Source: computed with a deterministic blackjack odds engine I built. No Monte Carlo simulation. Each hand's best play (hit/stand/double/split) is evaluated against the dealer's outcome distribution, then averaged across dealer up-cards weighted by how often each appears in a 6-deck shoe. Rules: 6 decks, dealer hits soft 17, 3:2 blackjack.

Tools: Python + pandas for the aggregation, HTML/CSS for the chart.

The full dataset (3,300 rows. every hand, action, deck count and soft-17 rule) is free under CC BY 4.0: https://blackjackoddstrainer.com/data/

—
stickygrippy

5 comments

  1. Nice visualization, would be nice to see a version of this with a player’s Blackjack winnings included as well as the chance to hit different hands. Player’s 20 is obviously fairly profitable, but also somewhat frequent since 4/13 cards in the deck are 10s. Would also be fun to depict how the odds change at different true counts for card counting purposes.

    Also, I assume you had Claude help you with this? The visualization reminds me quite a bit of what it’s generated for me in the past

  2. Seems weird to exclude A, 10, the most valuable starting hand in blackjack!

  3. I remember in business school we learned how American casinos added a 2nd green slot for a house winning roll, increasing their odds from 1/37 to 2/38. This 1/37th increase in their odds resulted in an extra 6.8B USD in income in a single year across the industry.

    House always wins, y’all.

  4. would be interesting now to have a 2D scatter plot of (worth of a hand; probability of getting this hand)

Comments are closed.