Casino Games

A casino game is a designed object. Its margin is not an outcome of play but a parameter fixed before the game reaches the floor, chosen alongside pace and volatility to produce a target revenue per position per hour. Reading these games well means reading them as engineering rather than as contests. This page works through where the edge sits in each major game, the two games where mathematics has genuinely broken the house at some point, why those exceptions closed, and the legal and operational machinery that determines what happens to anyone who finds a remaining one.

House edge, and how a game keeps its margin

A fixed mathematical advantage built into the rules of every casino game, expressed as a fraction of the amount wagered.

Compute the expected value of one decision under the posted rules, then scale by decisions per hour to get a rate.

Edge per decision is not comparable across games, because games differ in pace by an order of magnitude.

The big idea

Comparing any two games, and estimating the cost of a session before playing it.

The house edge is the expected loss per unit wagered, computed from the rules and nothing else. It does not depend on how anyone plays unless the game admits strategy, and it does not change with results. A game with a 1.4 percent edge returns 98.6 percent of the money wagered on it over enough decisions, and the operator’s revenue forecast rests on that arithmetic rather than on luck.

Edges vary widely. Baccarat’s banker bet sits near 1.06 percent, craps pass line at 1.41 percent, single-zero roulette at 2.70 percent, double-zero roulette at 5.26 percent, and keno frequently above 25 percent. The last figure is not a mistake. Some games are priced for a customer who is not comparing.

Element of risk

For games where the amount at risk exceeds the nominal wager, the edge can be stated two ways. House edge conventionally divides expected loss by the original bet; element of risk divides it by the total amount actually put at risk including doubles, splits, and odds.

The distinction matters when comparing a game where additional money goes in after the deal against one where it does not. Quoting house edge for blackjack and element of risk for craps, or the reverse, produces a comparison that is not a comparison.

Edge as a design parameter

The edge is chosen. A game designer selects a target hold, then sets paytables and rules to hit it, and a regulator approves the result. Nothing about the mathematics is discovered during operation.

This is why the phrase “the house always wins” is imprecise in a way that matters. The house wins because the game was built to have the house win, at a rate specified in advance, and the only genuine uncertainty is variance around a known mean.

Blackjack, and the game with a computable optimum

The one table game whose optimal strategy against a fixed rule set can be computed exactly, reducing the edge to a fraction of a percent.

Enumerate every player hand against every dealer upcard and choose the action with the higher expectation.

The dealer plays a fixed rule with no choices, so the player's problem is a finite optimization rather than a game against an opponent.

The big idea

Any blackjack table, and as the reference case for what an exactly solvable gambling game looks like.

Blackjack is unusual because the dealer has no discretion. The dealer’s play is fully specified by the house rules, which means the player is not facing an opponent but a fixed stochastic process. That reduces optimal play to enumeration: for every combination of player hand and dealer upcard, compute the expected value of each available action and take the best.

Baldwin, Cantey, Maisel, and McDermott published the first such computation in 1956 in the Journal of the American Statistical Association, working with desk calculators. Their result showed that correct play brought the game far closer to even than the prevailing belief held, and it established the template every later analysis followed.

Rule variations and their cost

Basic strategy is not one strategy. It is a family, one member per rule set, and the differences are worth real money. Whether the dealer hits soft 17, whether doubling after a split is allowed, how many decks are in play, and above all whether a natural pays 3 to 2 or 6 to 5 each shift the edge.

The 6 to 5 payout is the single largest change in modern casino blackjack. It adds roughly 1.4 percentage points to the house edge, which is more than the entire edge of a good 3 to 2 game. A player using perfect basic strategy at a 6 to 5 table is worse off than a careless player at a 3 to 2 table.

What errors cost

Deviation from basic strategy has an asymmetric cost profile. Most errors are cheap, because the two actions being compared have similar expectation. A few are expensive, and they cluster in the same places: standing on soft 18 against a strong upcard, failing to split eights, taking insurance.

