The Fire Bet is a shooter-long craps side bet. It wins when one shooter establishes and then makes several different point numbers before sevening out. The six numbers that can count are 4, 5, 6, 8, 9, and 10.
A long roll by itself is not enough. Repeating the same point can be exciting for line and odds bettors, but it does not add another Fire Bet number. The wager rewards variety: four, five, or all six unique points, depending on the posted paytable.
What must happen for a number to count
A point enters the Fire Bet record only after a complete point cycle:
- The shooter establishes 4, 5, 6, 8, 9, or 10 on a come-out roll.
- The shooter rolls that point again before rolling 7.
- The dealer marks that point as made for the Fire Bet.
Suppose a shooter establishes 6 and makes it. The 6 is now one unique Fire Bet point. If the shooter later establishes and makes 6 again, the Fire Bet total remains one. If the next completed point is 9, the total becomes two.
Come-out naturals and craps do not create Fire Bet points. They normally leave the shooter in control of the dice and the side bet alive. A seven that occurs after a point has been established ends the hand and resolves the Fire Bet according to the number of unique points already made.
The bet must generally be placed before that shooter’s first come-out roll. It cannot be added after the player has already seen a promising hand develop.
A sample hand
Consider this abbreviated sequence:
- Come out 6; shooter makes 6. One unique point.
- Come out 8; shooter makes 8. Two unique points.
- Come out 6; shooter makes 6 again. Still two.
- Come out 10; shooter makes 10. Three unique points.
- Come out 5; shooter makes 5. Four unique points.
- Come out 9; shooter seven-outs before making 9. The hand ends with four unique points.
If the posted schedule begins paying at four, the wager is settled at that level. It does not receive separate awards for one, two, and three points along the way.
Why the six point numbers are not equally difficult
Once a point is established, only that point and 7 matter to the eventual resolution. Other totals merely delay it.
The point-making probability is:
[ P(\text{make point }p)=\frac{w_p}{w_p+6} ]
where (w_p) is the number of dice combinations that produce point (p), and six combinations produce 7.
| Point | Ways to roll point | Chance point is made before 7 |
|---|---|---|
| 4 or 10 | 3 | 3/(3+6) = 33.33% |
| 5 or 9 | 4 | 4/(4+6) = 40.00% |
| 6 or 8 | 5 | 5/(5+6) = 45.45% |
Making 4 and 10 is therefore harder than making 6 and 8. The dice-combination guide shows why those totals have different weights.
The Fire Bet is more complicated than multiplying six independent point probabilities. The next established point may be one already made, and repeats create self-loops in the process. Exact analysis must track the full set of completed points after every successful cycle.
Exact frequency of four, five, and six unique points
Using fair dice and standard point procedures, the probability distribution for the final number of unique points is approximately:
| Unique points made before seven-out | Probability | About one in… |
|---|---|---|
| 0 | 59.3939% | 1.68 hands |
| 1 | 26.0750% | 3.83 hands |
| 2 | 10.1275% | 9.87 hands |
| 3 | 3.3434% | 29.91 hands |
| 4 | 0.8798% | 113.66 hands |
| 5 | 0.1640% | 609.78 hands |
| 6 | 0.01624% | 6,156.33 hands |
These figures describe the number of distinct points completed by the time the shooter seven-outs. The six-point result is rare because the shooter must complete every point category, while repeats do not advance the count.
The probabilities also explain why a large top award does not automatically make the bet favorable. A payout must be compared with the full distribution, not with the top result alone.
Why repeat points change the mathematics
The Fire Bet state is not determined only by how many points have been made. It also matters which points are already in the set. A hand with 4, 5, and 6 completed has three unique points, as does a hand with 6, 8, and 9, but the remaining paths are not identical. The first state still needs 8, 9, and 10; the second needs 4, 5, and 10. The difficult 4 and 10 have fewer combinations than 6 and 8.
A complete model therefore uses a set (S) of unique points already made. For each possible next point (p), the process has three relevant branches:
- the point is established and then seven-outs, ending the hand at (|S|);
- the point is established and made, adding (p) to the set if it is new;
- the point is established and made again when (p\in S), leaving the state unchanged.
