Two Plus Two Publishing LLC Two Plus Two Publishing LLC
 

Go Back   Two Plus Two Poker Forums > >

Notices

Commercial Software Discussion of commercial gambling-related / poker software & commercial graphics modifications

Reply
 
Thread Tools Display Modes
Old 03-12-2015, 02:47 PM   #51
punter11235
Carpal \'Tunnel
 
Join Date: Mar 2005
Location: solving poker
Posts: 6,780
Re: PioSOLVER - postflop equilibrium solver for Holdem

Quote:
(I say some because for a reason I'm not sure yet, it sometimes works and sometimes doesn't, depending on the range).
The thing is that for the author of the software it's a 15 minutes job to accept ranges in different format (as long as it's reasonable) but to make your ranges readable in other tools it's a lot of work and hidden traps like sometimes it works and sometimes it doesn't and preflop is different than postflop etc.
It really is his responsibility and 10 minute quick fix for him. We did our part by accepting CREV format.
punter11235 is offline   Reply With Quote
Old 03-12-2015, 04:50 PM   #52
Pouged
enthusiast
 
Join Date: May 2012
Posts: 62
Re: PioSOLVER - postflop equilibrium solver for Holdem

Is it possible to open multiple Pioviewer instances in order to compare different strategies side by side ?
Pouged is offline   Reply With Quote
Old 03-12-2015, 04:59 PM   #53
punter11235
Carpal \'Tunnel
 
Join Date: Mar 2005
Location: solving poker
Posts: 6,780
Re: PioSOLVER - postflop equilibrium solver for Holdem

Quote:
Is it possible to open multiple Pioviewer instances in order to compare different strategies side by side ?
It should work but take into account that PioVIEWER opens solver instance in the background. This means that if you build/load big trees in two PioVIEWERS it will takes 2x amount of memory.
If you want to compare strategies in the same tree just for different nodes there is "open new in new window" option. You can have as many of those as you like.
punter11235 is offline   Reply With Quote
Old 03-12-2015, 05:06 PM   #54
Pouged
enthusiast
 
Join Date: May 2012
Posts: 62
Re: PioSOLVER - postflop equilibrium solver for Holdem

Quote:
Originally Posted by punter11235 View Post
It should work but take into account that PioVIEWER opens solver instance in the background. This means that if you build/load big trees in two PioVIEWERS it will takes 2x amount of memory.
If you want to compare strategies in the same tree just for different nodes there is "open new in new window" option. You can have as many of those as you like.
Perfect, thanks.
Pouged is offline   Reply With Quote
Old 03-12-2015, 06:19 PM   #55
SiQ
Carpal \'Tunnel
 
Join Date: Jan 2008
Location: still ahead of the game
Posts: 6,993
Re: PioSOLVER - postflop equilibrium solver for Holdem

Hey Punter

Really cool project you have here.
I know you don't have any date set in stone, but I'm wondering if you have any guesses of a general time frame for adding additional bet sizes (for each decision point, obviously) - or if you are entirely sure that you'll be able to add them at all yet? Obviously the usefulness of the results would increase exponentially with each new sizing option (at least for quite a few of the first additions) but I can imagine that the complexity of the calculation is also increased exponentially, so I wonder if I should get my hopes up for this addition any time soon.

I'm considering making a video reviewing PioSolver. I've had a bit of fun playing around with it and I think I know my way around it but if you feel like there are any features I might miss that you want talked about feel free to shoot me a PM and we can swap details to have a quick chat.

edit: I think I have a fairly decent computer (16g ram, 6core 3.3 ghz) but doing flop analysis usually takes quite a while (not sure, maybe 20-30mins?) before the program stops itself and the nash distance has only reached as low as 0.3x while analyzing fairly shallow (~20-40bb) stack depths. Is this expected results for my machine running PioSolver Basic? Is this because the program uses less cores than the Pro version (if so - honestly I am not a big fan of this - selling a version that purposely runs slower than the upgraded version with no cost difference to your business. the lack of scripting makes sense, but a slower product doesn't - maybe you can explain if I'm just completely missing something here). Would Pro version run faster or is it mostly just my computer not being fast enough?

Last edited by SiQ; 03-12-2015 at 06:28 PM.
SiQ is offline   Reply With Quote
Old 03-12-2015, 06:59 PM   #56
punter11235
Carpal \'Tunnel
 
Join Date: Mar 2005
Location: solving poker
Posts: 6,780
Re: PioSOLVER - postflop equilibrium solver for Holdem

Quote:
I know you don't have any date set in stone, but I'm wondering if you have any guesses of a general time frame for adding additional bet sizes (for each decision point, obviously)
More flexible betsize input is likely to happen very soon. More bet sizes won't happen soon. Unfortunately it's as specific timeline as I have for now.

Quote:
or if you are entirely sure that you'll be able to add them at all yet?
Well, I can see the path. Walking it is another thing though. It will take a lot of time and work.

Quote:
I'm considering making a video reviewing PioSolver
That's great, see pm.