Insurance deserves a specific mention because it is the error most often marketed as prudence. It is a side bet on whether the dealer has a ten in the hole, priced at 2 to 1 when the true odds are worse, and it is a losing proposition independently of the main hand unless the composition of the remaining deck is known.

Splits, doubles, and surrender

Rules that expand a player’s options shift the edge more than most casual players realize, and less than a payout change does. Doubling after a split, resplitting aces, and late surrender each help the player by roughly a tenth of a percent, sometimes less. Cumulatively they can move the combined figure by nearly half a percent, so matching a chart to the specific rule set on the table matters.

Counting cards, and what it actually earns

Tracking the composition of the remaining shoe to detect the intervals in which the player, not the house, holds the advantage.

Assign a value to each card seen, divide the running total by decks remaining, and raise the bet when the normalized count is high.

Removing small cards from a shoe raises the frequency of naturals and of favorable doubling, which shifts expectation toward the player.

The big idea

Understanding advantage play, and calibrating expectations about what it earns per hour.

Thorp published the result behind card counting in the Proceedings of the National Academy of Sciences in 1961, showing that a strategy conditioned on the cards already dealt could produce a positive expectation. The insight is compositional rather than mystical: a shoe depleted of low cards favors the player, because naturals are more frequent, doubles are more often successful, and the dealer busts more often on stiff hands.

The high-low count assigns plus one to cards two through six, zero to seven through nine, and minus one to tens and aces. The running count is divided by the number of decks remaining to give a true count, and the true count maps roughly linearly to edge, with each unit worth in the region of half a percent.

What the numbers actually are

The realistic edge is small. A skilled counter in a good game holds an average advantage somewhere in the range of half a percent to about one and a half percent, available only during the fraction of the shoe where the count is favorable, and realized only if the bet is raised during exactly those intervals.

Against that, the variance is large. The standard deviation of a single blackjack hand is roughly 1.15 units, which means a session’s outcome is dominated by noise over any horizon a person actually plays. Kelly’s criterion gives the sizing that maximizes long-run growth under these conditions, and the fraction it recommends is small enough that meaningful hourly earnings require a bankroll most people do not have.

Why it is a job rather than a trick

Combining the edge, the pace, and the bet spread that a casino will tolerate before acting produces an hourly expectation that is a modest multiple of the average bet. Achieving it requires long hours, sustained accuracy under distraction, and tolerance of losing weeks that carry no information about whether the method is working.

The gap between the mathematics being correct and the activity being worthwhile is the honest lesson here. The mathematics has been correct and public since 1961. The activity is difficult, poorly paid relative to its skill requirement, and actively opposed by the counterparty.

Thorp’s 1961 PNAS paper

Thorp’s paper worked out in full a strategy keyed to the fives: once every five has been dealt, the player raises the bet and switches to a computed strategy with an expectation of 0.0329. It also outlined a stronger strategy keyed to the proportion of tens remaining, under which the player has an advantage almost half the time, and deferred its details to later work. His book Beat the Dealer generalized the approach into counts a person could actually run at the table. The mathematical result stood unchallenged. The operator’s response, principally multi-deck shoes, the cut card, and reshuffles on pit signal, reshaped the game the result applied to.

Roulette, and a game with no exploitable structure

A game of independent spins where the edge comes entirely from the zero pockets and no betting pattern can alter it.

Thirty-six numbered pockets pay 35 to 1, and the additional zero pockets are the entire margin.

Independence across spins means no history carries information about the next outcome, so no strategy conditioned on history has value.

The big idea

As the clean counterexample to progression systems, and to any claim that a betting pattern can beat a negative-expectation game.

Single-zero roulette has 37 pockets and pays 35 to 1 on a straight number, giving an edge of 1 divided by 37, or 2.70 percent. Double-zero roulette has 38 pockets and the same payout, giving 2 divided by 38, or 5.26 percent. Every bet on the layout carries that same edge, with the exception of the five-number bet on a double-zero wheel, which is worse at 7.89 percent.

