Every Baccarat Side Bet, Ranked: House Edge from 1.06% to 16.68%

2026-10-27

Here are all twelve bets, ranked by what they actually cost you per hour:

Cost per hour = house edge × $100 a hand × 70 hands.

The most expensive side bet costs nearly sixteen times what Banker costs (15.8×). Same hour, same $100 a hand: $74 against $1,168. That is not variance. That is pricing.

The one side bet that is almost defensible

Player Dragon Bonus at 2.65% is the only side bet in the table below 4%. It is still more than twice the cost of just betting Player — but it is in a different league from everything beneath it.

What is strange is that Banker Dragon Bonus runs on the identical paytable and costs 9.37% — three and a half times as much. Most people assume the paytables differ. They do not.

Why the same paytable costs 3.5× more

Dragon Bonus only pays on a non-natural win by 4 points or more. Split the wins by margin:

Watch the direction flip between the first three rows and the last six:

• Margins 1–3 (pays nothing): Banker 17.580%, Player 15.636% — Banker ahead by 1.944 points
• Margins 4+ (pays): Banker 12.021%, Player 12.730% — Player ahead by 0.709 points
• Natural wins (pay 1:1): Banker and Player exactly equal, 16.2589% each

Banker wins 1.235 percentage points more often overall, and it earns that through the drawing rules grinding out narrow wins. Dragon Bonus does not pay for narrow wins.

The Banker’s entire advantage sits in the one band Dragon Bonus refuses to pay for. Bet Banker Dragon and you are buying the Banker’s paytable to bet on the Player’s way of winning.

Pairs: the number everyone conflates

Rows one and two are not the same number, and they are constantly treated as if they were. 7.47% at 11:1 is a 10.36% edge; 14.38% at 5:1 is 13.71%.

None of this needs simulation. Eight decks, 416 cards, 32 of each rank, 8 of each exact card. After the first card is dealt, 31 of the remaining 415 match its rank. That is the whole derivation.

Big and Small: the pair that feels fairest

Big (the coup uses 5 or 6 cards) hits 62.113%; Small (4 cards) hits 37.887%. At 0.54:1 and 1.5:1 the edges are 4.35% and 5.28%.

These are high hit rate, low payout bets, so they feel like they win constantly. Per hour they cost four to five times what Banker costs. A high hit rate and a low cost are different things.

Lucky 6: last place

Banker winning with a total of 6 happens 3.7247% of the time on two cards and 1.6617% on three — 5.3864% in all. At 12:1 and 20:1 the house edge is 16.684%, the worst on the board.

EV = 0.037247×12 + 0.016617×20 − (1 − 0.053864) = −0.16683

Paytables differ between casinos, but the shape does not: a 5.4% event has to be carried by a 12–20× payout, and there is very little room left over for the player.

Method and verification

Every hit rate is exact, not simulated. A complete recursion over all 416 cards of an eight-deck shoe, dealt without replacement, through the full drawing rules: Banker 45.860% / Player 44.625% / Tie 9.516%. The pair bets are plain combinatorics.

Cross-check: eleven figures in this table reproduce well-established published values exactly — Banker 1.06%, Player 1.24%, Tie 14.36%, Tie at 9:1 4.84%, Pair 10.36%, Either Pair 13.71%, Perfect Pair 13.03%, Big 4.35%, Banker Dragon 9.37%, Player Dragon 2.65%. The method checks out before the rest of the numbers are worth anything.

Run it yourself:

• Enumeration code: github.com/baccaicom/baccarat-math — all 416 cards, 70 tests pinning every figure
• Expected-value calculator — change the payout, watch the edge move
• Risk-of-ruin calculator

For your casino’s paytable, the hit rates stay fixed; substitute the payouts:

house edge = 1 − Σ(hit rateᵢ × payoutᵢ) − push probability

Read on: Real probabilities and the house edge · The third card rule · Martingale’s two walls · The mathematics of betting systems