Quote:
I think I have a fairly decent computer (16g ram, 6core 3.3 ghz)
If it's Intel then it's huge hardware for PioSOLVER purposes. I am not sure about AMDs. For now my experience is that they are significantly slower than Intels for purposes of the solver.

Quote:
doing flop analysis usually takes quite a while (not sure, maybe 20-30mins?) before the program stops itself and the nash distance has only reached as low as 0.3x with this analysing fairly shallow stack depths.
Those numbers are only useful in relation to starting pot. 1% of the pot is quite good, 0.5% is very good, something like 0.3% is almost perfect.

Quote:
honestly I am not a big fan of this - selling a version that purposely runs slower than the upgraded version with no cost difference to your business
It runs slower mainly because i7s do hyperthreading. If I allow for 8 threads in that version then my selling point for pro version would be dimished.
To be honest we planned our main product to be ~500$ which was already considered too low for most people I've consulted. I wanted to give something cool and still very powerful at lower price point though because not everybody wants to shell out 500$ for product like that. If I allow for 6-8 threads in the basic version I would compete with my own product so to speak.
The choices were to make one expensive product or make additional cheaper one for people who don't expect to run heavy analysis or are not sure if 475$ is worth it for them. I don't agree about the point of costs to my business. By that logic I could add any feature we develop in the future to the cheapest version as I already have it developed.
It took a lot of time and effort to make it as fast as it is.

Quote:
Would Pro version run faster or is it mostly just my computer not being fast enough?
The difference on i7 quads should be about 30%. The difference on i5's should be less (because they can't run 8 hardware threads anyway). So yeah, if you want to utilize very powerful hardware then it's only available in pro version. I think this is pretty standard practice for similar tools - for example chess engines.
I mean, HM and PT offer "high stakes version" and that has 0 costs for them and no additional features (as far as I understand, correct me if I am wrong here).
punter11235 is offline   Reply With Quote
Old 03-12-2015, 08:31 PM   #57
punter11235
Carpal \'Tunnel
 
Join Date: Mar 2005
Location: solving poker
Posts: 6,780
Re: PioSOLVER - postflop equilibrium solver for Holdem

Quote:
I am not sure about AMDs. For now my experience is that they are significantly slower than Intels for purposes of the solver.
Ok, we confirmed this.
The AMDs are just slow for the purpose of the solver, at least popular models.
In official benchmarks that particular model is about 2x slower than i7 quad and for the solver it's a bit over 2x.
punter11235 is offline   Reply With Quote
Old 03-13-2015, 08:00 AM   #58
Karganeth
old hand
 
Karganeth's Avatar
 
Join Date: Jun 2008
Posts: 1,305
Re: PioSOLVER - postflop equilibrium solver for Holdem

Quote:
Originally Posted by SiQ View Post
Is this because the program uses less cores than the Pro version (if so - honestly I am not a big fan of this - selling a version that purposely runs slower than the upgraded version with no cost difference to your business. the lack of scripting makes sense, but a slower product doesn't - maybe you can explain if I'm just completely missing something here).
It's completely standard for software to have different pricing for versions that offer more features. Speed is probably one of the most desirable features so it makes sense for the higher priced versions to offer more speed. There's nothing wrong with this business model. It's probably more common than you think. Your processor most likely has more than 6 cores (probably 8) and those extra cores were disabled by intel. That's how it works. They stamp out one main chip design then disable a varying parts it then sell it at varying prices. All i7s and i5s are identical chips, they just have different amounts of stuff disabled on them (such as hyperthreading and cache size).

That being said, I do have an issue with one of the products on offer, PioSOLVER Edge. pro sells for $477 and edge is $1099, a difference of $624. Here is the description of what edge offers

Quote:
PioSOLVER edge is a product for high stakes players who just want a bit more for their analysis. Experimental features from development version will make it here before they are available in pro version. Some on-demand features are possible as well (contact us). Edge license gives you insight into our plans, planned features and experimental functionality. We will make custom compile for your hardware. Contact us before you buy this version to make sure the offer is sutiable for you.
This is extremely vague, it's not clear to the consumer what exactly he's getting. It doesn't say how much earlier experimental features will be available (and likely it can't - developers can't know with any precision when software features will be implemented). It's not clear what kind of features can be made on demand. It's very difficult for a user who doesn't have access to source code know how much work it would take for developers to implement a requested feature, so they aren't able to know what sort of features they can realistically request and expect to be implemented. It seems odd to charge $624 for someone to essentially beta test software. Normally I wouldn't be against giving the consumer more choices (however bad that choice might be), but edge undermines the confidence of people who don't own edge. Will their suggestions matter less than the suggestions of those who own edge? Will questions about upcoming planned features in this thread go unanswered because the person who asked them hasn't purchased edge?

In using the program I found a small bug. If you press the delete key twice in the "Weight for selection:" field (in the Select range part) it will display the following message


I have a few suggestions. My first suggestion is for PioViewer is to remove Hero and Villain terminology, and just replace it with IP and OOP. It was pretty confusing if I want to analyse a hand where I, hero, am in position. It's easy to think "I'm Hero so I'll put my range into the bit that says Hero". The terms Hero and Villain literally mean nothing in the way they are used.