There is no structure to exploit because spins are independent. The wheel has no memory, and the sequence of past results carries no information about the next one. This is not a claim about how much information there is; there is none.

Why progressions cannot work

A progression is a rule for varying the stake based on history. The Martingale doubles after a loss, the d’Alembert increments, and many others exist. Each changes the distribution of session outcomes, and none changes the expectation.

The reason is linearity of expectation. The expected value of a sum of bets is the sum of their expected values, and if every individual bet has negative expectation then so does any combination of them, regardless of how the stakes are chosen. A progression trades a high probability of a small win for a small probability of a large loss, and the trade is priced exactly at the house edge.

Physical prediction, and its conditions

The one genuine exception is physical. A wheel with a mechanical bias, or a prediction of the ball’s landing region from its velocity and deceleration, can in principle beat the game, and both have been done. Thorp and Claude Shannon built a wearable computer for this purpose in the early 1960s.

The conditions are demanding and largely historical. Modern wheels are manufactured to tight tolerances, monitored for bias by the operator, and rotated. Devices are illegal in Nevada under a specific statute and in most other jurisdictions. What remains is a real physical effect that is no longer practically accessible.

Independence and the gambler’s fallacy

The intuition that a run of one color makes the other color due for the next spin is the gambler’s fallacy in its purest form. Independence means the marginal distribution over pockets does not update on history, whatever the history is. A sequence of ten reds in a row occurs with probability of about one in 1,350 on a single-zero wheel, or one in 1,760 on a double-zero wheel, and carries the same predictive content for the next spin as a single red does, which is none. Displayed histories on the wheel exist because they sell bets, not because they inform them.

Craps, and the spread between its best and worst bets

A single table carrying bets from a fraction of a percent to over sixteen percent, offered simultaneously to the same player.

Line bets carry a low edge, the odds bet behind them carries none, and the proposition bets in the center carry the highest edges on the floor.

The layout is designed so that the worst bets are the most visually prominent and the best bet is not marked at all.

The big idea

Reading a craps layout, and as a demonstration that game selection within a game matters as much as between games.

The pass line carries a 1.41 percent edge and the don’t pass 1.36 percent, both among the better bets in a casino. The proposition bets in the center of the layout run far higher: any seven pays 4 to 1 against true odds of 5 to 1, giving an edge of 16.67 percent, and the hardways and one-roll propositions sit between those extremes.

Both sets of bets are available to the same player at the same table at the same moment. The difference between playing the line and playing the center of the table is larger than the difference between most pairs of casino games.

Odds, and the only bet with no edge

After a point is established, a player may back the line bet with an odds bet paid at true odds, meaning 2 to 1 on a point of 4 or 10, 3 to 2 on 5 or 9, and 6 to 5 on 6 or 8. The house edge on this portion is exactly zero.

Because the odds bet dilutes the edge across a larger total wager, a player taking maximum odds lowers the combined edge substantially: at 3-4-5 times odds, the blended figure on the pass line falls to roughly 0.37 percent. The bet is not marked on the felt and is not offered by the dealer, which is a design decision rather than an oversight.

Why the worst bets are in the middle

The center of the layout is the area the stickperson manages and calls attention to, and it carries the highest edges on the table. That arrangement is deliberate. Proposition bets are small, frequent, resolved in one roll, and priced for a player who is responding to the pace of the game rather than to the paytable.

This generalizes beyond craps. Within almost every casino game, the bets that are promoted, announced, or visually emphasized carry higher edges than the ones that are not.

Proposition bets and their pricing

Hardways, hop bets, and one-roll propositions are priced by paying less than true odds on outcomes the shooter can produce. Any craps pays 7 to 1 against true odds of 8 to 1, giving an edge of 11.11 percent. Big red on the any-seven pays 4 to 1 against true odds of 5 to 1, and hard six or hard eight pays 9 to 1 against true odds of 10 to 1. Each price is set slightly below the fair number, and each bet is priced for a customer watching the game rather than the numbers.

Baccarat, commission, and the largest table game

A table game with among the lowest house edges available, which nonetheless generates the majority of table revenue in several major markets.

Both hands are drawn by fixed rules with no player decisions, and the banker's small structural advantage is taxed by commission.

A low edge combined with very high stakes produces more revenue than a high edge at low stakes.

The big idea

Understanding where casino table revenue actually comes from, and the economics of high-limit play.

Baccarat has no decisions. Both the player and banker hands are drawn according to a fixed table, and the participant chooses only which side to back. The banker hand wins slightly more often because it draws last with knowledge of the player’s third card, and the standard 5 percent commission on banker wins exists to tax that advantage back.

After commission, the banker bet carries a house edge of about 1.06 percent and the player bet about 1.24 percent, both low by casino standards. The tie bet, paying 8 to 1, carries roughly 14.4 percent and exists to catch the difference.

Revenue concentration

Baccarat dominates table revenue in Macau, on the gaming regulator’s own published breakdown of win by game, and is the largest single table game in several Asian markets, with high-limit play concentrated among a small number of participants. Where a low-edge game is played at very high stakes, the resulting revenue exceeds that of higher-edge games played at ordinary stakes.

That concentration also changes the operator’s risk. A business whose revenue comes from a small number of very large positions has variance that a business built on many small positions does not, and the largest properties manage this explicitly.

Edge sorting and the Ivey litigation

Edge sorting exploits asymmetries in the printed pattern on the backs of playing cards. If cards can be rotated so that high cards are distinguishable from low ones by their back edges, a participant gains information about the next card, which in baccarat is enough to reverse the edge.

Phil Ivey used this technique and litigated the result on both sides of the Atlantic. In Ivey v. Genting Casinos UK Ltd, decided October 25, 2017, the United Kingdom Supreme Court held that his conduct amounted to cheating within the meaning of the contract and the Gambling Act 2005, notwithstanding that he had not touched the cards himself and had used only what the dealer did at his request. The court also restated the legal test for dishonesty in the course of doing so, which is why the case is cited well outside gambling law. He lost in the New Jersey litigation against Borgata as well, where a federal court ruled in 2016 that he had breached his contract and owed about $10 million; the case settled on undisclosed terms while on appeal in 2020.

The doctrinal point is worth stating precisely, because it is often misreported. The courts did not hold that Ivey lied or concealed his method. They held that deliberately arranging the conditions of play to obtain information the game was not designed to provide constituted cheating regardless of his belief that it was legitimate advantage play.

Commission and the banker’s edge

Without a commission, the banker bet would carry roughly a 1.24 percent advantage to the player because of the drawing rules, which let the banker react to the player’s third card. The 5 percent commission on winning banker bets taxes that gap back to a 1.06 percent house edge. Some tables market a no-commission variant that pays even money on banker except for a haircut on a winning banker six, and the resulting edge is comparable with a different distribution.

Slot machines, return to player, and volatility

Machines whose entire mathematical behavior is fixed in a specification sheet before manufacture and approved by a regulator.

A random number generator selects a stop combination, and the paytable and reel layout together determine return and variance.

Return to player is a long-run average over millions of spins and says nothing about the distribution of a single session.

The big idea

Interpreting an advertised return figure, and understanding why two machines at the same return feel entirely different.

A slot machine’s mathematics lives in a PAR sheet, a document produced during design that specifies reel strips, paytable, hit frequency, and theoretical return. Regulators approve it, and the machine is manufactured to it. Nothing about the outcome distribution emerges during operation.

