The game of red dog is a simple affair, it is actually based on a child's game. In the casino version, each player puts up a wager. Two cards are dealt. If they are consecutive, the hand is a tie, and no money changes hands. If they are of the same value, a third card is dealt. If this is also the same, this is a "triple" and winning bets are paid off at 11-1, otherwise the hand is a tie.
If the two cards are neither consecutive nor identical in value the value difference between them is calculated -1. This is called the spread eg a 6 and a 2 is a spread of 3. A four and six is a spread of 1 and so on.
Aces count as 14, while kings, queens and jacks count 13, 12 and 11 respectively.
Suits are of no basic significance.
After the spread has been determined, players are offered the option of doubling their wager. Then a third card is dealt. If it falls inside the value of the first two cards, the player wins. If the card falls outside the value of the spread, he loses his wager. The player is paid off at even money for spreads greater than 3. He is paid off at 2-1 for a spread of 3,4-1 for a spread of 2, and 5-1 for a spread of 1.
The best strategy is to raise with any spread of 7 or greater.
The house edge from a player using the best strategy is 2.33% with eight decks and 2.37% with six decks. A bit of irony: the game operators seem to believe the game is less favourable with more decks as with blackjack. Actually more decks favours the players at this game, mostly because there is a greater chance of a triple with multiple decks.
Efforts have been made to determine the viability of a card-counting at red dog in the usenet newsgroup rec.gambling other-games. Primary contributors to this study were Mr.Ted Alper and Dr.Yui-bui Chen. Mr.Alper told me in an e-mail that with perfect play the game could be profitable with deep penetration, but that until a significant % of the cards had been dealt out favourable opportunities were small and rare.
Dr.Yui-bui Chen conducted a simulation of 200 million hands for the purposes of testing his 5-level "C" count. His conclusion: you would need a 1-900 spread to beat the game consistently.
This study was a little pessimistic, since Chen assumed an eight-deck game with 75% of the cards dealt out. Sometimes red dog was dealt from six decks with only a deck cut out of play. Moreover Chen's results made no attempt to alter raising decisions with his card-count. Finally, a player would not have to play every hand.
However, my attempts to design a playing strategy for Chen's count yielded a gain of only .3% over basic strategy. Most of the gain comes from raising on spreads of 6, which is marginally unfavourable for a full deck. Tests against a six-deck game with 85% penetration revealed the system still required an unreallistic shift in bet size to be profitable, owing to the attrition of the many small negative expectation bets off the top of the shoe. Backcounting helps naturally, but not much.
It seems unlikely a practical card-count could be designed to give the player a worthwhile edge, owing to the games fundamentally non-linear structure and high house edge. There are other ways to skin a cat of course...