CRA 325: Why Low House Edge Still Loses Money

A low house edge reduces the average cost of each dollar wagered. It does not cap the amount a player can lose, make short sessions predictable, or turn a negative-expectation bet into a winning system.

In craps, three separate questions must be answered:

1. Price: What house edge applies to each wager?
2. Volume: How much total money is put into action?
3. Path: How widely can the actual result swing before the session ends?

Players often answer only the first question. The casino result is driven by all three.

House edge is a rate, not a bill

The basic expected-loss formula is:

$$ \text{Expected loss}=\text{Total action}\times\text{House edge} $$

A $10 Pass Line bet has a standard edge of $7/495$, or about 1.414%. Its expected loss per completed decision is:

$$ $10\times\frac{7}{495}=$0.1414 $$

The bet does not gradually surrender 14 cents. It ultimately wins $10 or loses $10. The 14 cents is the long-run average across many completed wagers.

If the player completes 100 separate $10 Pass Line decisions, total flat action is $1,000:

$$ $1{,}000\times1.414%=$14.14 $$

A $14.14 expected loss does not mean the player is likely to finish exactly $14 down. It states the center of a wide distribution of possible results.

The craps house-edge guide compares the price of individual wagers. This page explains why choosing a better-priced bet is only the beginning of cost control.

Count every bet, not the label you give your style

A player may describe the session as “Pass Line with odds” while also making Field bets, hardways, and one-roll propositions. Expected loss is additive:

$$ E(L)=\sum_{i=1}^{k}A_i h_i $$

Where:

• $A_i$ is total action on wager type $i$;
• $h_i$ is that wager’s house edge;
• $k$ is the number of wager categories used.

Consider this simplified session:

The low-edge Pass Line contributes less than 7% of the expected cost in this example. The repeated Any Seven action dominates the session even though each individual wager may be small.

This is why “I start with a $10 line bet” does not describe the real exposure. The table records every chip that goes to work.

Roll count and decision count are not the same

A Pass Line wager can remain unresolved through several rolls. A hardway may also stay active until its pair, easy total, or 7 appears. A Field or Any Seven bet resolves on one roll.

Estimating cost from “rolls per hour” without matching each wager to its decision pattern can be misleading.

• A $10 Pass Line bet placed once does not create another $10 of flat action on every neutral roll.
• A $5 Field bet repeated on every roll creates $5 of new action each time.
• A Place bet left up creates a new win-or-seven race after each win if the wager remains working.
• Odds are committed only after a point is established.

The dedicated expected-loss-per-hour guide shows how to convert actual betting frequency into an hourly estimate. Use a range when table speed and bet frequency are uncertain rather than presenting false precision.

Zero-edge odds can still produce the largest swing

Pass, Come, Don’t Pass, and Don’t Come odds are paid at true mathematical odds. Their expected value is zero before rounding or unusual restrictions.

That statement answers the price question. It does not answer the path question.

Suppose a player has:

• $10 Pass Line;
• $30 odds;
• $18 Place 6;
• $18 Place 8.

A seven-out can remove all $76 on one roll:

$$ $10+$30+$18+$18=$76 $$

The $30 odds wager adds no expected loss, but it increases the amount that can move on the point resolution. The Place bets also share the same losing event. These positions are not independent; they are positively correlated around 7.

The craps variance guide models these swings in detail. The useful distinction is:

• house edge measures average price;
• variance and standard deviation measure dispersion;
• exposure measures dollars at risk on a resolving event;
• correlation identifies wagers that can lose together.

A lower blended percentage can accompany a larger possible loss.

Why one session can ignore the average

A fixed $10 Pass Line decision wins with probability $244/495$ and loses with probability $251/495$. The average result is about $-0.1414$, but the standard deviation of one completed decision is almost $10.

For 100 equal, independent completed decisions, a rough session standard deviation is:

$$ SD_{100}\approx$10\sqrt{100}=$100 $$

Expected loss over those decisions is only about $14.14. The normal short-run swing is therefore many times larger than the mathematical cost.

This approximation assumes fixed stakes and independent completed decisions. Real craps play often includes overlapping wagers, shared seven-out risk, pressing, incomplete bets, and changing odds multiples. Those features can widen or distort the distribution.

