hi punter, i know its been asked before, but, in the next update, can we get better colour coding for betsizes? the current bet/betbig creates quite a gradient when its more the 3 sizes lol

Hi Punter. Thanks so much for your previous answers. Got another couple questions.

Question 1: If we run a preflop spot say 35bb deep where let's say SB opens 100% HU on BTN and we give BB a fixed strategy of 3betting 10% with the option to call or fold to a shove in which we don't allow flats and compare that same situation but SB opens only 50% and everything else stays equal for the BB, shouldn't the EV of a random hand stay the same since the BBs strategy is fixed for 3betting? I ran two SIMs only changing opening range (fixed strategy for BB stayed the same) and got different equities for AQs and some other hands and am not sure exactly why that is?

Question 2: When running preflop on AWS (Amazon) what specs should we be looking for in order to run preflop for simple spots and then for more complex spots for the following:

Family
Type
vCPUs
Memory (GiB)
Instance Storage (GB)
EBS-Optimized Available
Network Performance
IPv6 Support

Thanks

Feature request: In the Full Clone window, it would be nice if it also included the frequencies for each action like it does when looking at the 13x13 window in the main view. Otherwise, to compare frequencies after performing a node lock and such, you have to write down the frequencies or take a screenshot.

It's impossible to comment on a tree without seeing it with all the details. To share the configuration do the following:

1)go to treebuilding and calculation tab, fill the config (or load the save)
2)click "copy to clipboard" button
3)go to pastebin.com or similar service, paste there
4)link

This way I can copy the configuration and paste it to my Pio (by going to Tools->paste treebuilding configuration) and then solve it locally.

We will see, it's not clear what is "better" though as we think the gradient is pretty good. You can also edit it (from what color to what color the gradient is created). Requests like this should be accompanied by solutions you see as better

From your description it sounds like the EVs should be the same if both trees are solved to good accuracy. Feel free to share the config either here or by email if it's not too big (so I can run it myself).

Number of CPUs and Memory. Make sure they are all Intel Xeons as well.

It seems to me it works that way already, you can also choose ranges for specific actions, copy them and paste to node-locking interface.

Glad you asked,
Just add more bets that can get collored in different collours and that not creates gradient between them, something like:
bet1, bet2, bet3, bet4, bet5, bet6, allin
where bet1 is the smalles, bet6 the largest (except allin) and each can get any collours we make like.

This also takes care of another request that asked for bets to be represented (in the tree browser) as % of the pot rather then the actual chip amount.

I'm not sure if it's been asked before but is it possible to nodelock for future street action? If it's not available could it be in the future somehow?
I imagine if a player barrels turns a really low % Piosolver would want to checkraise flops more often for value and maybe call or float more with some hands.

Hey I am interested in buying your software but have a question. I've noticed that the current PioSolver version is 1.9.2, if I buy PioSolver right now and it gets updated to 2.0 will I be able to update my software for free? If this is not the case, is there an estimated time when version 2.0 will be available to buy? Thanks in advance.

Hi - having issue with getting piosolver edge onto my laptop (already have it on desktop) when I try to open it, it says "Cannot open because Turboactivate.dll file is missing, try reinstalling the program" I have done this but still no luck. Any idea? Thanks

Ok
1st tree
100% open range for OOP https://gyazo.com/002a79c9431d90fff3aa3aa5a0578902
3bet by IP player https://gyazo.com/66c6e8be336179b3472698c7de05b33d
results: https://gyazo.com/3d7514d6f4231ebab60cda85ea51fdad
AQs for specific example: https://gyazo.com/d5a8f999d0b4797f94ff6267b9f03a2f

2nd preflop tree:
18.55% opening range https://gyazo.com/f0a1ba03a9beeaa6b7a7b10a29ea6044
3bet IP player same range: https://gyazo.com/d8a0c49dd14434dd4bdeab818c6f3428
results: https://gyazo.com/1ddc3bf0f800507bee1a4ca2ce36dec4
AQs for specific example: https://gyazo.com/4f3db8f0c1e604321cbb2b121b211774

As you can see AQs for example have different results. I figured with all things being equal if 3bet range stays fixed that the EV of hands should stay the same regardless of the opening range. Maybe I am missing something and you can explain. Strictly talking about differences in EV, not adjustment in range % as I understand why those change.

I can't promise it happens in the next release but it will be on the to-do.

Node-locking in terms of % of actions likely won't happen as the naive approach doesn't work very well and at this point it's not worth researching ways to make it work (as that's time-consuming and without guarantees for success). Improvements to node-locking interface including locking in more general terms will happen though.

This is a problem with our business model (in comparison to subscription): not only we need to spend time developing the software without being paid first, we then need to both provide updates and make sure we are not left with nothing once the customer base stops growing fast enough. This is a bit of a ****ty spot but you are getting lifetime license in exchange.