That third branch is why a simple “six trials” calculation is wrong. A shooter can complete ten or more point cycles and still have only two or three distinct Fire Bet numbers. Repeats lengthen the hand without necessarily moving the side bet closer to its next award.
For practical comparison, the chance of reaching at least four unique points is the sum of the four-, five-, and six-point probabilities:
[ P(N\ge4)=0.8798%+0.1640%+0.01624%\approx1.0601% ]
That is about one paying hand in 94.3 shooter hands for a schedule that starts at four. Most wagers therefore lose in full, while a small minority produce the entire return.
Reading the paytable correctly
Fire Bet schedules are not universal. One table may pay only four, five, and six points; another may include a three-point award; top payouts may differ sharply. Always read the felt or posted rule card.
For illustration, consider a schedule that pays:
| Final unique-point count | Profit payout |
|---|---|
| 4 | 25 to 1 |
| 5 | 250 to 1 |
| 6 | 1,000 to 1 |
| 0-3 | Loss |
If those figures mean profit odds, the expected value of a one-unit wager is:
[ EV=(0.008798\times25)+(0.001640\times250)+(0.0001624\times1000)-0.989399 ]
[ EV\approx-0.1970 ]
The corresponding house edge is about 19.70%. If a schedule uses “for 1” total-return language instead of “to 1” profit language, the edge is higher because the listed amount includes the returned stake. The table’s exact wording matters.
A published Canadian casino rule sheet, for example, lists Fire Bet awards by the number of different points made. See the Loto-Québec Fire Bet rules. New Jersey’s rules also require dealers to mark each different point as it is made; the Fire Bet procedure in New Jersey regulations is a useful operational reference.
Fire Bet versus All Tall Small and Repeater bets
These side bets can appear similar because all remain active across many rolls, but they measure different events.
- Fire Bet: counts distinct point numbers that are established and then made.
- All Tall Small: tracks whether specified totals are rolled before 7; no point cycle is required.
- Repeater: requires selected totals to appear a designated number of times before 7.
A shooter could roll every box number and complete an All Tall Small bet without making four Fire Bet points. Conversely, a shooter could repeatedly establish and complete point cycles without satisfying all the totals required by ATS. The Repeater bet guide covers the separate repeated-total mechanic.
How the casino controls the wager
The Fire Bet adds a persistent record to an already busy game. Accurate handling normally requires:
- accepting the wager only during the approved entry window;
- associating each bet with the correct player position and shooter;
- marking a number only after the point is actually completed;
- ignoring repeat completions for the unique-point count;
- keeping markers visible and synchronized with the hand;
- checking the final payout against the approved table;
- obtaining supervisory verification for large awards when house procedures require it.
The dealer must distinguish “point established” from “point made.” Marking a number at establishment would award credit before the shooter has completed the required event.
Payout language also needs consistency. “25 to 1” means 25 units of profit plus the stake. “25 for 1” usually means 25 units returned in total. A one-unit difference may look small at the first level but becomes material in the theoretical edge and in large settlements.
Expected cost over repeated wagers
House edge applies to the amount staked on the side bet, not to the player’s total craps action. With a 19.70% example edge, one $5 Fire Bet has theoretical loss:
[ $5\times0.1970=$0.985 ]
If a player makes the same $5 wager for 40 shooters, total side-bet action is $200 and theoretical loss is about $39.40. Actual results will normally be either a long run of complete losses or a sudden large award; they will not resemble a smooth $39.40 decline. That difference is variance, not a change in expectation.
Comparing only the maximum prize can also distort the decision. A 1,000-to-1 top award is irrelevant on more than 99.98% of shooter hands. The more frequent four-point payout contributes heavily to the return, so a reduction at that level can change the edge even if the top sign remains unchanged.
What the large top number hides
A 1,000-to-1 sign naturally attracts attention. The more useful questions are:
- How often does each paying outcome occur?
- Are awards “to 1” or “for 1”?
- Does the schedule pay only the highest achieved level?
- What is the expected return across all shooter hands?
- How much will be wagered over the session?
The Fire Bet is not a lower-edge way to stay with a hot shooter. It is a high-variance side wager whose price is usually far above the core craps house edges. A small stake can produce a memorable payoff, but its expected cost should be treated as entertainment spending, not as a substitute for Pass, Come, or odds wagers.