Asian Casinos Lose Up to $700 Million a Year to AI-Assisted Card Counting in Baccarat

^{Players at the baccarat table}

Artificial intelligence is helping baccarat players identify moments when side bets start to offer an edge over the house. According to consulting firm Differential Labs, land-based casinos and online operators across Asia are losing an estimated $500–700 million a year as a result of these methods.

In baccarat, players bet on one of two hands — Player or Banker — or on a tie. The winning hand is the one with a total closest to nine. Alongside these main bets, casinos also offer side bets — for example, on the exact score of a tie or a specific outcome for either hand. Side bets account for roughly 5% of total baccarat wagering volume, but at some casinos they generate 40–60% of the game’s revenue.

Baccarat is played with multiple decks, and as cards are dealt, they are removed from the shoe. This changes the odds of certain combinations occurring.

By tracking which cards have already been dealt, a player can identify when a particular side bet starts to offer a mathematical edge over the house.

AI makes this process much easier: software can monitor several side bets simultaneously and identify which one currently offers the player an edge.

According to Differential Labs, when tracking a single side bet, a favorable opportunity typically arises one to three times during a shoe. When several side bets are analyzed at once, that number rises to around seven.

Differential Labs highlights two side bets where card counting is especially effective: High Tie, which pays out on a tie with a specific score — such as 8:8 or 9:9 — and Dragon Seven, a bet on a specific type of Banker win. Casinos offer high payouts on these rare outcomes, while their probability can shift noticeably as the shoe progresses.

Differential Labs says the number of players using computer-assisted card counting and AI tools has been growing since 2021.