My second suggestion is to add a view that allows me to see Range, Strategy and EV all in one view. So each hand cell would be split horizontally into 3 segments to show all the data. It would make it much easier than having to have 3 separate windows open and have to keep looking between them all to see all the data. It's very easy to lose my place doing this.

My third suggestion is to make the "Exploitable for:" show the value as a percentage of the pot. It is true that "Those numbers are only useful in relation to starting pot" so therefore it should always show those numbers as a fraction of the starting pot. I'm actually in favour of normalizing all the EVs to be a fraction of the pot too but others may disagree with this.

My final suggestion is that PioViewer should be open sourced. It would take a lot of pressure off the development team. Most of this thread will be filled with weird, unreasonable and nonsensical feature requests for PioViewer. The development team is probably more interested in working on PioSOLVER than PioViewer. I'm not just some hippie advocating for open source, it actually makes business sense to open source it. Open sourcing it gives people more reason to purchase PioSOLVER.
Karganeth is offline   Reply With Quote
Old 03-13-2015, 08:42 AM   #59
punter11235
Carpal \'Tunnel
 
Join Date: Mar 2005
Location: solving poker
Posts: 6,780
Re: PioSOLVER - postflop equilibrium solver for Holdem

Quote:
This is extremely vague, it's not clear to the consumer what exactly he's getting. It doesn't say how much earlier experimental features will be available (and likely it can't - developers can't know with any precision when software features will be implemented).
It is vague as at this point there are only plans, a bit better performance and priority support. That's why there is "contact us before you buy" in the description.
That being said I already have customers who are willing to pay 1100$ for that + insight into the plans + treating their feature request as priority. They trust we make it worth it for them anyway sooner or later and we will.
There is really no pressure for you to buy this version. You can always upgrade for more or less difference of the prices.

Quote:
It's very difficult for a user who doesn't have access to source code know how much work it would take for developers to implement a requested feature, so they aren't able to know what sort of features they can realistically request and expect to be implemented.
Again, I gave my edge customers honest look into state of the affairs. I don't know myself how long stuff will take but I have educated guesses.

Quote:
It seems odd to charge $624 for someone to essentially beta test software.
It is odd for you, it's great for others. Some of my customers are more than happy to take the risk of running some experimental version just to have earlier access to that. It will be happening with some major things coming in about 2-3 weeks.
Again, if you worry about it, just don't buy it for now. Wait and see, ask me in 2 months what is available.

Quote:
but edge undermines the confidence of people who don't own edge. Will their suggestions matter less than the suggestions of those who own edge?
Well, yes. That how "priority" is defined. If something is higher on the priority list then something has to be lower. We try to take care of all our customers, provide as much support as possible but yeah, edge customers will always be treated as higher priority.

Quote:
Will questions about upcoming planned features in this thread go unanswered because the person who asked them hasn't purchased edge?
Questions will be answered when reasonable estimation can be made.
The questions my edge customers have in mind are of the kind of "I need limit Holdem to run on 2 work stations 24/h, can you make it for me?" and then I tell them what kind of timeline we have in mind for that and consult them about which stuff is important.

Quote:
In using the program I found a small bug. If you press the delete key twice in the "Weight for selection:" field (in the Select range part) it will display the following message
Yes, this is a bug. Thanks for reporting.

Quote:
I have a few suggestions. My first suggestion is for PioViewer is to remove Hero and Villain terminology, and just replace it with IP and OOP
Tools -> Configuration and then at the bottom there is an option to use any names you please. We are currently shipping this version but I think free one is still not updated. It will be available for everybody in next PioViewer release which is planned for this Sunday.

Quote:
My second suggestion is to add a view that allows me to see Range, Strategy and EV all in one view.
Here is the screenshot from dev version which will make it to the next release (planned for Sunday):



Quote:
So each hand cell would be split horizontally into 3 segments to show all the data.
We did some experiments like that but things quickly become messy and not readable (for one such experiment see Tools -> Configuration -> show EV in main view)

Feel free to pass some design suggestions. I think it could work for ZOOM (to split into 3 parts) but not really for the main view. We are not exactly the best graphical designers though so any feedback on this is welcome.

Quote:
My third suggestion is to make the "Exploitable for:" show the value as a percentage of the pot. It is true that "Those numbers are only useful in relation to starting pot" so therefore it should always show those numbers as a fraction of the starting pot.
This is very nice idea. Adding to to-do.

Quote:
I'm actually in favour of normalizing all the EVs to be a fraction of the pot too but others may disagree with this.
If you mean the solver output normalizing all of them won't happen but adding this information at some point will.
If you mean the EVs displayed in the browser tab then it's already done. Tools -> Configuration and you can customize what is displayed in any way you please using the available variables. Like this:



However take into account that it's not exactly clear what "the pot" is. The configuration in that picture assumes that the pot is an amount which was in the middle before any actions on current street were taken.

Quote:
My final suggestion is that PioViewer should be open sourced
It's quite incredible that you suggest that as it was our plan until at least mid of January. We decided against it after doing some consultations with people better than us at business. We would love to do this and if we manage to convince ourselves that it won't directly harm us we will. I think the time to reconsider this option will come after we deal with initial fixes and features though.