Here are some things which hopefully help you make your decision:
-2.0 won't happen in next 3 months and likely longer if it happens at all
-there will likely be 1.10 version
-if there is 2.0 version there will be significant discounts for existing customers and we will take care to make it less painful for people who bought the license close to the cutoff point

That being said at some point any software is at the point that either it's worth working on it more and finding ways to get paid for that work or abandoning the software altogether. Pio became very popular and there is a lot demand for new things (in fact there is way more desirable things to implement than we will ever have time to do) so 2.0 version is likely at some point. Good news is that you are already getting software which took some years to develop and for which we provided a lot of free upgrades/new features.

The most common reason for that is that PioViewer was moved from the original folder. Make sure it's there along with all other files and you are running it from there. If that works you can make a shortcut to the desktop later.

As a rule I don't look at screenshots containing results/configs as that just makes too difficult to replicate locally and too easy to miss important details.
I've described a way to share the config. It's the best way because once I have it I will be able to recreate the tree exactly, verify it's solved and look at the results in all important spots. It's just not happening with screenshots.
It happened more than once that I digged into screenshots, spent time thinking for reasons for unexpected behavior, tried to create similar trees locally just to to discover after many emails/hours that the results are from 2 different configs, not solved to similar accuracy, one contained rake and other didn't etc. I am mentioning it to make sure you don't think I am nitpicking and refusing to look at it. It's just a way to share trees is to copy and provide the full tree building configuration

Hi punter,

I'm interested in taking piosolver results and exploring them in a custom UI. Is there documentation that explains the format the strategies are stored in? I didn't see anything in the scripting pdf: https://cdn.shopify.com/s/files/1/07...68544248083387.

Specifically, I want to know how to find a node (eg, "turn card is Ac and action is check") and then how to read the card distributions and strategies at that node (eg, "2h2c is held 50% of the time and strategy is bet 50%, check 50%, fold 0%").

Thanks.

Well, this pdf describes exactly how to find nodes and what format the results are in. It also describes the way to see how PioViewer obtains those results (you can log them and open it in console visible mode to see the communication as it's happening).

Maybe one thing that is not 100% clear is the order of 1326 numbers show_strategy/show_range etc. use. This order is described by "show_hand_order" command.

You need to use a proper NodeID, the way they are created is described at the top of the linked pdf. You can also see them in the viewer when browsing the trees (left upper corner).

Maybe I don't understand what you are asking though, feel free to ask more questions

Probably my fault. I just perused it quickly. I'll take a longer look tonight at the commands you mentioned and if I can't figure it out, I'll ask some follow up questions. Thanks.

By far the best way is to enable logging and open PioViewer in console visible mode (it's described how do it at the top of that pdf) and then just start solving/browsing a sample tree. You will quickly see what is happening and what commands are used.

I meant aggregate frequencies for the whole range, not individual hands.

I totally understand. Apologies for that.

Preflop Tree 1: http://pastebin.com/HSvGjFUE

Preflop Tree 2: http://pastebin.com/6tssujnR

Let me know if I did it right this time

Hey Punter,

Is there a way to lock preflop nodes? I see that it's possible to enter starting ranges, but I don't see a way to enter and lock preflop nodes.

This was helpful, thanks. I get how the commands work, and I understand how I can extract and interpret the info I want given that I'm connected to the engine. But I have two followup questions:

1. What is the preferred way of connecting to the engine from the scripting language of my choice? I could create a temporary text file with the UPI commands I want to execute in my ruby/node/etc, invoke the system command "PioSolver-pro19.exe temp.txt" and then parse the output that comes back in my scripting language -- but is there a better way?

2. Is there a way to access the strategy info I want directly from .cfr file, without needing the engine to interpret it? Or is the binary format too complex? If I did want an export of the strategy in a usable format, do you recommend I create UPI commands to grab the info I want and then save it in a format that's convenient for me? I just don't want to reinvent the wheel if I don't have to.

Thanks again.

Hi Punter,

thx for you reply, ill tried and used pastebin as you recommended, let me know if I did everything correct.

Spend quite a lot of time understand how pio thinks, here is a spot:

http://pastebin.com/zfM81tck

sb/bb 15bb deep limped pot, (so i discounted and weighted out some hands that preflop might wanna jam), flop Ah Jc 6s Turn Jh River 2c .

1) PIO suggestes 3barreling 8s7c but likes to check 8d7h on the river, what is the reason for that?

2) Similar question to my previous post, PIO suggest betting 87o 100% on the flop (the square is filled 70% with red, as I weighted this hand preflop 70%, which is fine)

On the turn the square is only filled a tiny amount red, and is mostly grey, but havent we still all our 87o hands in our range? And shouldnt it be 70% filled red instead?

https://gyazo.com/a626f5e8b9ada1f60118ddac619385a8

Thanks a lot for you help !

