Dice control is the claim that a shooter can use grip, starting orientation, arc, rotation, landing zone, and delivery to change the probabilities of final craps outcomes. A careful throw can look repeatable. That appearance is not evidence of a usable casino edge.
The practical conclusion is not that every physical influence is impossible. It is that ordinary players should treat casino craps as random unless a claimed influence survives a controlled, sufficiently large, independently reviewed test under real table conditions—and then produces an advantage on specific wagers after all rules and invalid throws are counted.
That is a much higher standard than rolling a long hand.
Setting the dice is not controlling the result
Dice setting means arranging particular faces before the throw. Dice control means the starting arrangement and delivery continue to affect the final distribution after the dice travel, strike the far end of the table, bounce, collide, and settle.
A shooter can demonstrate the first action by sight. The second claim requires outcome data.
This distinction prevents two bad arguments:
- “The casino lets players set the dice, so the method must work.”
- “The dice bounce, so no physical influence could ever exist under any conditions.”
Neither statement proves the relevant question. The question is whether a repeatable method changes probabilities enough, under permitted casino conditions, to overcome the house edge on an identified betting strategy.
Start with the fair-dice baseline
Two fair six-sided dice have 36 ordered combinations. The total 7 has six combinations:
P(7) = 6 ÷ 36 = 1 ÷ 6 ≈ 16.67%The complete distribution is:
| Total | Combinations | Probability |
|---|---|---|
| 2 | 1 | 2.78% |
| 3 | 2 | 5.56% |
| 4 | 3 | 8.33% |
| 5 | 4 | 11.11% |
| 6 | 5 | 13.89% |
| 7 | 6 | 16.67% |
| 8 | 5 | 13.89% |
| 9 | 4 | 11.11% |
| 10 | 3 | 8.33% |
| 11 | 2 | 5.56% |
| 12 | 1 | 2.78% |
Random play does not produce those percentages neatly in every 36 rolls. It produces clusters, short hands, long hands, repeated points, and stretches with unusually few or many sevens. The dice-combinations page explains the baseline; the variance simulator shows why short sequences can look designed.
Long hands are expected to occur
The chance of avoiding a 7 for n consecutive rolls is:
P(no 7 in n rolls) = (5 ÷ 6)^n| Consecutive non-seven rolls | Probability under fair dice |
|---|---|
| 10 | 16.15% |
| 20 | 2.61% |
| 30 | 0.42% |
A 30-roll stretch without a seven is uncommon for one specified starting point, but casinos run many tables, shooters, and sessions. Someone will eventually produce an attention-grabbing hand. Selecting that hand afterward is not the same as predicting it beforehand.
The seven-to-roll ratio does not settle the question
Dice-control discussions often use a seven-to-roll ratio, commonly calculated as non-seven rolls divided by sevens:
SRR = Non-seven outcomes ÷ Seven outcomesFor fair dice, the long-run baseline is:
SRR = 30 ÷ 6 = 5.0An observed ratio above 5.0 does not automatically prove influence. It may be ordinary sampling variation. It may also be distorted by excluded short rolls, unrecorded invalid throws, changed table conditions, stopping rules, or selective publication of the best sessions.
More importantly, a lower seven rate does not describe which other totals increased. Craps wagers depend on the entire distribution and the game state, not on one number alone.
A simple sample-size check
Let:
- n = number of valid recorded rolls;
- S = observed sevens;
- p = 1/6 under the fair-dice model.
The expected number of sevens is:
E(S) = n × pThe standard deviation of the seven count is:
SD(S) = √[n × p × (1 - p)]Suppose a shooter records 3,600 valid rolls. Fair dice predict:
E(S) = 3,600 × 1/6 = 600 sevensand:
SD(S) = √(3,600 × 1/6 × 5/6) = √500 ≈ 22.36If the record contains 550 sevens, the result is 50 below expectation, about 2.24 standard deviations from the fair-dice mean:
z = (550 - 600) ÷ 22.36 ≈ -2.24That result deserves examination. It still does not prove a casino advantage by itself. The claim may have been chosen after reviewing many totals; sessions may have been stopped selectively; throw conditions may not match casino rules; rolls may not be independent; or the result may fail in a new sample. A proper test states the hypothesis, sample size, valid-roll rules, betting consequence, and analysis method before data collection.
The NIST guidance on sample sizes for proportion tests reflects the same principle: a proportion claim needs a defined hypothesis and enough observations to distinguish a real shift from sampling noise.
Fewer sevens do not help every craps bet in the same way
A global claim such as “I avoid sevens” is incomplete because 7 has different roles during a craps cycle:
- On the come-out roll, 7 wins Pass Line and Come wagers.