Thank you for all your suggestions, it was really a pleasure to see such detailed look at our software.

Last edited by punter11235; 03-13-2015 at 09:10 AM.
punter11235 is offline   Reply With Quote
Old 03-13-2015, 11:03 AM   #60
chinagambler
veteran
 
chinagambler's Avatar
 
Join Date: Jun 2011
Location: Czech Republic
Posts: 2,514
Re: PioSOLVER - postflop equilibrium solver for Holdem

Would really appreciate if you could at least create a way to copy/paste the ranges into Flopzilla. Shouldn't be too hard to convert. A range like AsAh:0.123093396, KdKc:0.679629326, 4h4d:0.0908142105 would look like this in Flopzilla: [12]AsAh[/12], [68]KdKc[/68], [91]4h4d[/91]
chinagambler is offline   Reply With Quote
Old 03-13-2015, 11:39 AM   #61
Chabra
journeyman
 
Chabra's Avatar
 
Join Date: Apr 2013
Location: Above EV
Posts: 355
Re: PioSOLVER - postflop equilibrium solver for Holdem

Quote:
Originally Posted by chinagambler View Post
Would really appreciate if you could at least create a way to copy/paste the ranges into Flopzilla. Shouldn't be too hard to convert. A range like AsAh:0.123093396, KdKc:0.679629326, 4h4d:0.0908142105 would look like this in Flopzilla: [12]AsAh[/12], [68]KdKc[/68], [91]4h4d[/91]
Flopzilla can't handle more than 5 weights, and also can't handle 2 hands of the same holecards but different suits having different weights, so it won't work anyway. I've already tried it, and wrote a script to convert into the right format, but Flopzilla won't accept.
Chabra is offline   Reply With Quote
Old 03-13-2015, 02:13 PM   #62
xcEmUx
enthusiast
 
xcEmUx's Avatar
 
Join Date: Feb 2013
Posts: 57
Re: PioSOLVER - postflop equilibrium solver for Holdem

Probably a stupid question, but why is IP acting first on the flop ?

Spoiler:


€/: got it, had to check "Include first donkbet"

And another stupid question:
Let´s say CO opens 3bb, BU calls 3bb, Blinds fold.
-> Starting Pot otf should be 3+3+1,5 = ~8bb is that right ?


Thanks a lot in advance !

Last edited by xcEmUx; 03-13-2015 at 02:29 PM.
xcEmUx is offline   Reply With Quote
Old 03-13-2015, 03:00 PM   #63
punter11235
Carpal \'Tunnel
 
Join Date: Mar 2005
Location: solving poker
Posts: 6,780
Re: PioSOLVER - postflop equilibrium solver for Holdem

Quote:
€/: got it, had to check "Include first donkbet"
Yeah. This is useful for trees where OOP would check 100% anyway. The solution won't be (much) different then and the calculation will be faster and the tree smaller.
Usually those spots are BTN/CO/MP/UTG vs BB where leading the flop OOP isn't usually part of optimal play.
Quote:
Let´s say CO opens 3bb, BU calls 3bb, Blinds fold.
-> Starting Pot otf should be 3+3+1,5 = ~8bb is that right ?
It is right but because of some decisions we made in the code you can only use whole numbers for the pot and bets. It's the best to use 5/10 or 50/100 blinds to allow for enough flexibility in bet sizes. The maximum stacks you can use are 65536. The reason for that is that I was paranoid about memory usage when I was starting to code this solver.
punter11235 is offline   Reply With Quote
Old 03-13-2015, 07:08 PM   #64
marakesh
newbie
 
Join Date: Mar 2015
Posts: 22
Re: PioSOLVER - postflop equilibrium solver for Holdem

Am confused about the speed limitations of the basic version:

1 I have i7 6 cores with 12 threads. Will I be able to use 4 cores and 8 threads in basic version? What speed improvemnts should i expect in edge version?

2 Will a not coder be able to handle the scripting functions of edge version for nightly scheduling?

Also, =1 to opensourcing viewer. The current presentation layer is very rough and multiple people working on it should allow for exponential leaps. On the other hand, you will be able to work on optimizing speed and adding multiple bet sizings for the solver.
marakesh is offline   Reply With Quote
Old 03-13-2015, 07:29 PM   #65
+VLFBERH+T
grinder
 
+VLFBERH+T's Avatar
 
Join Date: Mar 2013
Location: at the crossroads
Posts: 529
Re: PioSOLVER - postflop equilibrium solver for Holdem

Quote:
Originally Posted by SiQ View Post
I'm considering making a video reviewing PioSolver. I've had a bit of fun playing around with it and I think I know my way around it but if you feel like there are any features I might miss that you want talked about feel free to shoot me a PM and we can swap details to have a quick chat.
Excellent !! Are you a coach at one of the coaching sites and will post it there, or will the video be publicly available, e.g. on your YouTube channel ?