Return to player is the long-run expected fraction of wagered money returned, commonly between 85 and 98 percent depending on jurisdiction, denomination, and venue. It is an average over a number of spins far larger than any individual will play, which makes it a poor guide to what a session will look like.

Volatility and hit frequency

Two machines with identical return can have entirely different session distributions. Hit frequency is the proportion of spins returning anything at all; volatility describes the spread of outcomes. A low-volatility machine returns small amounts often, and a high-volatility machine returns almost nothing most of the time and occasionally a great deal.

Since return is fixed, volatility is the free design parameter, and it is set to produce a target session experience. The industry has moved toward higher volatility over time, which is a choice about how the same expected loss is distributed across players.

Near misses and the regulatory line

A near miss is a displayed outcome that appears close to a win, such as a jackpot symbol landing just above the payline. Where reel strips are weighted so that these appear more often than the underlying probabilities would produce, regulators have treated the practice as deceptive, and Nevada and other jurisdictions have restricted virtual reel mapping that manufactures the effect.

The distinction regulators draw is between a near miss that arises naturally from the reel layout, which is permitted, and one engineered to appear at a rate unrelated to the actual probability of the win, which is not. The line is about whether the display represents the underlying process honestly.

PAR sheets and regulatory certification

A PAR sheet lists every reel stop, every paytable line, and the resulting probabilities and returns. Manufacturers submit it, along with source code, to an accredited testing laboratory such as GLI or BMM, which verifies that the shipped game behaves as the sheet claims. Regulators in Nevada, New Jersey, and comparable jurisdictions certify the combined package, and audited hold that drifts materially from the certified figure is grounds to pull the machine.

Video poker, and machines with a positive return

Machines whose return can be computed exactly from the paytable, and which at some paytables returned more than one hundred percent.

Enumerate all draws from every dealt hand, choose the holding with the highest expectation, and compute the exact return.

The deck is a known finite object with no hidden state, so the expectation is a finite sum rather than an estimate.

The big idea

Reading a video poker paytable, and understanding why exactly computable games get withdrawn.

Video poker is the one machine game whose return is exactly computable by anyone. The deck composition is known, the draw is uniform, and the paytable is printed on the machine. For each of the 2,598,960 possible dealt hands there are 32 holding decisions, and evaluating all of them gives the exact expected return under optimal play.

Because the calculation is exact, paytables can be compared precisely, and small differences in the payout for a single hand type change the return by measurable amounts. Full pay Jacks or Better, paying 9 for a full house and 6 for a flush, returns about 99.54 percent. Reducing those to 8 and 5 drops it to roughly 97.3 percent while changing nothing else on the machine.

Positive paytables and their disappearance

Some historical paytables returned above 100 percent under optimal play. Full pay Deuces Wild is the standard example, returning slightly over 100.7 percent to a perfect player. These machines existed in quantity in Nevada for years.

They have largely disappeared, and the reason is straightforward: they were unprofitable against players who played them correctly, and the population of such players grew as the analysis became public. A game that can be solved exactly cannot retain an edge once the solution circulates.

Cost of imperfect play

The return figures assume optimal holding decisions on every hand, and optimal play in video poker is genuinely difficult. Strategy charts run to dozens of ranked holdings, and the correct decision on marginal hands is frequently counterintuitive.

Ordinary players give up somewhere between half a percent and several percent to strategy error, which is more than the difference between a good paytable and a mediocre one. The exactness of the mathematics does not transfer to the player unless the player is exact.

Promotions, comps, and advantage outside the game

The marketing layer attached to casino games, where most genuine advantage play actually occurs.

Value each offer as an expected return, subtract the expected loss from any required play, and take the difference.

Comps are computed from theoretical loss rather than actual loss, so the calculation is fully specified in advance.

The big idea

Evaluating a match play or an online bonus with a wagering requirement.

Casinos compute player value as theoretical loss, meaning average bet multiplied by decisions per hour multiplied by hours multiplied by house edge. Comps are issued as a fraction of that figure, typically somewhere between a fifth and a half depending on the property and the game.

Because theoretical loss uses the house edge rather than the actual result, the comp calculation is deterministic and known in advance. That makes it computable by the player as precisely as by the operator, which is the basis of most advantage play that does not involve beating a game.

Valuing a match play

A match play coupon pays the bet twice on a win but is surrendered either way. On an even-money bet with probability p of winning, the coupon’s value is approximately p times the coupon face value, less the house edge on the underlying bet.

On a near even-money proposition the coupon is worth close to half its face in expectation. It is one of the few offers whose value can be computed in a single line, and it is a good illustration that the arithmetic of promotions is usually simpler than the arithmetic of the games.

Wagering requirements

An online bonus typically carries a requirement to wager some multiple of the bonus before withdrawal. The expected cost of meeting it is the required turnover multiplied by the house edge of the permitted game, and the offer is positive when the bonus exceeds that cost.

The terms matter more than the headline figure. Game weighting, maximum bet limits during the requirement, and time limits each change the calculation, and the games permitted at full weighting are usually the ones with the highest edge. An offer that looks positive against blackjack is often not available against blackjack.

Penetration, shuffling, and countermeasures

The operational defenses against advantage play, which are cheaper to deploy than the techniques are to execute.

Reduce the fraction of the shoe dealt, shuffle continuously, and monitor bet variation to identify who is counting.

Every technique that requires a specific game condition can be defeated by removing that condition.

The big idea

Assessing whether a game is beatable in practice rather than in principle.

Penetration is the fraction of the shoe dealt before a shuffle, and it is the single largest determinant of whether counting is worth anything. A count only becomes reliably informative deep into a shoe, so a game cut at half is worth far less than one cut at three quarters, and below a certain penetration the edge disappears entirely regardless of skill.

Continuous shufflers

A continuous shuffling machine returns played cards to the shoe, which eliminates the depletion that counting depends on. Against it, the count carries no information at all.

The device costs the operator something real: it changes the game’s feel, and it increases the hands per hour, which cuts both ways since it also increases the player’s expected loss per hour. Operators weigh those against the advantage-play exposure of the particular game.

Bet spread as the identifying signal

A counter must raise the bet when the count is favorable, which produces a correlation between bet size and shoe composition that is visible to anyone tracking it. The bet spread, meaning the ratio between the largest and smallest bets, is the primary signal used to identify counters.

The economics of concealment follow directly. A wide spread earns more and identifies faster; a narrow spread earns less and survives longer. Flat betting earns nothing. Every technique for disguising the correlation, whether cover plays, team structures, or shortened sessions, reduces the earn rate in exchange for tenure, and the trade is unavoidable rather than a matter of skill.

Penetration as the cheapest defense

Moving the cut card from three quarters of the shoe to half of it removes most of the value of counting at the cost of a few hands per hour of dealer productivity. No new equipment is bought and no player is identified. The operator can adjust penetration by pit, by shift, or by table, and reverse the change without ceremony. Cheap defenses that require no attribution tend to be deployed first.

Decisions per hour, and the real cost of a game

The rate at which a game resolves, which converts an edge per decision into a loss per hour and differs across games by an order of magnitude.

Multiply house edge by decisions per hour by average wager.

Pace is a design lever, and increasing it raises revenue without changing any advertised number.

The big idea

Comparing games honestly, and evaluating any change that speeds a game up.

Edge per decision is the number usually quoted and the less useful of the two. Multiply it by the number of decisions per hour and by the average wager, and the result is expected loss per hour, which is what the player actually experiences and what the operator actually forecasts.

Decision rates differ enormously. A crowded blackjack table might deal 60 to 80 hands an hour to a given seat, a heads-up table over 200, a roulette wheel perhaps 40 to 60 spins, a craps table around 100 rolls with far fewer resolutions per player, and a slot machine 600 to 800 spins.

Multiplying through, a slot at 6 percent and 600 spins an hour on a 1 unit bet costs 36 units an hour. Blackjack at 0.5 percent and 70 hands an hour on a 100 unit bet costs 35 units an hour. A twelvefold difference in the advertised edge produces the same hourly cost, because pace and stake absorb all of it.

Speed as a revenue lever

Raising pace increases revenue proportionally without altering the paytable, the return figure, or anything a regulator reviews as a fairness question. Faster spin cycles, automatic shufflers, side bets that resolve without slowing the main game, and multi-hand play all operate on this margin.

Pace is the lever the industry has pulled hardest over the past three decades.

This is the clearest case in the industry of a change that is invisible in the disclosed statistics and material in the outcome. A machine advertising 96 percent return is not comparable across two pace regimes, and pace is not on the glass.

Why a low edge can cost more

A slot machine at 6 percent edge running 600 spins an hour costs far more per hour than a baccarat hand at 1.06 percent running 70 hands an hour, at the same stake. The edge is often the smallest term in the hourly-cost calculation. A player choosing between games on edge alone, without considering pace and stake, will routinely choose the more expensive option.

Decision rates by game

The ranges below are for busy tables and full machines, and shift with crowd, dealer, and specifically with side-bet participation, which resolves alongside the main game and multiplies the effective decision rate.

Game	Decisions per hour	Typical edge	Loss per hour, 10 unit average
Slot machine	600 to 800	6 percent	360 to 480 units
Video poker	400 to 600	1 percent	40 to 60 units
Double-zero roulette	40 to 60	5.26 percent	21 to 32 units
Baccarat, banker	60 to 80	1.06 percent	6 to 8 units
Blackjack, basic strategy	60 to 80	0.5 percent	3 to 4 units

Backoffs, exclusion, and the legal position

The legal and operational position of a player who is winning by skill, which in most jurisdictions permits the casino to refuse service.

Casinos rely on common law rights to exclude, subject to state statutes and civil rights limits.

Counting uses only public information and no device, so it is not a crime, but the casino is generally under no obligation to deal to anyone.

The big idea

Understanding what actually happens to a successful advantage player, and where the criminal line sits.

Card counting is not a crime in Nevada or in any other US jurisdiction that has considered it squarely. It involves observing cards that are dealt face up and performing arithmetic. No statute prohibits thinking about a game in progress.

What follows from that is narrower than it sounds. In most states a casino may refuse to deal to anyone, and the standard remedies are a backoff, meaning a request to stop playing a particular game, or exclusion from the property enforced through trespass law. The player wins the legal question and loses the game anyway.

New Jersey as the exception

Uston v. Resorts International Hotel, Inc., 89 N.J. 163, decided in 1982, is the significant departure. The New Jersey Supreme Court held that Resorts could not exclude Ken Uston for counting cards, on the ground that the Casino Control Act gave the Casino Control Commission exclusive authority over methods and rules of play, displacing the common law right to exclude in that domain.

The reasoning is jurisdictional rather than a general endorsement of advantage play. The court did not hold that counters have a right to play anywhere; it held that in New Jersey the legislature had assigned control over the rules of casino games to a regulator, and a casino could not enact its own rule by exclusion. Casinos in New Jersey responded with countermeasures within the rules, principally shuffling and penetration, which remain available to them.

Where the criminal line sits

The line moves sharply once a device is involved. Nevada Revised Statutes 465.075 makes it a felony to use any electronic or mechanical device to assist in projecting the outcome of a game, keeping track of cards played, or analyzing probabilities or strategy. Similar provisions exist in other states.

The distinction is therefore not between honest and dishonest play but between unaided cognition and instrumented assistance. Counting in one’s head is lawful; the identical calculation performed by a concealed computer is a felony. Marking cards, past posting, and colluding with an employee are separately criminal in every jurisdiction.