aka23
Dormant account
- Joined
- Jun 11, 2006
- Location
- Planet Earth
I am working on an optimal strategy calculator for Pai Gow. It decides on how to split cards by adding up EVs for all the valid splits and selecting the highest one. My working test version is at
The calc is showing slight differences in a few situations from the most commonly recommended strategy. An example is 4 of a Kind. The standard B&M strategy for 4 of a kind is:
2 through 6: Always keep together.
7 through 10: Split unless a king or better can be played in front.
Jack through king: Split unless an ace can be played in front.
Aces: Always split.
My calc suggests:
2 through 5: Play the four of a kind
6 through 9: Split into two pairs unless 2-card hand contains a king or better
10 through Queen: Split into two pairs unless 2-card hand contains an ace
King: Split into two pairs unless 2-card hand contains ace, then split KKK / AK
Ace: Split into two pairs
I believe my math is correct, so I am uncertain about the reasons for the slight difference. It does not relate to optimizing for player vs banker, as I have checked. It might related to recommended strategy optimizing for KJ or AJ + 4 of a kind, rather than K9 or A9. I suppose, you'd want to optimize for A10 or K10, given that there are 2 cards left, which is right on the borderline between the two strategies listed above. It might also relate to optimizing for different dealer play strategies.
Does anyone know of optimal strategy tables for Pai Gow that were determined via computer simulation, rather than following dealer statagies? This would be helpful for verifying results.
You do not have permission to view link
Log in or register now.
. The calc is showing slight differences in a few situations from the most commonly recommended strategy. An example is 4 of a Kind. The standard B&M strategy for 4 of a kind is:
2 through 6: Always keep together.
7 through 10: Split unless a king or better can be played in front.
Jack through king: Split unless an ace can be played in front.
Aces: Always split.
My calc suggests:
2 through 5: Play the four of a kind
6 through 9: Split into two pairs unless 2-card hand contains a king or better
10 through Queen: Split into two pairs unless 2-card hand contains an ace
King: Split into two pairs unless 2-card hand contains ace, then split KKK / AK
Ace: Split into two pairs
I believe my math is correct, so I am uncertain about the reasons for the slight difference. It does not relate to optimizing for player vs banker, as I have checked. It might related to recommended strategy optimizing for KJ or AJ + 4 of a kind, rather than K9 or A9. I suppose, you'd want to optimize for A10 or K10, given that there are 2 cards left, which is right on the borderline between the two strategies listed above. It might also relate to optimizing for different dealer play strategies.
Does anyone know of optimal strategy tables for Pai Gow that were determined via computer simulation, rather than following dealer statagies? This would be helpful for verifying results.