Either way, please post the link to the video here, looking very much forward to it !
+VLFBERH+T is offline   Reply With Quote
Old 03-13-2015, 07:31 PM   #66
punter11235
Carpal \'Tunnel
 
Join Date: Mar 2005
Location: solving poker
Posts: 6,780
Re: PioSOLVER - postflop equilibrium solver for Holdem

Quote:
1 I have i7 6 cores with 12 threads. Will I be able to use 4 cores and 8 threads in basic version? What speed improvemnts should i expect in edge version?
Basic version runs 4 threads, that's it. The pro one runs w/e it's given (although I have to admit implementation for 32 hardware threads is not very efficient as of now).

Quote:
2 Will a not coder be able to handle the scripting functions of edge version for nightly scheduling?
The scripts are available in pro version. If you have 0 coding experience you will be able to run simple scheduling tasks (run a tree on 30 different boards, save all of them to files) but probably nothing more advanced than that.
It almost surely doesn't make sense for you to buy an edge version now unless you plan on heavy analysis/advanced scripting and need help with that + some additional minor features (I just run a tree with some branches cut with one customer for example, experimenting with measuring value of removing various options).

Quote:
Also, =1 to opensourcing viewer. The current presentation layer is very rough and multiple people working on it should allow for exponential leaps. On the other hand, you will be able to work on optimizing speed and adding multiple bet sizings for the solver.
Thank you for your suggestion. That decision will be reconsidered in about 6-8 weeks. Our preference is to open-source but we need to make sure it makes business sense for us.
punter11235 is offline   Reply With Quote
Old 03-13-2015, 08:11 PM   #67
Joshua
journeyman
 
Joshua's Avatar
 
Join Date: Sep 2002
Location: Monkey Tilt
Posts: 276
Re: PioSOLVER - postflop equilibrium solver for Holdem

I tried to sign up for the newsletter on your homepage but got an error message: "404 Page Not Found".

I also would like to add me to the list of people who would like a mac version

But I'm thinking of buying a stationary PC anyway. Do your software run alot better on an Intel Xeon processor or is Intel i7 enough for running simulations during the night?
(I'm guessing Intel i7 is enough now but maybe not once you add more complex features?)
Joshua is offline   Reply With Quote
Old 03-13-2015, 08:14 PM   #68
+VLFBERH+T
grinder
 
+VLFBERH+T's Avatar
 
Join Date: Mar 2013
Location: at the crossroads
Posts: 529
Re: PioSOLVER - postflop equilibrium solver for Holdem

Hi, first of all: even though I haven't learned yet how to fully utilize it, this is one amazing piece of software, for sure ! I bought it yesterday, and still need a little bit to get the hang of it, so here are my PioSolver noob questions:



In the above example (from your samples that come pre-loaded), the starting pot is "0 0 180", what does that mean ? The explanation states "One whole number representing starting pot", why are there three numbers here ?



When I use the "Choose Pot and Bets" button, I wish to use bet sizes of 70% of the pot on each street, but the way I would intuitively think I have to input that, apparently is wrong ...

So - as an example, say we play heads-up, blinds are $5/$10, effective stacks pre-flop are $1100 (pre-flop). Button (which is SB) raises to 2.5bb, i.e. $25, BB calls, so $50 are in the pot, and $1100 - $ 25 = $1075 effective stacks remaining. SB/BTN now bets flop, turn, river, each for 70% of the pot (I want to study which ranges BB calls down with on each street) -> how do I enter this blinds / stack size situation ? Sorry for this, as it might seem a goofy question to you, and those who have figured it out, but it doesn't seem all that intuitive to me [It doesn't seem like you are too fond of CREV, but in this detail anyway, their way of putting in blinds/rake, then "set stacks" is pretty straightforward / intuitive, but once I know exactly how it works with your program, I'll be just fine, only need to avoid inputting the starting situation and having the program understand it in a different way than I meant ...]

When we do not use the "Choose Pot and Bets" button, we can apparently only input the starting pot and bet sizes, but not effective stacks - is there some default effective stack value assumed, and if so, which one is it ?

I don't know who SiQ is, but he's gonna post a video on PioSolver somewhere. Are you planning explanatory videos as well ? They could be short clips, but would surely be appreciated by many (also, e.g. when replicating strategies shown in such videos, it helps that videos can be paused, which cannot be done with gifs).

Lastly: What does "MES" mean (as in "HERO's MES") ?

I don't know if the way of inputting the starting situation is intuitive to most others, so again, sorry if the answers might seem "obvious" to some.

Thank you for all the work that you did (and will) put into this project !
+VLFBERH+T is offline   Reply With Quote
Old 03-13-2015, 09:29 PM   #69
samooth
veteran
 
samooth's Avatar
 
Join Date: May 2009
Location: gutrotting
Posts: 3,111
Re: PioSOLVER - postflop equilibrium solver for Holdem

^
0 0 180 means Hero invested 0, Villain invested 0, and the pot is 180. you can just type 180 into the pot field and that's fine, though. you can also see what those numbers mean as you click through the game tree under the browser, at the upper left corner.

70p 70p 70p 1075 -- the last number represents effective stacks as the bet amounts are cumulative (!). alternatively, you can just try and figure out what those bet sizes turn out to be in bbs, in your case ~ 35 120 320 1075 -- all this is for the case of 4 bets in (35, 85, 200, 755). the last number or whatever the last number corresponds to when you use Xp is the effective stack size.

MES means something like maximal-exploitative strategy -- when we fix the current strategy for Villain, and Hero plays his best response to that, then Hero's MES denotes the EV of this best response strategy. you can also take EV Hero and add the "exploitable for" number to it, it'll be equal to Hero's MES. this number also means that Hero can't do any better than that, now matter how he plays, against the current strategy the solver gives for Villain.
samooth is offline   Reply With Quote
Old 03-14-2015, 05:17 AM   #70
punter11235
Carpal \'Tunnel
 
Join Date: Mar 2005
Location: solving poker
Posts: 6,780
Re: PioSOLVER - postflop equilibrium solver for Holdem

Quote:
I tried to sign up for the newsletter on your homepage but got an error message: "404 Page Not Found".
Yeah, that thing doesn't work. I will look into it.

Quote:
I also would like to add me to the list of people who would like a mac version
Our customers made it working on Macs via both Bootcamp and Parallels. Bootcamp is probably a safer bet but it requires booting your computer into Windows for using the solver which is far from ideal for many users.
Unfortunately I can't offer any support for those options. Maybe try a free version first.


Quote:
But I'm thinking of buying a stationary PC anyway. Do your software run alot better on an Intel Xeon processor or is Intel i7 enough for running simulations during the night?
(I'm guessing Intel i7 is enough now but maybe not once you add more complex features?)
Some benchmarks I have available (those are i7 specific compiles but generic version is only marginally slower):
-i7 3770, 3.4GHZ, DDR3 800Mhz - 15.2 seconds
-hexa core i7 (newer than the one above but not sure exactly which model): 11 seconds
-16 core Xeon (32 threads): 7s

The difference isn't bigger because the multi-threading implementation isn't the most efficient right now. My quick attempt to make it better failed yesterday so it needs to wait.
I think I would go for hexa core i7 personally.

Few things to add to samooth's answers as to interface questions:

Quote:
In the above example (from your samples that come pre-loaded), the starting pot is "0 0 180", what does that mean ? The explanation states "One whole number representing starting pot", why are there three numbers here ?
So this 0 0 180 thing is a format used internally by the solver. First number is what OOP player invested so far postflop, 2nd is what IP invested so far and 3rd one is what the pot was on the flop (so "dead" money).
That format is also used in node description in the browser tab. Originally it was planned that you could use that to force a postflop bet for example (as I did in some early tests) but then we've realized that 99.99% of use cases don't need that.
The pot field now accepts single number and internally translates it to 0 0 that_number but it can still accept 3 numbers in old format.

Quote:
When I use the "Choose Pot and Bets" button, I wish to use bet sizes of 70% of the pot on each street, but the way I would intuitively think I have to input that, apparently is wrong ...
Well, yeah only absolute numbers work there.


Quote:
[It doesn't seem like you are too fond of CREV, but in this detail anyway, their way of putting in blinds/rake, then "set stacks" is pretty straightforward / intuitive
The problem is that with decision tree software you usually build very limited trees. The full postflop tree however has a lot of obscure branches like: bet-call, check, check, check-bet-raise-allin. The solver needs to know what the sizing is there as well. That means that setting every bet sizes manually would require about 15-20 minutes of work and we wanted to make very simple automatic way to build reasonable trees so people can actually start using the program.
It is true though that this interface is both limited and not very intuitive. We are working on improving both factors. It takes time however as this is one thing we really need to get right as big part of the experience depends on it and it will be very difficult to change later once people have their habits and saved configs depending on it.

Quote:
When we do not use the "Choose Pot and Bets" button, we can apparently only input the starting pot and bet sizes, but not effective stacks - is there some default effective stack value assumed, and if so, which one is it ?
Those are calculated from the bets assuming the last bet is all-in. So yeah, something like:

65p 65p 72p 975 means bet 65%, 65%, 72% and then all-in for 975 (total). It's a decent description of typical raised pot with 97.5bb effective stacks at 5$/10$

Or in 3bet pot with starting pot of 180 and effective stacks of 910 you can do:

100 310 910 which means bets of 100, 210 and 600. The amounts are cumulative.

Quote:
Are you planning explanatory videos as well ? They could be short clips, but would surely be appreciated by many (also, e.g. when replicating strategies shown in such videos, it helps that videos can be paused, which cannot be done with gifs).
Yeah, I haven't realized the GIFs are hard to follow for that reason. The videos are likely to appear at some point but I still hope other people do it for me
I am not a native English speaker and I am not good at talking continuously to the microphone without having someone responding to what I say. I will try one day but it would really better for everyone if we can get some experienced coaches on board.

