GTO PioSOLVER - postflop equilibrium solver for Holdem - Page 121 - Poker Software -- Two Plus Two Poker Forums

Two Plus Two Forums Online Poker Sites & Marketplaces Poker Software

[GTO] PioSOLVER - postflop equilibrium solver for Holdem

Post Reply Subscribe

...

Page 121 of 299

First

21 71 101 111 116 117 118 119 120 121 122 123 124 125 126 131 141 171 221

Last

Page 121 of 299

First

21 71 101 111 116 117 118 119 120 121 122 123 124 125 126 131 141 171 221

Last

09-16-2017 , 08:16 PM

#3001

HappyLuckBox

old hand

Join Date: Feb 2013 Posts: 1,491

Quick tangent, I noticed that when I have a large tree loaded in pio (ex 120gb tree) and I leave my computer Idle for a bit and return, my ram load drops down to 60-70gb or so. Attempting to interact with the tree/browser will then load the ram back up to 120gb, and it seems to be drawing the information from my C: ssd drive.

So it looks like pio offloads some of the data from ram into my primary ssd while idle. Is there a way to prevent this from occuring, or to redirect it to my D: hdd storage drive instead. I'd like to prevent wearing out my ssd if possible

Last edited by HappyLuckBox; 09-16-2017 at 08:25 PM.

Quote

09-16-2017 , 08:39 PM

#3002

Kitsune101

centurion

Join Date: Sep 2011 Posts: 133

Questions regarding PLOCalc (and solver)

1. Is there a tentatively date for the release of the PLOSolver?
2. You mention it requires very powerful software to run. What specs are we talking about for a preflop solution?
3. Are these two separate products or does purchasing the PLOCalc give us access to the solver when it does comes out?

Thanks

Quote

09-16-2017 , 09:44 PM

#3003

HappyLuckBox

old hand

Join Date: Feb 2013 Posts: 1,491

Quote:

Originally Posted by Falcore

Ok I think I've got it, so just to clarify then, this is my understanding of how I would model a heads up hand, a SB vs BB 6max hand, and a BTN vs BB 6max hand in Pio (assuming 5/10 blinds):

HU:
OOP money in pot: 10
IP money in pot: 5
Extra dead in pot: 0

6max (SB v BB)
OOP money in pot: 5
IP money in pot: 10
Extra dead in pot: 0

6max (BTN v BB)
OOP money in pot: 10
IP money in pot: 0
Extra dead in pot: 5

So my roomate and I have been trying to figure this out for the last few hours, especially with regards to the live cash game we play in $5-$5-$10 with mandatory $20 straddle. $5 is mandatory blind on the button.

I've been unable to use:

OOP money in pot: 5
IP money in pot: 5
Extra dead in pot:30
bb size: 20

because the solver then assumes oop is first to act preflop as I stated in prior posts.
So I've been using the following config instead:

OOP money in pot: 5
IP money in pot: 0
Extra dead in pot:35
bb size: 20

I know this is a bit less accurate, but by moving the IP blind from 5 -> 0 I simply added it to extra dead money to try and help account for that.

But my roomate is making the argument that I should instead use:

OOP money in pot: 5
IP money in pot: 0
Extra dead in pot:15
bb size: 20

He reasons that extra dead money in pot of 15 will be 10 from the big blind + 5 for the button, and the bb size = 20 will be the straddle. I'm pretty sure that the bb size is only used for accuracy settings, and not something that the solver assumes is an actual blind for calculation purposes.

So which method would provide the results we are looking for?

Quote

09-17-2017 , 03:21 AM

#3004

punter11235

Carpal \'Tunnel

Join Date: Mar 2005 Posts: 8,210

Quote:

So it looks like pio offloads some of the data from ram into my primary ssd while idle.