- After a point is established, 7 loses Pass and Come wagers.
- Don’t Pass and Don’t Come generally reverse those incentives, subject to their come-out rules.
- Place and buy bets usually lose when 7 arrives before the selected number.
- Any Seven wins only when 7 appears on the next roll.
Reducing sevens could therefore remove some favorable come-out results while reducing later seven-outs. The net effect depends on which totals replace the missing sevens, how often the shooter is in each game state, and which wagers are made.
To claim an edge, a test must move from “the seven rate changed” to a complete expected-value calculation:
EV = Σ [P(outcome i) × Net payoff(outcome i)]Each relevant outcome probability must be estimated, and the payoffs and rules must match the actual table. A high SRR alone is not a betting system.
Casino throw procedure makes the claim harder to demonstrate
Casino rules vary, but regulated procedure is designed to make the throw observable and to prevent placement, sliding, switching, and other improper techniques.
The official Massachusetts craps rules, for example, require the two selected dice to leave the shooter’s hand simultaneously in a manner intended to make them strike the far end of the table. They allow a “No Roll” call when the dice fail to leave simultaneously, fail to strike an end, involve a cheating or fixed technique, or are otherwise considered improper.
Those requirements do not mathematically prove that influence is impossible. They create more rotation, collision, and condition variability, while giving the crew authority to reject nonconforming attempts. A method demonstrated on a short practice surface without equivalent wall contact does not establish performance on a live casino table.
Table conditions can also change:
- surface material and bounce;
- table length;
- back-wall construction;
- dice condition and replacement;
- chips in the landing area;
- required arc and wall contact;
- crew enforcement;
- fatigue and delivery consistency.
A claimed effect that disappears when any of those ordinary conditions changes is not a robust casino advantage.
What credible evidence would look like
A serious test should include all of the following:
- A claim specified in advance. For example, a lower seven rate, not a vague promise of “better numbers.”
- A fixed sample plan. The tester should not stop only when the current results look favorable.
- All attempted throws recorded. Invalid, short, off-table, and rejected throws matter operationally and cannot be hidden.
- Realistic conditions. Table length, approved dice, wall contact, landing surface, and pace should resemble the claimed environment.
- Independent observation. A third party should verify rolls and prevent selective recording.
- Full outcome data. Every ordered dice result is more informative than reporting only sevens or hand length.
- Out-of-sample replication. A second data set should test the result without changing the method.
- A wager-specific EV calculation. Statistical difference is not enough; the shift must overcome the relevant edge and practical betting limits.
- Costs and execution included. Tokes, travel, mistakes, rejected throws, table limits, and bankroll volatility affect whether an apparent edge is usable.
Video highlights, testimonials, one exceptional casino hand, and practice rolls selected for smoothness do not satisfy these conditions.
The operational issue is valid roll control, not debate
Casino staff do not need to decide whether a shooter believes in dice influence. They enforce the approved throw and protect the game.
The stickperson controls the dice and watches the delivery. The boxperson or supervisor can address repeated failures, delay, unusual handling, or disputed calls. Dealers protect the betting layout and late-wager cutoff. Surveillance may review suspected sliding, dice substitution, collusion, or repeated nonconforming throws.
A shooter arranging the dice without delaying the game may be treated differently from one who slides them, conceals them, repeatedly misses the required end, or uses an unapproved device. Exact procedure and enforcement vary by jurisdiction and property. Players should follow the crew’s instructions rather than assume that a practice seen elsewhere is permitted.
Why the myth is financially dangerous
The dice-control story can change betting behavior even when it does not change dice probabilities. A player may:
- increase wagers for a shooter who looks skilled;
- buy expensive instruction based on testimonials;
- ignore the cost of high-edge proposition bets;
- attribute wins to technique and losses to distraction;
- keep playing to “find the throw” again;
- exclude failed sessions from personal records.
That pattern turns an unverified claim into higher total exposure. The expected-loss definition shows how wager size and number of decisions convert an edge into money. The safest betting assumption is that the standard probabilities apply.
A practical way to evaluate a claim
Ask five questions:
- What exact outcome probability is claimed to change?
- Was the test designed before the rolls were observed?
- Were all throws and table conditions recorded?
- Did an independent repeat sample confirm the result?
- What specific legal wager becomes positive after payoffs, rules, errors, and costs?
If the seller or shooter cannot answer those questions, there is no reliable basis for increasing a wager.
Dice control remains an interesting physics and statistics question. That does not make it a demonstrated casino strategy. Until credible evidence closes the gap, a smooth delivery should be treated as presentation—not an edge.