Warm-up opponent ranges about deepholdem

[happypepper]

1. is correct
2. I believe for aggressive approach, it means they flip a coin before reconstructing ranges. If coin is heads, they will feed uniform range as initial state to gadget. If coins is tails they will use previous opp range as initial state to the gadget.

[felagund08]

From DeepStack article:

Another option is more aggressive and sacrifices the re-solving guarantees when the opponent’s >range estimate is wrong. It forces the opponent with probability b to follow through into the game
with a hand sampled from the estimated opponent range. With probability 1 - b they are given
a uniform random hand and can choose to terminate or follow through.

That means that with probability b they avoid the gadget, doesn't it?

[happypepper]

Ah I think you're right. I don't understand why it's safe to use do that though when DeepStack is first to act and opponent called a bet on previous street.

I think it should only be used if opponent bet/raised and DeepStack called on previous street. Then DeepStack is first to act on current street.