Quote:
I don't know if the way of inputting the starting situation is intuitive to most others, so again, sorry if the answers might seem "obvious" to some.
Again, it is not very intuitive when you are starting but it has advantage of simplicity and make it possible to build big reasonable trees fast.
We are working on making it better

Last edited by punter11235; 03-14-2015 at 05:27 AM.
punter11235 is offline   Reply With Quote
Old 03-14-2015, 09:03 AM   #71
+VLFBERH+T
grinder
 
+VLFBERH+T's Avatar
 
Join Date: Mar 2013
Location: at the crossroads
Posts: 529
Re: PioSOLVER - postflop equilibrium solver for Holdem

Quote:
Originally Posted by samooth View Post
70p 70p 70p 1075 -- the last number represents effective stacks as the bet amounts are cumulative (!). alternatively, you can just try and figure out what those bet sizes turn out to be in bbs, in your case ~ 35 120 320 1075 -- all this is for the case of 4 bets in (35, 85, 200, 755). the last number or whatever the last number corresponds to when you use Xp is the effective stack size
Quote:
Originally Posted by punter11235 View Post
Again, it is not very intuitive when you are starting but it has advantage of simplicity and make it possible to build big reasonable trees fast.
We are working on making it better
Ha, cool, thanks, got it now But then again, looking forward to a possible overhaul of this part, which I am sure will come anyhow when you implement more flexible bet sizes.

Quote:
Originally Posted by punter11235 View Post
Yeah, I haven't realized the GIFs are hard to follow for that reason. The videos are likely to appear at some point but I still hope other people do it for me
I am not a native English speaker and I am not good at talking continuously to the microphone without having someone responding to what I say. I will try one day but it would really better for everyone if we can get some experienced coaches on board.
I am not a native English speaker myself, but seeing that you write English very well, I am sure you speak just fine, and for the sake of getting the most out of this program, I am very willing to put up with a very heavy accent and / or otherwise less than perfect presentation.
+VLFBERH+T is offline   Reply With Quote
Old 03-14-2015, 02:14 PM   #72
NL Loki
adept
 
Join Date: Feb 2015
Location: Southern Hemisphere
Posts: 774
Re: PioSOLVER - postflop equilibrium solver for Holdem

Are we making a group to investigate frequencies with different flops? Will u make a forum to allow researchers to connect and share results
NL Loki is offline   Reply With Quote
Old 03-14-2015, 05:06 PM   #73
chinagambler
veteran
 
chinagambler's Avatar
 
Join Date: Jun 2011
Location: Czech Republic
Posts: 2,514
Re: PioSOLVER - postflop equilibrium solver for Holdem

Quote:
Originally Posted by NL Loki View Post
Are we making a group to investigate frequencies with different flops? Will u make a forum to allow researchers to connect and share results
Would be interested in joining a Skype group or private forum for people who bought the software, would be great added value.
chinagambler is offline   Reply With Quote
Old 03-14-2015, 05:10 PM   #74
+VLFBERH+T
grinder
 
+VLFBERH+T's Avatar
 
Join Date: Mar 2013
Location: at the crossroads
Posts: 529
Re: PioSOLVER - postflop equilibrium solver for Holdem

Quote:
Originally Posted by chinagambler View Post
Would be interested in joining a Skype group or private forum for people who bought the software, would be great added value.
Yes, me too !
+VLFBERH+T is offline   Reply With Quote
Old 03-14-2015, 07:11 PM   #75
Karganeth
old hand
 
Karganeth's Avatar
 
Join Date: Jun 2008
Posts: 1,305
Re: PioSOLVER - postflop equilibrium solver for Holdem

Quote:
Originally Posted by punter11235 View Post
It is vague as at this point there are only plans, a bit better performance and priority support. That's why there is "contact us before you buy" in the description.
That being said I already have customers who are willing to pay 1100$ for that + insight into the plans + treating their feature request as priority.

[...]

Well, yes. That how "priority" is defined. If something is higher on the priority list then something has to be lower. We try to take care of all our customers, provide as much support as possible but yeah, edge customers will always be treated as higher priority.
I agree that supporting them first makes sense, but I don't see adding in new features as a form of support, I see it as development of the software that affects all users. It seems strange that as a form of support to 1 particular user changes are made that affect all users.

In my view, prioritizing the feature requests would only make sense in a case where you had already sold all the licenses you were ever going to sell, and it was just a matter of what order you added the features in. There it would be fine to add the features of the premium users first. But that's not the case here.

Your profits are partially dependent on attracting new users. If the suggestions of edge users always perfectly aligned with what suggestions would attract new users then would be no ambiguity: you would simply always choose to add whatever suggestion edge users said. But edge users suggestions might not align this way. So with the existence of the edge license people are able to pay to take the development of PioSOLVER and PioViewer in a direction it wouldn't have gone in otherwise (that is, it takes it away from the direction that would sell more licenses).

If you give someone more priority it implies other people are losing priority: it isn't free to give priority to edge users because it means other license owners lose value - their feedback matters less so they expect less value from the product, so fewer people will buy it. In my view suggestions should be added according to the merit of the suggestion alone, not on who suggested them.

Quote:
Tools -> Configuration and then at the bottom there is an option to use any names you please. We are currently shipping this version but I think free one is still not updated. It will be available for everybody in next PioViewer release which is planned for this Sunday.
Thanks for the help. I renamed OOP Player to OOP and IP Player to IP. I don't see how being able to name players anything other than OOP and IP adds any value, so in my view this is just adding unnecessary complexity to the program and is a feature that should perhaps be removed.

Quote:
Here is the screenshot from dev version which will make it to the next release (planned for Sunday):

I strongly dislike this view for one simple reason: the value of one data set (range) affects my ability to see the value of a separate data set (strategy). The less often a player has a hand the harder it is for me to see what he does with those hands. That's not ideal.

Quote:
We did some experiments like that but things quickly become messy and not readable (for one such experiment see Tools -> Configuration -> show EV in main view)
This view is almost what I want but the way it displays EV is not consistent with how it is normally shown. Here is the normal EV next to the combined EV view.





As you can see, in the second case the coloring of the EV is split according to what the player is doing with that hand. So for 98s instead of going form yellow to orange to red in a smooth way, it goes from yellow to orange to yellow to red. This makes it harder to read. I think it should be not be split like this, and just be displayed in exactly the same way it is for the full EV view.

Quote:
Feel free to pass some design suggestions. I think it could work for ZOOM (to split into 3 parts) but not really for the main view. We are not exactly the best graphical designers though so any feedback on this is welcome.
One of my favorite subreddits to frequent is http://www.reddit.com/r/dataisbeautiful/ which describes itself in the following way: "DataIsBeautiful is for visualizations that effectively convey information.". It's full of novel ways to display data and debates over those very visualizations and their merits. If you want to learn what makes good data visualization then I recommend visiting here. I still believe that giving the user an option to have 3 views would be best. Instead of having 3 buttons for Range, EV and Strategy there could be a check box that you tick for each of the views, and if you ticked all 3 it would show all 3. This would also remove the need for the "Ev + strategy in main view" setting, which reduces complexity. However this layout comes at the price of an extra mouse click for someone who wants to switch between a pure view of one to another pure view (as it requires them to unclick then click the other one instead of just clicking the button once).

Quote:
If you mean the solver output normalizing all of them won't happen but adding this information at some point will.
If you mean the EVs displayed in the browser tab then it's already done. Tools -> Configuration and you can customize what is displayed in any way you please using the available variables. Like this:



However take into account that it's not exactly clear what "the pot" is. The configuration in that picture assumes that the pot is an amount which was in the middle before any actions on current street were taken.
Thanks I had forgotten about that feature to input my own ev formula. What you have here is almost what I want. I'm currently using the formula (ev + current) / dead for the value, making all the evs in terms of a fraction of the starting pot. But what I really wanted was to have evs as a fraction of the current pot. There is a variable that stores how much money the active player has put into the pot on previous streets and the current street, but there is no variable to store how much the inactive player has put into the pot. It is possible to deduce how much he has put in on previous streets by simply doubling how much the active player has put in, but it's not possible to know how much he has put in on the current street.

So if the pot was $100 I checked then he bet $50 and my ev was $75, I want this to be displayed as 0.5 (half the current pot), not as 0.75 (pot at the start of the street). A new variable must be added in order for this to be possible.

I do have an issue with how piracy is dealt with. On the website it says

Quote:
There is nothing we can do against people cracking our software. Here is some useful information to make a decision about supporting such a venture easier:

[...]
Attempts to tamper with the binary will result in the following:
Tampered PioSOLVER will produce non-sense results in subtle ways (not immediately obvious to tampered binary user)
This approach to combating pirates is extremely ineffective and incredibly harmful to honest buyers of your software. First of all, pretty much anyone attempting to crack PioSOLVER is always going to be comparing the results to a copy that he hasn't tampered with, so he will know with certainty if the software has detected his tampering and begun to garble its results. But much more importantly, this is extremely bad for people who are honest buyers. It would be incredibly naive to think that this defence mechanism will never be activated against someone who has bought a license in good faith. There is always collateral damage. Maybe some antivirus software thinks PioSOLVER is some kind of virus and attempts to mess with its binary, inadvertently triggering PioSOLVER's anti pirating mechanism. Or it could be anything, there's no way to anticipate every possible scenario.

But even if it did never mess up the results of someone who bought a license, license owners don't believe (and shouldn't) that it has 100% perfect accuracy. They believe it might mistakenly trigger on their machine, messing up their results, and they would never know. This passive aggressive way of dealing with pirates is not only ineffective, it also harms the perceived value of the product (and thus fewer licenses are sold). The engine should just disable itself if tampering is detected. This way if it is accidentally triggered the user knows and can contact support.

Anyway, there is a typo in PioViewer. When "Load recently used tree" is used, the title changes from PioViewer Unnamed to PiovViewer (followed by the address of the file). How has no one else noticed this?

Last edited by Karganeth; 03-14-2015 at 07:20 PM.
Karganeth is offline   Reply With Quote

Reply
      

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off


Forum Jump


All times are GMT -4. The time now is 10:18 AM.


Powered by vBulletin®
Copyright ©2000 - 2017, Jelsoft Enterprises Ltd.
Copyright © 2008-2010, Two Plus Two Interactive
 
Poker Players - Streaming Live Online