calculating "GTO preflop ranges" in PLO - Poker Theory

Two Plus Two Forums Poker Strategy Poker Theory & GTO

calculating "GTO preflop ranges" in PLO

Post Reply Subscribe

...

07-03-2018 , 06:17 AM

md46135

grinder

Join Date: Mar 2015 Posts: 420

There's a software which claims it knows GTO preflop ranges for PLO, it's called PLO Matrix. How can they have GTO ranges preflop for a game which doesn't stop at preflop?

Quote

07-03-2018 , 05:50 PM

browni3141

Pooh-Bah

Join Date: Aug 2015 Posts: 5,678

It sounds like false marketing.

There aren't even known "GTO pre-flop ranges" for NLHE, a less complex and more heavily studied game.

Quote

07-04-2018 , 05:39 AM

samooth

veteran

Join Date: May 2009 Posts: 3,350

read the FAQ:

Quote:

How were the underlying ranges created?
The ranges were created using MonkerSolver. We used three 512gb servers and ran simulations on them continuously for months to create the most accurate possible ranges for the Matrix.

Quote

07-05-2018 , 07:20 PM

RustyBrooks

Carpal \'Tunnel

Join Date: Feb 2006 Posts: 24,647

I can't think of any situation where it makes sense to call something a "512 gb server." Is that ram? You can rent servers on AWS with almost 2000gb of ram for a few bucks an hour.

Quote

07-06-2018 , 05:52 AM

samooth

veteran

Join Date: May 2009 Posts: 3,350

calling it this way makes perfect sense in the context of using solvers since cpu and ram is usually the bottleneck. price per hour is a no bueno representation for server costs, you are likely underestimating how long big sims need to run.

Quote

08-01-2018 , 04:12 AM

PastyDonkey

centurion

Join Date: Jul 2014 Posts: 188

I missed this somehow, but this is my software. 512gb is ram. It is the norm is the monker community to refer to your servers by how much ram they have because ram is the limiting factor for running sims. The ranges are GTO in the same sense that a pio output is gto. For the how to get preflop ranges: Monker runs an abstracted postflop that gives a good estimation of preflop ranges.

If there is a server with 2TB of ram that costs a couple dollars an hour please pm. The ones I have seen are significantly more expensive.

Quote

08-01-2018 , 09:54 AM

RustyBrooks

Carpal \'Tunnel

Join Date: Feb 2006 Posts: 24,647

X1 instances are like $4/hour but I realize now that this is probably a price with a reserved instance under contract. It's more like $13/hour otherwise. They have even bigger ones now.

Quote

08-06-2018 , 10:30 PM

BDHarrison

Pooh-Bah

Join Date: Nov 2016 Posts: 4,529

Are most online pots contested heads-up? Because I just don't believe there is such a thing as GTO preflop ranges for a game where most pots are multi-way.

At least their FAQ seems to confirm my opinion that hands with three of a suit are usually garbage.

Quote

08-07-2018 , 01:25 AM

Abbaddabba

Pooh-Bah

Join Date: Jul 2005 Posts: 5,807

Put it this way - if you’re opening on the button and you get flatted by the sb, do you not see how it’s possible to compute optimal responses? Its already many times many times more complex than heads up but it’s very clearly theoretically possible.

Break it down further to a one street game and the solution can be done with a pen and paper.

Quote

08-07-2018 , 11:50 PM

#10

PastyDonkey

centurion

Join Date: Jul 2014 Posts: 188

You guys are mostly questioning monkersolver which is what we used to generate the ranges. Monker has some secret sauce in it's abstractions where it dramatically lowers the size of the game tree and still comes up with results that match pio. It is the first commercially available plo and multiway solver. There is a little more info on the monker webiste http://www.monkerware.com/solver.html but nobody really knows exactly what is happening behind the scenes.

Quote

08-18-2018 , 02:38 PM

#11

samooth

veteran

Join Date: May 2009 Posts: 3,350

lol @ secret sauce

Quote

Post Reply Subscribe

...