Pio (or any other Windows program) doesn't have control over this unfortunately. There is a concept of virtual memory in Windows. Windows uses RAM and swap file for that. Programs may request allocation in virtual memory but they can't enforce their program staying in a RAM part (well, generally so, there are ways to allocate specific type of memory which can only be in RAM but that solution has its own problems). Windows tries to optimize things for speed but it's not easy to predict if you are going to use another program (and then RAM will be needed for it) or continue using Pio. Apparently it thinks Pio is idle now so it offloads it.

Quote:

Is there a way to prevent this from occuring

In general: no. You can try reducing swap file size so it doesn't have anywhere to offload the RAM.

Quote:

or to redirect it to my D: hdd storage drive instead. I'd like to prevent wearing out my ssd if possible

You can move swap file to another drive. I am not going to give specific instructions because it might be different under various Windows versions. Google: "how to move swap file windows".

Quote:

1. Is there a tentatively date for the release of the PLOSolver?

No. We are working on the interface still. We are also working on 1.10 Holdem solver release at the same time. Development time is impossible to predict because new problems and feedback from testers arise as we go.

Quote:

2. You mention it requires very powerful software to run. What specs are we talking about for a preflop solution?

PLO preflop without abstractions won't run on a single machine and is very unlikely to be available. In general PLO preflop only seems feasible with quite heavy abstractions.

Quote:

3. Are these two separate products or does purchasing the PLOCalc give us access to the solver when it does comes out?

It will be another version of the same product so you will be able to upgrade. Please understand that a market for Omaha tools is a small one but developing quality tools for it is more difficult than for Holdem. That means things are going to be expensive and likely targeted to a narrow group, especially when we talk about project like PLO preflop.

Quote:

So which method would provide the results we are looking for?

As I said: I will get back to you with it after the weekend because I need to verify if/why the natural way of forcing a check doesn't seem to work.

Quote

09-17-2017 , 03:51 AM

#3005

HappyLuckBox

old hand

Join Date: Feb 2013 Posts: 1,491

Quote:

Originally Posted by punter11235

As I said: I will get back to you with it after the weekend because I need to verify if/why the natural way of forcing a check doesn't seem to work.

All right, looking forward to next update. For now I will see if I can manage the pagefile and maybe do some other modifications like cooling my VRM mosfet (it runs even hotter than my cpu!) under load.

Quote

09-17-2017 , 09:34 AM

#3006

shahrad

old hand

Join Date: May 2008 Posts: 1,590

Quote:

Originally Posted by punter11235

It will be another version of the same product so you will be able to upgrade. Please understand that a market for Omaha tools is a small one but developing quality tools for it is more difficult than for Holdem. That means things are going to be expensive and likely targeted to a narrow group, especially when we talk about project like PLO preflop.

I was always for an Omaha Solver but meanwhile I doubt it will be useful but for a specific very narrow group.
The main issue is:
1)How good is a flop strategy when you cannot execute turn river properly? You cannot follow the strategy in a practical game.
2)How good can one estimate opponents preflop and postflop strategy? I think not good at all. And for regs it is so easy to switch strategies.

Quote

09-17-2017 , 11:30 AM

#3007

punter11235

Carpal \'Tunnel

Join Date: Mar 2005 Posts: 8,210

Quote:

I was always for an Omaha Solver but meanwhile I doubt it will be useful but for a specific very narrow group.

I meant the preflop solver as this will require several machines to run without abstractions.
Postflop should be usable with 128GB or maybe even 64GB of RAM although you won't be able to use many bet sizes.

Omaha is a big game so things are going to be slow and will take a lot of memory. Maybe we are going to go into abstractions later but developing a reliable solver without abstractions is important to be able to tell how well specific simplifications hold in practice. Without it it's a guessing game.

Quote:

2)How good can one estimate opponents preflop and postflop strategy? I think not good at all. And for regs it is so easy to switch strategies.

Well, I think postflop solver for Holdem is very useful even though you often need to calculate spots with various assumptions. It deepends your understanding of the game. It's not very practical to just follow the solutions in-play luckily.

Quote

09-18-2017 , 05:28 AM

#3008

HappyLuckBox

old hand

Join Date: Feb 2013 Posts: 1,491

So the preflop solver is fantastic when looking at spots like SB vs BB, or BTN vs BB, where the player who is calling preflop is closing the action. We can freely input 100% range and see what the solver thinks is an optimal 3bet,flat,fold range.

But when we want to look at spots like COvsBTN or SBvsBTN where the player isnt closing the action preflop, it gets a little more tricky. We cannot input 100% ranges for say SB in SBvsBTN, because the solver will give us a flatting range, but this of course makes us susceptible to being squeezed by BB in real life.

With regards to SBvsBTN therefore, I think the best approach is to force the tree to either 3bet or fold, since many good players dont flat from the SB in real life anyways. Would you agree this is good way to analyze SBvsBTN play?

As for COvsBTN, (Or any other earlier positions like UTGvsMP, HJvsCO, etc) I guess the best we can do is to input a standard CO opening range (say 25%), and a standard button defending range of say 15%? From there we can let pio analyze how to optimally play with "standard ranges." Would this be the best way to use pio preflop to analyze optimal preflop play from earlier positions?

Last edited by HappyLuckBox; 09-18-2017 at 05:46 AM.

Quote

09-18-2017 , 06:18 AM

#3009

punter11235

Carpal \'Tunnel

Join Date: Mar 2005 Posts: 8,210

We are happy to announce a small update to PLOCalc. It includes:

-new syntax in range selector
-equity graphs improvements (selecting and showing parts of the range)
-options to split hand categories in range explorer
-performance improvements and bug fixes

More in the blog post, here:
https://www.piosolver.com/blogs/news/plocalc-1-1

Quote

09-18-2017 , 07:57 AM

#3010

punter11235

Carpal \'Tunnel

Join Date: Mar 2005 Posts: 8,210

Quote:

I need to know what exactly you want to run here. As the pot is 3-way there are a few possibilities. If it's BTN vs SB with defined BTN's range then you would just do it like here:
https://gyazo.com/a3b8bd7a009b8007c29b30719c71c9cd

That is you ignore forced money put in by the button and you just add it to the forced raise.

Quote:

But my roomate is making the argument that I should instead use:

OOP money in pot: 5
IP money in pot: 0
Extra dead in pot:15
bb size: 20

It's not correct because 5$ from the button is not "dead", it's still live (is part of the bet which "plays"). The money on the table is either "dead" that is in the middle and not counting to other players bets or "live". From my understanding of the rules the straddle is "live".

Quote:

I'm pretty sure that the bb size is only used for accuracy settings, and not something that the solver assumes is an actual blind for calculation purposes.

This is correct.

Quote

09-18-2017 , 08:45 AM

#3011

HappyLuckBox

old hand

Join Date: Feb 2013 Posts: 1,491

Quote:

Originally Posted by punter11235

Okay, the exact parameters I am trying to sim in my game are:
$5 otb (IP player who is raising to $60)
$5 sb (OOP player who is the defender)
$10 bb (Presumably Folds)
$20 live straddle (Presumably Folds)

and we play $5000 effective usually.

So I was thinking something like this?

I noticed that your tree has SB being able to flat. I thought forcing sb to 3bet or fold would produce more useful results to prevent SB from being squeezed by BB or Straddle in reality. As for the extra dead money in pot, I used 30, to account for $20 from the Straddle and and $10 BB who both presumably fold after SB 3bets.

Quote

09-18-2017 , 09:14 AM

#3012

pingeras

Guest

Posts: n/a

Let's imagine that I run a script of XXX boards without weights and I get the XXX + script.txt
If I delete the script.txt I create a new script with the same XXX boards but with the weights (so the Solver don't calculate the new boards just "skip_if_done ok!") I will get a new script.txt file

Will I able to run a multiflop aggregation report properly?
Maybe the real question is: Is the weight information store in the .CFR file or into the .txt file?