The NIST statistical handbook’s measures-of-scale section explains why standard deviation is expressed in the original measurement units while variance uses squared units. In a craps session, dollars of standard deviation are easier to compare with a bankroll than squared-dollar variance.

The long run does not rescue the player

A common misunderstanding says that playing longer allows the low house edge to “settle down” and therefore protects the bankroll.

Longer play makes the average result per dollar more likely to approach the mathematical expectation. At the same time, it creates more total action and therefore more expected dollar loss.

If a player makes $10,000 of Pass Line action:

$$ \text{Expected loss}=$10{,}000\times1.414%=$141.41 $$

At $100,000 of action:

$$ \text{Expected loss}=$100{,}000\times1.414%=$1{,}414.14 $$

The percentage stays low. The dollar cost scales with volume.

Convergence does not mean the player’s balance returns toward the starting bankroll. It means the average loss per dollar tends toward the edge as the sample grows.

A low edge does not mean a high chance of winning

Pass Line wins about 49.29% of completed decisions. Don’t Pass under Bar 12 has a slightly lower house edge because one come-out result pushes, yet it still loses more money than it wins over the complete contract.

House edge and hit rate answer different questions:

• hit rate asks how often a wager pays;
• payout determines how much a win returns;
• house edge combines every outcome and payment into average cost;
• session-win probability also depends on stopping time, stake changes, and the sequence of outcomes.

A high hit rate can still accompany a poor payout. A low-edge wager can still lose several decisions consecutively. No single percentage summarizes the whole session.

Pressing can turn a low-edge plan into variable exposure

Suppose a player starts with $12 each on Place 6 and Place 8 and increases the winning number by $6 after every hit. The underlying probability and house edge of each properly paid Place wager do not improve. The amount exposed to the eventual 7 rises.

A long roll can create a large visible rack and larger working bets. When 7 appears, the session result depends on how much profit was collected and how much was repeatedly pressed back into action.

“Playing with winnings” is an accounting phrase, not a probability rule. Once chips are won, they belong to the player. Recommitting them creates fresh exposure.

The same issue appears after hardway wins, Come-bet odds increases, and loss-chasing progressions. Bet quality does not cancel stake escalation.

Bankroll failure can occur before expectation becomes visible

Risk of ruin is the probability that a bankroll reaches zero—or a practical stop level—before the planned play ends. It depends on:

• starting bankroll;
• wager size;
• volatility of the bet mix;
• correlation among active wagers;
• session length;
• pressing or chasing rules;
• the player’s stop conditions.

A $200 bankroll supporting a $10 flat Pass Line bet can absorb more ordinary losing decisions than the same bankroll supporting $10 Pass, $50 odds, and $36 across 6 and 8. The second approach may have a lower blended house-edge percentage, but one seven-out can remove $96.

The bankroll-risk guide focuses on this survival question. The variance simulator is more appropriate than expected value alone when comparing possible short-term paths.

Comps do not reverse the game

Casino rewards are normally based on estimated theoretical value, tracked action, time, and internal reinvestment policy. A low-edge player may earn less per dollar wagered than a high-edge player because the expected casino value is lower.

A meal, room discount, or free play offer can reduce the net entertainment cost. It should not be treated as proof that more wagering is profitable. The value of the benefit must be compared with:

• expected gambling loss;
• travel and incidental spending;
• restrictions and expiration;
• the player’s realistic use value, not the retail price.

Playing additional hours solely to earn a reward creates new action and can cost more than the benefit.

A practical low-edge session plan

Choosing lower-edge bets is still the correct mathematical preference when the alternatives provide similar entertainment. Pair that choice with controls that address volume and path:

1. set a cash loss limit before buying in;
2. define the maximum dollars allowed on the layout at one time;
3. decide an odds multiple that the bankroll can absorb;
4. exclude unplanned center and one-roll bets;
5. set a time or decision limit rather than waiting to get even;
6. count chips taken off the table as no longer available for pressing;
7. stop when the preset limit is reached, regardless of how favorable the next roll feels.

The broader low-house-edge hard truth applies the same principle across casino games. The craps-specific lesson is more operational: one reasonable line bet can sit beside enough fast, correlated, or high-edge action to make the complete layout expensive.

Low edge is valuable because it reduces expected cost. It works only as advertised when the player also controls total action, simultaneous exposure, and time. It is a pricing advantage relative to worse bets—not a promise of a small loss.