### Presentation

```By: Jonathan Quenzer

To have a computer learn how to play Blackjack
through reinforcement learning
◦ Computer starts off with no memory. After each hand is
played, the computer learns more.
◦ Goal is to have computer make the best possible decision of
how much to bet and when to hit/stay
 Splitting hands and doubling down will not be included. This
decreases the odds of winning.


The dealer has a 5-8% advantage depending on the
specific rules without using card counting
The player through correct strategy and using card
counting can obtain at most a 2% advantage on the
dealer



I wrote a Matlab program to simulate Black Jack.
Feature vectors were generated by running the
program and analyzing each hand played.
All of the features were scaled to have a mean of ½,
minimum of 0, and maximum of 1.
Hand History
Ace
2
The number of cards remaining in the
deck for each value of card
…
10
PT
Player Total Without Aces
PA
Number of Aces in Player's Hand
DSC
The Dealers Showing Card
Hit/Stay
+1 = Hit, -1 = Stay
Win
1.5 = Black Jack, 1 = Win, 0 = Lose
Lose
1 = Lose, 0 = Win
Push
1 = Push, 0 = Win
= Classification
Feature Set
Hit/Stay
Ace
2
The number of cards remaining in the
deck for each value of card
…
10
CC
Calculated Card Count: Hi-Opt II used
PT
Player Total Without Aces
PA
Number of Aces in Player's Hand
DSC
The Dealers Showing Card
+1 = Hit, -1 = Stay: Opposite of Hit/Stay
Hit/Stay
from Hand History if the player lost
Feature Set
Bet Amount
Ace
2
The number of cards remaining in the
deck for each value of card
…
10
CC
Calculated Card Count: Hi-Opt II used
+1.5 if hand was Black Jack, +1 if won,
Bet Min/Max
-1 if lost


Example of 5 nearest
neighbors
Neighbors sum to +3,
so decide to Hit


Computer started
with no knowledge
The player gained