Quote

09-18-2017 , 11:03 AM

#3013

punter11235

Carpal \'Tunnel

Join Date: Mar 2005 Posts: 8,210

Quote:

Yes, this is a good point. It's better to only have raise or fold for SB. 30$ dead money in the pot is correct as long as the straddle folded as well (so is not in the button).

Quote:

Will I able to run a multiflop aggregation report properly?
Maybe the real question is: Is the weight information store in the .CFR file or into the .txt file?

It's stored in the script file (so .txt file). You can try editing it manually. Make sure to make a back-up copy first and experiment on very simple scripts (preferably with 2-3 flops) first so you don't waste too much time. Notice that the newest version of PioViewer shows you loaded flops when you try to run aggregation report so you will be able to verify if your changes took desirable effect.

Quote

09-18-2017 , 02:29 PM

#3014

HappyLuckBox

old hand

Join Date: Feb 2013 Posts: 1,491

Quote:

Originally Posted by punter11235

Yes, this is a good point. It's better to only have raise or fold for SB. 30$ dead money in the pot is correct as long as the straddle folded as well (so is not in the button).

Awesome! Glad to know that my parameters were correct involving sb play for the live game I was attempting to model. The last few days worth of compute were not useless afterall

Quote

09-18-2017 , 07:15 PM

#3015

DoctorZoidberg

stranger

Join Date: Sep 2017 Posts: 6

I have PIO edge version and I was using PIO for postflop only so far. So I'm new to preflop solver
I wanted to solve one preflop solution.
100BB, BU open 2.5x with some range and I wanted to see BB response.
I give BB option to call fold or 3bet 4x
I build pure preflop tree click on go/stop button and see the results.
But my results must be wrong - only 4.4% of 3bet and no fold
Am I missing something?
https://imgur.com/a/wffMb
https://imgur.com/a/ahT4P

Quote

09-18-2017 , 08:28 PM

#3016

magorko

veteran

Join Date: Jul 2010 Posts: 2,209

Quote:

Originally Posted by punter11235

No. We are working on the interface still. We are also working on 1.10 Holdem solver release at the same time.

what will be different between piosolver and holdem solver?

Quote

09-19-2017 , 02:31 AM

#3017

punter11235

Carpal \'Tunnel

Join Date: Mar 2005 Posts: 8,210

Quote:

I build pure preflop tree click on go/stop button and see the results.
But my results must be wrong - only 4.4% of 3bet and no fold
Am I missing something?

Pure preflop tree is one without postflop play (check-down to the end). Those solve very fast but aren't very useful in practice.
You need to choose a subset of flops and build normal preflop tree. You are also missing an option to call a 4bet or at least to fold after it (right now only raise is possible in your tree).

In general if you would like us to verify if the tree is correct we need to see the whole config. The way to copy it is to click "to clipboard" button (visible on your screenshot) and then paste it to pastebin.com and link here (or paste here directly enclosing it in [ code ] [ /code ] tags (without spaces inside)).

There is a tutorial about it, here:
https://www.youtube.com/watch?v=LX8n2E3OGok

Quote:

what will be different between piosolver and holdem solver?

It's the same. Current version is 1.9, the next one is going to be 1.10. It's going to be a free update.

Quote

09-19-2017 , 08:49 AM

#3018

DoctorZoidberg

stranger

Join Date: Sep 2017 Posts: 6

Hi punter

I recreate the tree to be completed as u recommended
I choose 10 flop subsets and fill the postflop betsizing data.
After I click on "build tree" I receive an error: "Server Error: ERROR: preflop solving is not supported by this version"
I'm connect to "preflop-browser" and I also run PIO Updater begore to check the latest version (1.9.2 (preflop))
As mentioned before I have an edge version of PIO. No idea whats wrong but as mentioned earlier I have never use preflop feature before

https://imgur.com/a/A9dmr

Quote

09-19-2017 , 10:54 AM

#3019

punter11235

Carpal \'Tunnel

Join Date: Mar 2005 Posts: 8,210

Quote:

I'm connect to "preflop-browser" and I also run PIO Updater begore to check the latest version (1.9.2 (preflop))

If you have preflop-browser.exe to begin with you likely had basic or pro version before. You need to connect to PioSOLVER-preflop.exe (or PioSOLVER-preflopAVX.exe). Those should be in your solver's folder as well.

If that doesn't help or you can't find them please pm me so I can help you with that either here or on Skype.

Quote

09-20-2017 , 07:48 AM

#3020

DoctorZoidberg

stranger

Join Date: Sep 2017 Posts: 6

Hi punter

No I dont have PioSOLVER-preflop.exe neither or PioSOLVER-preflopAVX.exe.
Only preflop-browser.exe as displayed on the last screenshot (on the right side next to PIO). I also wrote on the skypegroup but no response there.

Quote

09-20-2017 , 07:59 AM

#3021

punter11235

Carpal \'Tunnel

Join Date: Mar 2005 Posts: 8,210

Quote:

No I dont have PioSOLVER-preflop.exe neither or PioSOLVER-preflopAVX.exe.
Only preflop-browser.exe as displayed on the last screenshot (on the right side next to PIO). I also wrote on the skypegroup but no response there.

It looks like the upgrade to edge didn't go well then. You need check "force update" option and make sure the destination folder (default C:\PioSOLVER). I go through it on the video, here:
https://www.youtube.com/watch?v=1PEmsGaGLvk (point 1, see the video description).

If that still doesn't put the desired files in your folder please PM me here so I will add you on Skype and go through it with you.

Quote

09-20-2017 , 08:13 AM

#3022

monkey1

centurion

Join Date: Aug 2012 Posts: 144

Why is Pio folding some Qx in this river spot?

https://imgur.com/qaYRBU8

Code:

