You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Im very suprised that the blackjack agent doesn't include bets. The whole point of trying to beat the house in blackjack is making the right bets (rewards scale on bet size) and thus timing them with the deck which brings me to the next point.
Why isn't there a deck? From my understanding the cards are drawn randomly and thus there isn't a deck from which the cards are drawn. The major difference comes once you remove a played card from a deck, thus reducing the possible card combinations in the next game. This is a major learning mechanism for any RL since it allows for it to remember which cards have been drawn from a deck and then make a custom bet based on what it thinks is going to be the most probable cards played the next round.
I just want to ask if there is any plans to making something like this in RLCard?
The text was updated successfully, but these errors were encountered:
@sword134 Thanks for the proposal! Our goal of blackjack is a simple example to get started. Thus, we make it as simple as possible (in the current version, we have smaller observation space).
I agree with you that including deck information in Blackjack is helpful. The possible extension could be using a game-specific parameter to suggest using a deck or not. We will hold this issue for a while to see whether any guy would like to contribute. If unfortunately, no one is interested in implementing this, we will implement the deck later :) Thanks!
Im very suprised that the blackjack agent doesn't include bets. The whole point of trying to beat the house in blackjack is making the right bets (rewards scale on bet size) and thus timing them with the deck which brings me to the next point.
Why isn't there a deck? From my understanding the cards are drawn randomly and thus there isn't a deck from which the cards are drawn. The major difference comes once you remove a played card from a deck, thus reducing the possible card combinations in the next game. This is a major learning mechanism for any RL since it allows for it to remember which cards have been drawn from a deck and then make a custom bet based on what it thinks is going to be the most probable cards played the next round.
I just want to ask if there is any plans to making something like this in RLCard?
The text was updated successfully, but these errors were encountered: