long survivors drag the average up. It tells you almost nothing about tonight.

The chance of losing the whole stake

Start with a case small enough to check by hand. Bring $30 to a $10 line and the stake dies within three decisions whenever all three lose. The pass line loses 251 decisions in 495, so that’s 0.5071 × 0.5071 × 0.5071, or 13.0%. I carried the same path counting through every possible sequence of an assumed three hours, step by step, with the stake counted as gone once the running loss reached it.

Session stakePatternChance stake is lost
$100$10 pass line33.8%
$200$10 pass line4.7%
$300$10 pass line0.2%
$300$10 line, $20 in odds26.3%
$550$10 line, $20 in odds4.0%
$200Place 6 and 8 at $12 apiece28.3%
$380Place 6 and 8 at $12 apiece4.2%
Chance of losing the whole stake over an assumed three hours

The two-swing stakes from the last section all land near 4% to 5%, about one session in 20 to 25. A $100 stake for a flat $10 line dies a third of the time, and I’d call that a short session in disguise.

How long $300 lasts

Flat at $10, a $300 stake almost never dies in an assumed three-hour session, about one time in 435. Put double odds behind the same line and it’s gone about one session in four, at the same 14 cents a decision. The odds didn’t raise the cost. They raised the stakes on every point.

Even with the right stake, you can’t steer how a 90-decision flat session ends. With $200 in the rack you finish ahead 40.5% of the time, dead even 8.3% and behind 51.2%.

Loss limits and win goals

A loss limit and a win goal are stopping rules. They decide when you leave, and they shape how your nights look, but they never change the 1.41% you pay on each decision. I use both anyway, because a plan made at home holds up better than one made at 2 a.m.

Setting a stop-loss

My stop-loss is the session stake itself. When the envelope is empty, the session is over, and I don’t visit the cage for more. Quitting at half the stake and keeping the rest for another night works as well, provided it’s set before the buy-in.

Is a stop-loss worth it? Not as a money-saver, since every decision you do play still costs the same 1.41%. It caps the worst night, and the worst night is what wrecks a trip. Chasing a loss, an oversized buy-in and a creeping table minimum each get priced in common craps mistakes, and a fixed stop is the plain cure for all three.

Pricing a win goal

Say you play a flat $10 line, quit if you’re down $100, and quit if you’re up $100. Both exits sit ten units away. The classic ruin formula gives your chance of reaching the goal first as (r^10 − 1) ÷ (r^20 − 1), where r is 251 ÷ 244, the pass line’s losses over its wins. With r^10 at 1.32690 and r^20 at 1.76067, that’s 0.32690 ÷ 0.76067, or 42.98%.

Stretch the goal to +$200 with the same $100 stop and your chance falls to 24.46%. Shrink it to +$50 and it climbs to 61.86%. A small goal gets hit often and a big one rarely.

What’s a good win goal? I like one I’ll reach more often than not, about half my stop-loss. When I hit it, I color up and leave with the stake and the win together. Chips you’ve won become your money the moment the dealer pays them, and a $10 bet made with them still costs 14 cents.

Why the average still doesn’t move

Take the ±$100 plan. You win $100 42.976% of the time and lose $100 the other 57.024%, an average of −$14.05. The matching duration formula puts the average length of that session at 99.35 decisions, and 99.35 × $10 × 1.414% is also $14.05. The loss always equals the action times the edge. Limits only decide how much action you give it.

Locking up part of a win

Another approach keeps you at the table after a good run. Say your $200 stake has grown to $300. Pocket the original $200 plus $50 of the win, and play on with the other $50 as a fresh stop-loss. The worst case from there is a $50 profit for the night, and the best case stays open. The locked chips are safe from the swings. The chips you keep betting still cost a $10 bettor 14 cents per decision. I put the locked chips in a different pocket so they never drift back into the rack.

Betting progressions and the house edge

Every progression rearranges bet sizes, and each bet inside it still carries its own edge. Your average loss is the total you wager times that edge, whatever order the bets come in. The Martingale, the Paroli, 3-2-1 regressions and press-and-pull all change the shape of a night and leave its price alone.

The Martingale meets the table maximum

The Martingale doubles the bet after each loss and drops back to the base after a win, so each cycle tries to net one base bet. Table limits are what stop it. Pennsylvania requires the minimum and maximum wagers to be posted at each craps table under 58 Pa. Code § 623a.3(c), and it bars any wager over that maximum, with odds as the one carve-out. Picture a $10 pass line at a table posted at $10 to $2,000.

Losses in a rowNext betLost if it losesChance of reaching it
0$10$10100%
1$20$3050.71%
2$40$7025.71%
3$80$15013.04%
4$160$3106.61%
5$320$6303.35%
6$640$1,2701.70%
7$1,280$2,5500.86%
The Martingale doubling sequence on a $10 pass line at a $10 to $2,000 table

Lose the $1,280 bet too, and the next doubling calls for $2,560, which is $560 over the posted maximum. The cycle ends $2,550 down with no legal way to double again. Eight straight pass line losses come up 0.5071 to the eighth power, or 0.437% of the time, about once every 229 cycles.

The other 99.56% of cycles each win $10. Average the two and you get 0.9956 × $10 − 0.00437 × $2,550, a loss of $1.19 per cycle. The average cycle also puts $84.07 in action, and $84.07 × 1.414% is the same $1.19. Across an average 229 cycles, the doubling buys about 228 small wins and one loss worth 255 of them, at the line’s price.

The odds carve-out doesn’t rescue it, because a Martingale player doubles the flat bet, and the flat bet is what the posted cap stops. You’d also need $2,550 behind a $10 bet to run the full sequence once.

Paroli and pressing after wins

The Paroli flips the Martingale. You double after each win and reset after three wins or any loss. Starting at $10, a cycle averages $10 + 0.4929 × $20 + 0.4929 × 0.4929 × $40, or $29.58 in action, and costs $29.58 × 1.414%, about 42 cents. It feels thrilling when it lands. Per dollar it costs exactly what a flat bet does.

Should you bet more when you’re winning? Only if you’d have bet that size anyway. A shooter’s run of made points doesn’t tilt the next decision, and a raise only raises what each decision costs you.

3-2-1 regression and press-and-pull

A regression starts big and cuts back after a win, such as $36 on the 6 dropping to $24 and then $12, aiming to lock in an early profit. Press-and-pull alternates pressing a place bet after one hit and pocketing the win after the next. The calls themselves are covered with place bets. For your bankroll, each one hands the house 1.515% of every dollar decided on the 6 or 8, in any order you choose.

Picking a pattern that wins most nights is easy. A pattern that risks a lot to win a little will do it. None of them can change the average, because the dice settle each decision without knowing what you bet on the last one.

Tracking your play

A session log turns memory into numbers. Write down the date, the venue, the table minimum, your betting pattern, the hours played, the buy-in and the cash-out. After a dozen sessions you can set your results beside the expected cost and see whether your limits held.

What to logWhy it helps
Buy-in and cash-outThe session’s result, without the rounding memory adds
Hours and decisions playedLets you price the session at your own table’s pace
Pattern and bet sizesCatches a $10 line that crept to $15
Whether your limits heldThe discipline check that matters most
What a craps session log should hold

Expect the log to wander. Twenty sessions of 90 flat $10 decisions carry an average cost of 20 × $12.73, or $254.60, with a swing of $94.86 × √20, about $424.23. A winning log after twenty nights still turns up about one time in four, and it proves nothing about the method. Keep logging anyway. Over a hundred sessions the average starts to show through.

Your log also measures your own pace. Divide the decisions you played by the hours, and the result beats my assumed 30 an hour for every hourly figure above. My own count at a crowded Saturday table has come in well under that, and a quiet weekday game well over it.

I keep mine in a notes app and fill it in before I leave the casino, while the numbers are still honest.

Bankroll rules for online craps

Online craps charges the same edge per decision, but a computer game settles them much faster than a live crew, so each hour costs more. Shrink the bet or the session, and set your account’s deposit and time limits before the first roll.

Faster pace, same formula

Say a computer game gives you an assumed 90 line decisions an hour against my assumed 30 at a live table. A flat $10 line then costs 90 × $10 × 1.414%, or $12.73 an hour, a whole assumed three-hour live session’s cost, and the hourly swing grows to about $95.

I budget online sessions in decisions instead of minutes. Ninety decisions is 90 decisions whether it takes an assumed three hours live or about an hour on a screen at my assumed online pace, and the $200 stake from the table above fits either. The live dealer streams I’ve tried ran closer to casino speed, and the other gaps between formats are laid out in online versus live craps.

Deposit limits, time limits and cooling-off breaks

Pennsylvania’s responsible gaming self-limit rule, 58 Pa. Code § 814a.2, requires its licensed online casinos to offer deposit and spending limits by the day, week or month. It also requires the option of a single-wager cap, a daily time limit and a suspension of at least 72 hours. Lowering a limit takes effect by your next login. Raising one waits until the earlier limit’s period has run out and you confirm the change.

That delay is the useful part. I set my monthly deposit limit at my monthly craps budget, so the envelope plan from the top holds even late at night with a phone in my hand. Pennsylvania’s daily time limit lets you finish the round in progress, and then the session is over whether you’re up or down.

When to take a break entirely

Stop for weeks, not minutes, when the plan stops holding. Warning signs include playing with money meant for bills, raising the stake after every losing session, keeping results out of the log or away from family, and breaking a limit you set yourself.

A cooling-off suspension or a longer self-exclusion from your account settings or through your state’s program takes the decision out of a tired moment. I’ve taken a month off after a trip where I moved my own stop-loss twice. The tables were all still there when I came back.

Up next, taking the dice

With the money side settled, the last piece of the felt to learn is the dice themselves. Being the shooter covers getting the dice, the line bet a shooter makes and how a throw has to land.

If a term there is new, the index of how to play craps lessons starts from the basic rules.

Craps bankroll FAQ

Is one long session cheaper than several short ones?

No. Cost follows total decisions, so six one-hour sittings and one six-hour marathon cost the same at the same pace. Short sittings do give you more natural places to stop, and they keep fatigue from creeping into your bet size.

Where do unspent chips from a cut-short night go?

Back into the total bankroll, not into the next session. If Friday’s $200 envelope comes home holding $260, the extra $60 rejoins the trip money and Saturday still gets $200. Fattening the next buy-in lets one good night raise your bet size by stealth.

Can a couple share one craps bankroll?

Yes, if each player draws a separate session stake from the shared total. Two players betting the same side of the same dice win and lose together, so a shared rack swings harder than either stake would alone, and neither of you can tell whose limit just broke.

Should a gambling budget get its own checking account?

I keep one. A separate checking account funded once a month at your budget turns the monthly total into a wall instead of a figure in your head. My deposits to a licensed online casino come from that account and nowhere else.

A bankroll plan only works if every dollar in it is money you can afford to lose. If you catch yourself borrowing, hiding results or moving the stop-loss mid-session, call 1-800-MY-RESET (on a keypad, 1-800-697-3738). The person who answers will go through it with you, free and confidential.

Steve Venclauskas

Steve Venclauskas

Updated October 29, 2026