Ok, added to the to-do but no promises.

It looks correct. I've run it without postflop play for now and the results check out:
https://gyazo.com/7e760b9900c7bb4263535d3913c54c52

and:
https://gyazo.com/3d243253c19d1571294f624d375f7a25

I will run it with postflop play later today (it will take a while) and see if anything changes.

It works the very same way as locking postflop: build the tree (without solving it), go to the browser, choose the node you want to lock, enter nodelocking interface (by hitting ctrl+n) and change strategies/lock nodes there.

It's a good idea to test it on trees without postflop play first (pure preflop trees) because they are solved instantly so you can experiment and see if locking works the way you want.

Every programming language have an interface to connect to console programs. If you are using C#, drop us an email I think we still have some old connection code which would be a good start. In Python you have popen interface in standard library but there are other options as well. In general you should look for opening processes and rediricting their stdin/stdout so you can work with them.

There isn't and it's a bad idea. The format changes, it doesn't contain all the information (a lot is recalculated on the fly) so not only you would need to understand the format but also write significant parts of the solver's logic yourself to be able to access all the information.

As a cautionary tale some people reverse engineered the format in 1.2 version which was way simpler than it is today (because many optimizations weren't there). The result is that they are still stuck at 1.2. It's a very bad idea to base anything on the save format, it will be slower, it won't work and you will regret it. That being said the format is not a huge secret and isn't obfuscated in any way so I can answer some questions about it privately if you are determined following this route

Yes, that's the way or you can just ask for information once it's needed and don't store anything yourself (base everything on UPI commands), that has advantage of being more universal and will work with future versions as well as other solvers (for now Jesolver).

In general the why question is very difficult to answer. Pio just follows the process converging to an equilibrium but doesn't really have any reasons. It just adjust strategies at every step and arrives at the result. To answer the why question you need to look for patterns/justifications but they will always be just human explanation and there is no guarantee you don't miss deeper reason.

It seems it wants to bet spades more, looking at calling range after the river shove it in fact looks like 8s7c blocks more hands (8s6x mainly) than the other combo but the differences are small (in EV). Those things are hard to understand and maybe it's not the most important thing as even very small differences in EVs of actions may result in huge differences in strategies (if you need to include some hands in 3barrel range you will end up barreling the best ones even if they are only marginally better).

That being said it seems Pio thinks the difference between good 3barreling 87o combos like 8s7c and worse ones like 8d7h is quite big. I honestly don't know why, it makes sense to look what exact opponents 86/76 combos are blocked and why but I wasn't able to spot it after a few minutes of looking at it. It only reminds me that we need to add an option to show how much given combo blocks the calling range.

It wants to bet 87 100% but you have two bet sizes and it choose the bigger one almost exclusively, look:
https://gyazo.com/879183dca905536d0720b28f229d4f01

You are then following a path of smaller bet so you don't have 87o that often there.

Coming back to it:

It seems you removed some hands from OOP range which would like to call. I solved both trees with 74 flop subset and I got this for full OOP range:
https://gyazo.com/9b0015d7971f4740ceb8db6042ce1792

and this for limited OOP range:
https://gyazo.com/b27f7de7ebb3ce084e7eb6c3f707faa7

As you can see hands like low suited connectors/1gappers would call if they were in range. That means the overall strategy is different and of course EVs will be different as well (as you have a different range in the tree you play it differently in optimal settings).

Say I'm interested in how often an out of rhythm check raise is optimal as the OOP player. So spots where I c/r the turn as the last flop aggressor, or where I c/r the river as the last turn aggressor. I could manually construct all the action sequences meeting these criteria, and then iterate over all possible turn and river cards, calling "calc_line_freq" for each relevant nodeID. Which is a fine solution, but I was just curious if there was an even easier way to address questions such as this?

Calc_line_freq returns frequency for the whole line (so aggregated for all turn/river cards following a given line). If you want individual frequencies you need to call "show_strategy" with relevant NodeID.
You may list all lines first using show_all_freqs command, maybe it's easier to work with that. In the future versions there will be also "show_all_lines" command to make the task for listing all lines fast and convenient.

I emailed the developer today about running Pio using wine on macOS (Crossover). He told me some guys managed to get it to run, but he cant give me their contact info. Would be awesome if some of those guys could help me out if they read this. I managed to launch PioViewers GUI but now am not able to activate it. The program crashes when submitting the key (.NET 4.5.1 is successfully installed). Also I am getting an exception error when launching (but I do think this does not prevent PIO from working, seems like a font error): System.ArgumentException: Value of 'Infinity' is not valid for 'emSize'.