The Brutal Math Behind Two Dice Combinations
Thirty-six. That is the only number you need burned into your skull when you sit down at a craps table or fire up a digital dice game. Two standard six-sided dice produce exactly 36 possible outcomes, not 11, not 21, and certainly not whatever random figure the guy next to you at the casino bar just mumbled. The moment you understand that the 36 outcomes are not equally distributed across the sums, you stop being a mark and start thinking like an auditor. π²
The sample space for a pair of fair dice is a Cartesian product: 6 faces on die A multiplied by 6 faces on die B equals 36 ordered pairs. Ordered is the critical word here. A roll of (3, 4) is a fundamentally different physical event from (4, 3), even though both land on a sum of 7. Bookmakers and software engines count them separately because the probability mass depends on that distinction. Strip away the ordering and you collapse the entire mathematical foundation of two dice probability into a puddle of bad assumptions.
Why the Sum of 7 Dominates Everything
The distribution of sums across those 36 combinations is a perfect bell curve, and it is ruthlessly asymmetrical. The sum of 7 appears in 6 of the 36 combinations: (1,6), (2,5), (3,4), (4,3), (5,2), and (6,1). That is a 16.67% probability, the single highest of any sum. The sums of 2 and 12 each appear only once, giving them a pathetic 2.78% chance. This is not a design flaw. It is the entire economic engine of craps, Sic Bo, and every digital dice game that has ever been coded.
Any operator who does not internalize this distribution is bleeding money or, worse, getting bled. The house edge on a pass line bet in craps is 1.41%, and that number is derived directly from the 36-combination matrix. The odds on a hard way bet, the payout on a field bet, the true odds on a place bet β every single one traces back to how many of those 36 ordered pairs satisfy the winning condition. If you are auditing a dice game for exploitable gaps, you start by mapping the payout table against the 36-cell grid and hunting for cells where the payout exceeds the true frequency. That is the only edge that matters.
The Ordered Pair Trap and Software Exploitation
Digital dice games introduce a variable that physical tables cannot: the random number generator. A properly seeded RNG using a uniform distribution over 1 to 6 for each die will reproduce the 36-combination matrix faithfully over a large sample. But flawed implementations exist, and they are more common than the industry wants to admit. Some engines use a single random integer from 2 to 12 and then map it to a visual representation. That approach destroys the natural frequency of each sum and creates a detectable bias. π°
If the engine generates a uniform integer from 2 to 12, then the sum of 7 has the same 9.09% probability as the sum of 2. That is a catastrophic deviation from the true dice combination distribution. A sharp player who tracks outcomes over a few thousand rolls can detect this within hours. The expected frequency of a 7 should be 16.67%. If you are seeing anything close to 9%, you have found a broken game, and broken games are ATMs for anyone paying attention.
Building the 36-Cell Grid in Your Head
Memorize the frequency table cold. Sum 2: 1 way. Sum 3: 2 ways. Sum 4: 3 ways. Sum 5: 4 ways. Sum 6: 5 ways. Sum 7: 6 ways. Sum 8: 5 ways. Sum 9: 4 ways. Sum 10: 3 ways. Sum 11: 2 ways. Sum 12: 1 way. That sequence β 1, 2, 3, 4, 5, 6, 5, 4, 3, 2, 1 β sums to 36. Every bet on a dice table is a wager on one or more of these frequencies. The true odds of rolling a 7 before a 6 are 6-to-5. The true odds of rolling a 4 before a 7 are 3-to-6, or 1-to-2. If the payout does not reflect these ratios, you are paying a tax for the privilege of playing.
Crunch the numbers yourself. Map the 36 cells. Audit the payout. The only thing standing between you and a positive expectation is the gap between the true frequency and the offered odds. π―