#TreeBuilding#V2
#Range0#AA,KK,QQ,JJ,TT,99,88,77,66,55,44,33,22,AK,AQ,AJ,AT,A9s,A8s,A7s,A6s,A5s,A4s,A3s,A2s,KQ,KJ,KTs,K9s,K8s,QJ,QTs,Q9s,JTs,J9s,T9s,T8s,98s,97s,87s,76s,65s,54s
#Range1#QQ:0.25,JJ:0.5,TT:0.75,99,88,77,66,55,44,33,22,AKo:0.25,AQ:0.75,AJs:0.75,AJo,ATs,A9s,KQs:0.75,KQo,KJs,KTs,K9s,QJs,QTs,Q9s,JTs,J9s,T9s,T8s:0.5,98s,97s:0.5,87s:0.5,76s:0.5
#Board#Qs 3s 2d
#Pot#75
#EffectiveStacks#970
#AllinThreshold#67
#AddAllinOnlyIfLessThanThisTimesThePot#500
#MinimumBetsize#0
#UseUnifiedBetAfterRaise#False
#UnifiedBetAfterRaise#
#ForceIPBet#False
#ForceOOPBet#False
#Cap#4
#CapEnabled#True
#CapMode#NoLimit
#RemovedLines#Bet, Raise, Raise 68@Bet, Raise, Raise, Call, Bet 2e@Bet, Raise, Raise, Call, Check, Bet 2e@Bet, Raise, Call, Check, Bet 75@Bet, Call, Bet 2e, Raise 68@Check, Bet, Raise, Raise 68@Check, Bet, Raise, Raise, Call, Bet 2e@Check, Bet, Raise, Raise, Call, Check, Bet 2e@Check, Bet, Raise, Call, Check, Bet 75@Check, Bet, Call, Check, Bet 2e, Raise 68@Check, Check, Bet 2e@Check, Check, Check, Bet 2e@Check, Check, Bet, Raise, Raise 68@Check, Check, Bet 75, Call, Bet 1e@Check, Check, Bet 75, Call, Check, Bet 1e@Check, Check, Check, Bet, Raise, Raise 68@Check, Check, Check, Check, Bet 1e@Check, Check, Check, Check, Check, Bet 1e
#ExtraLines#Bet, Raise, Raise 2e@Bet, Raise, Call, Check, Bet, Raise 1e@Bet, Call, Bet 2e, Raise 2e@Check, Bet, Raise, Raise 2e@Check, Bet, Raise, Raise, Call, Bet 1e@Check, Bet, Raise, Call, Check, Bet, Raise 1e@Check, Bet, Call, Check, Bet 2e, Raise 2e@Check, Check, Bet 2e, Raise, Raise 1e@Check, Check, Bet 75, Raise, Raise 2e@Check, Check, Check, Bet 75, Raise, Raise 2e@Check, Check, Check, Bet 2e, Raise, Raise 1e@Check, Check, Check, Check, Bet 175@Check, Check, Check, Check, Check, Bet 175@@Check, Check, Check, Bet, Call, Check, Bet, Raise 65@Check, Check, Check, Bet, Call, Bet, Raise 65@Check, Check, Bet, Call, Check, Bet, Raise 65@Check, Check, Bet, Call, Bet, Raise 65
#FlopConfig.BetSize#60
#FlopConfig.RaiseSize#68
#FlopConfig.AddAllin#False
#FlopConfig.DonkBetSize#
#TurnConfig.BetSize#2e 75
#TurnConfig.RaiseSize#68
#TurnConfig.AddAllin#False
#TurnConfig.DonkBetSize#33
#RiverConfig.BetSize#1e 83
#RiverConfig.RaiseSize#92
#RiverConfig.AddAllin#False
#RiverConfig.DonkBetSize#1e 83
#FlopConfigIP.BetSize#60
#FlopConfigIP.RaiseSize#68
#FlopConfigIP.AddAllin#False
#FlopConfigIP.Dont3bet#False
#TurnConfigIP.BetSize#2e 75
#TurnConfigIP.RaiseSize#68
#TurnConfigIP.AddAllin#False
#TurnConfigIP.Dont3bet#False
#RiverConfigIP.BetSize#1e 83
#RiverConfigIP.RaiseSize#92
#RiverConfigIP.AddAllin#False
#RiverConfigIP.Dont3bet#False
#Rake.Cap#30
#Rake.Fraction#0.05
#Rake.Enabled#True

Last edited by monkey1; 09-20-2017 at 08:28 AM. Reason: added tree config

Quote

09-20-2017 , 08:17 AM

#3023

punter11235

Carpal \'Tunnel

Join Date: Mar 2005 Posts: 8,210

Quote:

Why is Pio folding some Qx in this river spot?

https://imgur.com/5RPsBcP

Can you paste the tree config itself? (The stuff that starts with #). It's easier to re-create the tree (going to Tools->paste tree config). I could re-run the script but it requires editing out the paths and then I still don't see filled in options in the tree builder.

Looking at your screenshot though it looks like AQo is a borderline call/fold and is a mixed action.

Quote

09-20-2017 , 08:31 AM

#3024

monkey1

centurion

Join Date: Aug 2012 Posts: 144

Quote:

Can you paste the tree config itself?

^

I guess it folds some since IP has almost no bluffs.

Quote

09-20-2017 , 08:53 AM

#3025

punter11235

Carpal \'Tunnel

Join Date: Mar 2005 Posts: 8,210

Quote:

I guess it folds some since IP has almost no bluffs.

Again, it's hard to say without seeing the config and re-running it locally.
Sometimes AQ/KQ are actually worse calls than lower queens because they block potential bluffs of the opponent (if they have Ax/Kx in their range to bluff with but not other hands). You will need to look at it yourself or paste the config so I (and other people) can re-run it

Quote

Page 121 of 299

First

21 71 101 111 116 117 118 119 120 121 122 123 124 125 126 131 141 171 221

Last

Post Reply Subscribe

...

Page 121 of 299

First

21 71 101 111 116 117 118 119 120 121 122 123 124 125 126 131 141 171 221

Last