A craps comp is not a refund for losing. It is a marketing reinvestment decision based primarily on the casino’s estimate of a player’s value.
The cleanest way to think about it is:
Rated action → theoretical loss → reinvestment → comp value
The complication is that craps contains many wagers with different house edges, different resolution speeds, and different rating practices. Two players with the same amount of chips on the layout can therefore receive very different theoretical values.
The core comp model uses estimated play, not the final result
A simplified casino-side model is:
Rated action = average rated wager × rated decisions
Then:
Theoretical loss = rated action × assigned house edge
And finally:
Comp budget = theoretical loss × reinvestment rate
The reinvestment rate is a property policy. It is not a universal percentage guaranteed to every craps player.
Suppose a casino rates a player at $100 of house-edge-bearing action per decision, estimates 50 decisions per hour, and records two hours of play:
Rated action = $100 × 50 × 2 = $10,000
If the assigned blended house edge is 1.5%:
Theoretical loss = $10,000 × 1.5% = $150
If the casino reinvests 30% of that theoretical value:
Illustrative comp budget = $150 × 30% = $45
Every figure except the arithmetic can vary by property. The casino may rate the average bet differently, use another pace assumption, assign a different game factor, exclude or discount certain wagers, or apply a different reinvestment policy.
For the underlying concepts, see theoretical loss and craps rating and comps.
Craps is harder to rate than a single-wager table game
A blackjack player may make one main wager, occasionally adding a double or split. A craps player can have money in several places at the same time.
A single layout might show:
- $25 Pass Line;
- $50 odds behind the line;
- $30 Place 6;
- $30 Place 8;
- a Come bet;
- odds behind the Come bet;
- an occasional proposition wager.
Those chips do not all create the same expected casino win.
The Pass Line has a relatively low house edge. Free odds pay at true mathematical odds and therefore do not contain a built-in house edge. Place bets have their own edges. Proposition bets can be substantially more expensive.
If a rating system simply treated every dollar on the felt as identical theoretical value, it would misstate the economics of the play.
Odds bets create the biggest rating misunderstanding
A player might have $25 on the Pass Line and $100 in odds behind it. Visually, $125 is committed to the point.
But the two components are mathematically different:
- the $25 Pass Line wager has a house edge;
- the $100 odds wager pays at true odds and has zero house edge.
That means the player can increase total dollars physically wagered without increasing theoretical loss in direct proportion.
Casinos handle this in different ways. A property may exclude odds from rated average bet, discount them, record them separately, or use a standard rating convention. That is an operational policy choice, not a mathematical statement that the odds suddenly have a house edge.
This is one reason a player should not assume that “I had $125 working” means the rating system treated the full $125 as equal-value action.
Bet mix changes theoretical value
Consider two players who both appear to have about $120 working on the layout.
Player A concentrates on low-edge line bets and odds.
Player B uses more place, hardway, and proposition action.
Their visible total exposure may be similar, but the theoretical casino value can be different because the wager mix has a different blended edge.
A simplified blended-edge model is:
Blended edge = total expected loss across rated bets ÷ total rated action
Suppose $60 of rated action carries a 1.5% edge and another $40 carries a 4% edge:
- expected loss from first component = $60 × 1.5% = $0.90;
- expected loss from second component = $40 × 4% = $1.60;
- total expected loss = $2.50 on $100 action;
- blended edge = 2.5%.
This is a teaching example, not a universal casino rating formula. Actual systems may use fixed game factors rather than reconstruct every wager decision.
Average bet is an estimate at a live table
A craps supervisor cannot necessarily enter every chip movement made by every player after every roll.
Ratings may therefore depend on periodic observations or system inputs such as:
- average house-edge-bearing wager;
- start time;
- end time;
- position;
- player identity;
- game type;
- assumed or observed pace;
- notable changes in bet size.
If a player starts at $25 and later plays $300, the final rating depends on how accurately that change is observed and recorded.
This is why a player can remember “I was betting $300” while the casino record shows a lower average. The rating is intended to represent the session, not only the highest wager seen.
For the generic rating concept, see average bet.
Pace matters because craps decisions do not resolve like slot spins
The number of dice rolls per hour is not automatically the same as the number of rated decisions for every wager.
Some wagers resolve on one roll. Others remain active across multiple rolls. A Pass Line bet can stay unresolved while the shooter makes several point-cycle rolls. Place bets can win and remain up, be pressed, be taken down, or be off under house procedures.
A comp model therefore needs a sensible convention for converting time and observed wagers into expected action.
This is also why a published “rolls per hour” figure should be treated as an assumption. Table occupancy, player pace, buy-ins, payouts, dice changes, disputes, and prop-bet volume all affect game speed.
See craps expected loss per hour for the exposure side of the same problem.
Actual loss and theoretical loss can move in opposite directions
Suppose two players generate identical rated theoretical loss of $200.
- Player A wins $4,000.
- Player B loses $5,000.
If the casino’s comp policy is theo-based, the starting comp budget can still be similar because both players generated similar expected value.
Properties may use actual loss in special circumstances or blend actual and theoretical measures, but the player should never assume that a large recorded loss is automatically the comp formula.
The distinction matters most after a volatile session. A player who lost heavily may feel that a $100 meal is obviously “owed.” A casino rating system may see a much smaller theo and therefore a much smaller normal reinvestment allowance.
Nevada controls show why “the rating” is not one universal regulated formula
Nevada’s current Table Games Minimum Internal Control Standards contain detailed requirements for computerized player-tracking systems that accumulate redeemable points. The same standards explicitly note that those requirements do not apply to player-rating-only systems where evaluation of play and the choice or dollar amount of complimentaries are solely the result of employee judgment.
That distinction is useful because it shows why readers should not assume one Nevada-wide comp percentage or one mandatory craps rating formula. Controls can govern systems and records while the commercial valuation method remains property-specific. See the Nevada Table Games MICS.
Comp value is not the same as retail value
A casino may describe an offer as a $100 dinner, $250 room, or $500 ticket package. That is the retail or face value presented to the guest.
The casino’s internal cost can be different.
This matters when evaluating whether gambling extra to earn a benefit makes sense. If a player adds $2,000 of negative-expectation action to earn a meal with a $75 menu price, the player has not made the meal “free.” The extra expected gambling cost may exceed the value of the benefit.
The correct comparison is:
Incremental expected gambling cost versus value the player would actually pay for the benefit
Not:
Retail comp label versus zero
A practical way to estimate a craps comp
A player trying to understand an offer can use a transparent estimate:
- estimate the average rated wager, not every chip on the table;
- estimate the number of rated decisions or use time × an assumed pace;
- apply a reasonable house-edge factor for the rated bet mix;
- calculate theoretical loss;
- apply an assumed reinvestment percentage;
- compare the result with the actual offer.
If the estimate differs sharply from the casino’s offer, the most likely explanation is not secret mathematics. It is usually a difference in one of the inputs: average bet, time, pace, bet mix, rating of odds, tier status, reinvestment policy, or discretionary host adjustment.
What a craps comp is really worth
The useful answer has two layers.
From the casino’s side, the comp is worth the amount of reinvestment the property is prepared to return against the player’s expected value.
From the player’s side, the comp is worth what the player would genuinely pay for that room, meal, free play, or other benefit—not necessarily the advertised retail price.
That is why chasing comps is usually a poor reason to increase action. The comp is a percentage of a theoretical gambling cost, not a mechanism that reverses that cost.
The disciplined way to read the offer is therefore: rated action first, theo second, reinvestment third, comp last. Keeping those four steps separate prevents most of the confusion around what a craps comp is actually worth.