Poker-AI.org

Re: Would an AI that had no memory of previous streets donk

2017-12-22T23:17:56+00:00

You are right, mayority of the players never donk in HU spot. But it doesn't mean that it is wrong to donk. The reason is probably because several years ago it was considered that donking is totally bad play because guy with initative would cbet almost always so why donk with value hand doesn't make sense. Today, regulars don't cbet that bluff heavy, take care about being more balanced, so they are not going to cbet that often, which increases the EV of donking with value and bluffs.

Statistics: Posted by mlatinjo — Fri Dec 22, 2017 11:17 pm

Re: Would an AI that had no memory of previous streets donk

2017-12-22T16:48:36+00:00

mlatinjo wrote:

Not only polarized. Also donk pairs that need protection. Anyway those go mostly in bet / fold range. If you bet strong hands and some solid draws and some hands that need protection, then you cna defend draws + strong hands vs raise and fold hands that need protection.
Did you ever look at solver programs? Like e.g. pio solver? It would answer your question / questions regarding optimal strategy

I'm of course familiar with pio but I don't have the program myself. I suppose my main mistake was conflating GTO play and winning online poker play. That's why I thought these small % polarized donk ranges are uncommon, because I base it on population tendencies and what I've seen players (+streamers,coaches) do.

Statistics: Posted by mediacalc — Fri Dec 22, 2017 4:48 pm

Re: Would an AI that had no memory of previous streets donk

2017-12-22T01:46:17+00:00

Statistics: Posted by mlatinjo — Fri Dec 22, 2017 1:46 am

Re: Would an AI that had no memory of previous streets donk

2017-12-20T20:37:47+00:00

mlatinjo wrote:

In poker it makes sense to donk with some hands. E.g. hands that are strong enough to donk turn, bet river, but not strong enough to raise turn bet river. So if a donk range is polarized, IP player can't raise many combos, because he can raise only strongest hands for value.

So if we let a strategy converge as best we can to the nash equlibrium, then this strategy will have a donking %. Understood. I suppose my question was bad because it didn't assume people donk polarized ranges already (I see this as uncommon)

Thanks!

Statistics: Posted by mediacalc — Wed Dec 20, 2017 8:37 pm

Re: Would an AI that had no memory of previous streets donk

2017-12-19T22:07:03+00:00

Statistics: Posted by mlatinjo — Tue Dec 19, 2017 10:07 pm

Would an AI that had no memory of previous streets donk more

2017-12-18T21:33:34+00:00

Imagine we're on the turn, we have the concept of an IP player and an OOP player. And the different ranges for them. Let's assume IP player's range is stronger on the turn.

An OOP AI that doesn't know position would still realise that donking is bad, right? Because it would bet and get raised by the stronger range a lot.

But would the donk % be very low like <2% or a much higher value like %20?

Statistics: Posted by mediacalc — Mon Dec 18, 2017 9:33 pm