Dice setting is the practice of arranging the dice to a specific starting position before the shooter throws them, as part of a broader technique some players call controlled shooting or precision shooting — the claim that a practiced, consistent throw can shift the odds of craps away from pure randomness. It's one of the most argued-about topics in the game, and this page isn't going to settle it. What it will do is lay out the technique itself, the research that exists on both sides, and how you can test it yourself if you're curious, without telling you which side is right.
What Dice Setting Actually Is
Setting the dice means arranging them to a specific configuration of faces before the throw — for example, positioning both dice so the 2, 3, 4, and 5 show on the four side faces, leaving the 1 and 6 as the top and bottom. That particular arrangement is often called a hardways set, because it puts hard 4, 6, 8, and 10 (matching pairs) in front of the shooter and, in theory, makes a 7 harder to land if the dice stay on that same axis when they land. Other sets are built to favor different numbers — one common set aims to increase the frequency of 6 and 8 specifically, since those are the most profitable place numbers on the layout.
Setting the dice is only step one. The second, more contested step is "controlled shooting" or "precision shooting" — gripping the two dice together so they move as a single unit, then throwing them with a soft, consistent arc and backspin so they land just short of the back wall, touch it gently, and (the theory goes) stay on the same axis of rotation instead of tumbling randomly. Setting the dice without a controlled throw does nothing on its own — a die that tumbles freely ends up random regardless of where it started. Every serious account of the technique, skeptical or not, agrees on this much: the grip and release have to be genuinely repeatable, throw after throw, or there's nothing consistent left to measure in the first place.
The Physics Argument for Controlled Shooting
The case for controlled shooting rests on a simple mechanical claim: if a die is thrown so that it only rotates around one axis, several of its six faces can never end up on top, which changes the probability of each total compared to a fully random toss. Proponents compare it to a golfer's putt or a bowler's release — a skill that takes real practice but is, in principle, learnable and repeatable.
Casinos are built around defeating exactly this. The back wall of a craps table is lined with a pyramid-patterned rubber diamond surface specifically designed to make the dice bounce unpredictably no matter how they arrive, and the rule requiring both dice to hit that wall exists partly to prevent a soft, controlled slide from skipping the randomizing bounce altogether. Whether a human throw can survive that wall consistently enough to preserve any bias is the entire question, and it's the point where the physics argument and the practical evidence start to pull apart.
The Published Research
The most cited real-world test came from Stanford Wong, a well-known professional gambling author who was openly skeptical of dice control until 2004, when he took a two-day seminar with Golden Touch Craps and came away convinced enough to write a book on it, Wong on Dice. Wong later took part in a widely discussed live test: a 500-roll session that appeared to show a controlled shooter beating the random expectation for sevens, followed by a second, 1,000-roll session with the same shooter that went the other way. Added together, the combined 1,500 rolls came out close to what pure randomness would predict — a result usually cited by skeptics as evidence against control, and by believers as evidence that one session, even a large one, isn't enough to prove anything either way.
The standard metric in this research is the ratio of total rolls to sevens rolled, often called the SRR (sevens-to-rolls ratio) or RSR depending on the source. A purely random shooter should land on 6 rolls per seven, on average. Believers point to shooters who sustain a ratio above 6.3 or so over large samples as evidence of skill; skeptics point out that with a small enough sample, plenty of random shooters will land above that number by chance alone.
The most rigorous attempt to settle the question mechanically came in 2020, when researchers built a purpose-designed dice-throwing machine, nicknamed "Lucky Lil," specifically engineered to reproduce the biomechanics controlled shooters describe — on-axis rotation, backspin, a consistent throwing angle — with a precision no human hand can match. Filming with high-speed cameras to confirm the throws were landing on-axis as intended, the researchers recorded 7,557 throws on a real craps table and ran chi-squared tests against random expectation. The machine failed to produce a statistically significant deviation from randomness. Skeptics read that as strong evidence against the whole concept — if a machine built to execute the technique perfectly can't produce a measurable edge, a human hand seems unlikely to do better. Believers push back on the comparison itself, arguing that a machine replicating fixed mechanical parameters isn't the same thing as a trained shooter's feel and adjustment over thousands of live throws, and that the debate hasn't ended there — statisticians have continued publishing formal models for how a test like this should even be structured, which tells you the underlying question is still considered open enough to be worth studying.
The Skeptic Position
The most prominent skeptical voice on this topic is Wizard of Odds, the gambling-math site run by mathematician Michael Shackleford. His stated position isn't an outright dismissal — he's said publicly that he doesn't rule out the possibility of dice control, and that he respects some of the people who believe in it, including Wong. But he's also run his own small-scale tests, describes the results as landing close to random expectation, and has pointed out a structural problem with testing this at all: distinguishing genuine skill from random variance requires an enormous number of rolls, likely in the tens of thousands, and casinos aren't set up to let anyone run that kind of controlled experiment at the table. His broader point is that extraordinary claims call for correspondingly rigorous evidence, and in his assessment, nobody has produced it yet.
The Practitioner Position
Dice-setting instruction is its own small industry, with schools like Golden Touch Craps and authors like Frank Scoblete and the writer known as Sharpshooter teaching the technique through books, seminars, and coaching. Their position is that dice control is a real, learnable physical skill, comparable to any other precision-throwing discipline, and that it simply takes far more practice than casual players assume — often cited as 10,000 to 20,000 practice throws before a shooter's numbers become meaningful. Practitioners typically track their own SRR over these large practice samples as their evidence, and some use third-party software built to analyze whether a throw is landing on-axis, treating a sustained, favorable SRR as proof of skill rather than luck. From inside that community, the 2004 Wong conversion is frequently cited as a case of a genuine skeptic changing his mind after firsthand instruction, not as a marketing story.
How to Practice Dice Setting at Home
If you want to try developing a controlled throw, the practice itself is simple to describe even if it's hard to execute consistently. Set the dice to whichever configuration you're working on, hold them together so they move as one unit rather than two, and throw with a soft, consistent arc and backspin toward a soft surface — many practitioners use a bed, a folded blanket, or a dedicated craps practice pad rather than a hard table. The goal on each throw is the same: land the dice gently, keep them together, and watch whether they land on the same axis you started them on.
You don't need a casino to test any of this seriously, and you don't need to build anything custom either. Practice felts and diamond-backed practice pads made specifically for home dice work are widely available and inexpensive, and they're built to mimic the pyramid-textured back wall a real casino table uses to randomize a bounce. A plastic folding table works fine as the base — stable, cheap, and easy to fold away between sessions. Lay the felt across the top, prop the diamond backstop at the far end, and you've got a setup close enough to a real table to make your practice numbers worth something, all for a modest cost compared to what dice-setting seminars and coaching typically run.
The one variable entirely within your control, whatever you believe about the outcome, is how you pick the dice up and set them in your hand — the same grip, the same finger placement, the same starting orientation, every single time. Skeptics and believers disagree about almost everything else in this debate, but neither side argues that an inconsistent grip helps. If your hand does something slightly different on every throw, you've introduced a variable nobody can account for, and any data you collect afterward is measuring that inconsistency as much as anything else.
Log every throw. Track how many total rolls you've made and how many came up seven, and watch that ratio over a large sample rather than judging any short run. A diamond-backed pad gets you much closer to real table conditions than a bare blanket, but even inside the dice-setting community, there's an honest caveat worth repeating: no home setup is an exact match for a specific casino table's own rubber diamond backstop, worn in its own particular way, and some practitioners themselves note that numbers built up at home don't always transfer perfectly once you're throwing on an actual table. Treat home practice as a way to build consistency in your grip and release, not as a guarantee that your at-home numbers will hold up in a casino.
Tracking Your Results with the Simulator
Whatever you conclude about the technique, the statistics behind testing it are the same statistics behind any craps question, and the simulator is a useful tool for understanding them even though it generates purely random rolls. Run a large batch of simulated rolls — a few thousand, if you want a meaningful comparison — and track the sevens-to-rolls ratio the same way you would at a real table. That gives you an honest picture of what pure randomness actually looks like at the sample size you're working with, including how much a random ratio can wander above or below 6 over a few hundred or even a few thousand rolls.
Then compare that baseline to your own logged results from real practice or real play. If your real-world ratio sits inside the range the simulator shows for random samples of the same size, you don't have evidence of anything beyond ordinary variance — which is exactly the trap both the Wong test and the machine study were built to avoid by using much larger samples than most players ever log. If you want to understand the underlying game better before you get into any of this, How to Play and the glossary cover the fundamentals, and the strategy page covers approaches that don't depend on any of this being true.
The Honest Conclusion
Here's where the evidence actually sits: a widely cited live test came out close to random. A machine built to execute the technique with better precision than any human also came out close to random. A community of practitioners with their own tracked results and their own converted skeptics says otherwise, and a careful, well-known skeptic says the sample sizes needed to prove it either way are larger than almost anyone has produced. Neither side has a smoking gun, and that's not a dodge — it's genuinely where the data is. What isn't in dispute is that a repeatable grip and release are the starting point for testing any of this seriously — without that, you're not measuring a technique, you're just measuring noise. If you're curious, the only honest way to find out is to try it yourself, throw the same way every time, track a large enough sample to mean something, and let the numbers say what they say.