Neutrino
Neutrino
Joined: Feb 20, 2014
  • Threads: 84
  • Posts: 515
September 13th, 2015 at 12:59:14 AM permalink
Quote: MaxPen

I don't think it's possible to solve NLHE. There were some claims that it was then some pros proved the developers wrong. Forget the name of that study. Will try to find the details on that and post.



I looked into the cephus project and I got some questions about its effectiveness.

Does it take advantage of the opponent's mistakes and leaks? Or does it simply play regardless of the tendencies of the opponent?

I believe profitable poker is all about exploiting other people's mistakes. This is why I hate playing against tight players. You can bitch and moan about them playing badly all day but in the end you just simply can't take money from someone who doesn't put money in.

So, playing an "optimal" strategy that only seeks to make itself non-exploitable rather than actively exploiting its opponents, is kinda like saying "The optimal strategy to no-limit holdem is to have 4 friends sit at the same table by themselves and everyone fold every hand, except limp all pocket pairs and suited connectors. This way the house gets no rake and you can just farm the shit out of the bad beat and high hand jackpots." Sure, that's optimal and +EV, and everything else you do in poker is literally -EV because of rake, but obviously something is wrong here otherwise everyone would be doing it. What's wrong is that it's actually a waste of time because playing poker is a much bigger +EV if you can exploit other people's mistakes.

I know I'm not the only one who think exploiting opponets' mistakes are important. Afterall, I would think almost everybody would find it silly if the Cephus bot plays a Nit the same way it plays a LAG.
MaxPen
MaxPen
Joined: Feb 4, 2015
  • Threads: 13
  • Posts: 3634
September 13th, 2015 at 1:31:13 AM permalink
Cepheus solved heads up LHE. Nits don't win short handed let alone heads up long term. The principles of Cepheus work long term heads up because it is the optimal play for every possible situation possible without error. Opponents playing style has no relevance heads up.
Your original question was about heads up LHE. It has been solved definitively. Cepheus did not solve full table play.
Neutrino
Neutrino
Joined: Feb 20, 2014
  • Threads: 84
  • Posts: 515
September 13th, 2015 at 10:13:06 PM permalink
I looked into it further. I was quite surprised to find that, the bot would almost never (<1%) cap the bet aka 4bet, and also almost never limp-reraise the SB.

It doesn't even 4bet with AA!
I can't figure out why this is the right play. Because it really just doesn't make sense.

On the contrary, I'd like to ask how much a error is when a human 4bets. As in is 4betting an acceptable play or dead wrong.


Another thing I would like to know is, the bot probably has data on EV gain of SB and EV loss of BB. I can't seem to find it though.
MaxPen
MaxPen
Joined: Feb 4, 2015
  • Threads: 13
  • Posts: 3634
September 13th, 2015 at 11:06:05 PM permalink
I think the overall idea is that it is about who makes the least mistakes. Capping preflop with extreme advantage gives the disadvantaged opportunity to fold therefore limiting the amount of units won on average the limited number of times you actually have that premium hand. On the flipside constant pressure from raising gives an opponent excuse to fold even when you are not favored or slightly disadvantaged.
frisbee25
frisbee25
Joined: Oct 8, 2014
  • Threads: 12
  • Posts: 27
October 26th, 2015 at 4:06:15 PM permalink
Can someone explain why the hand range of Cephus is a fixed set of hands heads up and not a range, and is that true other than heads up?

  • Jump to: