You don't full range of yours as well or at least what the opponent expect you to have at given point. The reason for this is that optimal strategy depends on the range of both players not only one of them.

The free version solves one flop (Qs Jh 2h) or any turn/river tree. You would normally specify just the flop and just browse the tree as the solution contains all turns/rivers. In case of turn solutoins it contains all the rivers so no need to specify that.

With no programming experience, how much time would you expect it to take to get reasonably competent at scripting? Is basic automation fairly straight forward? What kind of advanced use cases are there?

Thanks for any input.

Thanks for the explanation. Is it possible to setup a scenario like in pokersnowie (it assumes ranges) where you define the two ranges and you can calculate the ev for each type of move on flop, turn and river based on your two hole cards?

For the most basic case (calculating the same tree structure on many flops and saving the results) there is script generation tool which is really straightforward to use.
This is explained here:
https://www.youtube.com/watch?v=VhG_YsEvfwU (question number 6)

A bit more advanced use case of combining several scripts into one and running it all at once you need a bit of trial and error but that's pretty straightforward as well.

In general there is no limit for scripting as all of the solver's functionality is exposed so you can for example code a full graphical interface using the solver. That will require being a fairly competent programmer though.

That's in general how the solver works, you define:

1)starting ranges
2)board (for example flop)
3)betting structure

The solver then calculates strategies/EVs/EQs for all the hands in all the possible spots.
I invite you to watch the quick start video and follow it with the free version so you can get a hang of it (the free version solves Qs Jh 2h flop so the video is possible to follow).

Hi punter, awesome program! I just bought the basic version and really liking it so far. A few questions.... is there anyway to analyze why piosolver does certain things with certain combinations of hands faster/efficiently?

For example, I looked at a sb vs btn 3! pot, on a flop of Jc 9d 2d

It took me a while to figure out why it prefers slowplaying middle set without a club (9h9s) vs fastplaying middle set with a club (9c9h,9c9s). After analyzing the flop raising ranges for both players, it dawned on me that the reason is because with 9h9s we block both combos of top two, Jh9h and Js9s.

But when taking a look at top set, something strange happened, sb prefers fastplaying without the Jd (JsJh) and slowplaying with the Jd (JdJh, JdJs). This seemed counter intuitive at first, because with JsJh we block both top two combos (Js9s, Jh9h) but then after some time I realized this is because when you dont have the Jd, you unblock hands like AJdd, KJdd, QJdd, JTdd etc that you want to raise gii with otf.

But for some odd reason the btn prefers an opposite strategy! Pio says that btn should fastplay with the Jd (JdJh, JdJs), and slowplay without it! This doesnt make sense!

I have been trying to figure out why for the past hour, is there a tool or something that shows/explains why it takes certain actions with certain combos? I'm guessing not, and thats part of what makes piosolver such an amazing tool for top level players, it really makes you critically analyze every detail/nuance



Btw, on a seperate topic, I am thinking of building a new pc to take advantage of piosolver edge.

I know I need alot of ram, should I aim for 64gb or 128gb?

Also, What cpu would you recommend? Does edge make use of more cores? Will I notice a big difference between 10 core i9-7900x vs the upcoming 18-core, 36-thread i9-7980XE?

Or perhaps would you recommend purchasing AMD Ryzen Threadripper 1950X? It offers a 16 core cpu at 3.4 ghz and is available now

Unfortunately the why question is often very hard to answer. The differences are small as well. One important thing is to compare EVs for both actions. If it "prefers" fastplaying 9c9h but the EV difference between raise and a call is miniscule (or there is not a meaningful one at all in case of mixed actions) then it doesn't make as much sense to focus on decision with this specific combo but it makes more sense to focus on overall range composition. On the other hand, especially on the turn and river you will see clear differences between various combos based on blocker effect.

One thing to realize is that you most likely need some kind of balance when it comes to blockers in betting and checking range. This means some hands must go to a bet and some must go to a check. Which one specifically go to which range may often be caused by very small differences in EV. Maybe sometimes you will be able to detect the pattern and answer the why question but very often you just won't be as the tree is huge and it may be impossible see the clear reason why the balance ends up being the way it is.

The kind of decisions you are talking about (bet/check on the flop) won't be easy to make clearer for reasons above. We have some ideas for turn/river decisions though where usually the blocker effect is more pronounced.

That depends on a budget as RAM is expensive. You will be able to solve a lot of practical trees with 64GB and maybe you won't need more. If you like more detailed solutions and experiments though you will need more. In my view there is little reason to run trees bigger than 128GB although many users disagree

One thing to consider is renting a dedicated server or cloud instance instead of buying your own hardware. This way you have more maneuvering room. You can for example rent 64GB server for 2months and then if you want to solve some bigger trees you can decide to rent a 256GB one for a while. This is a question of preferences and logistics but you should at least consider renting.

Pro version works with up to 16 threads. The edge one doesn't have a cap (technically it's 128 at the moment but we will increase once 64+ core CPUs become available).
It scales pretty well as well. Up to 32 cores it's almost linear and it's pretty decent iwth more than that as well.

With CPUs from similar generation the speed is more or less proportional to number of physical cores multiplied by base frequency so:

1)10 core 7900x has 10 x 3.3 = 33
2)18 core 7980XE has 18 x 2.6 = 46.8

7980XE is going to be about x1.41 faster then. In general value of Xeons is in dual box setups (so 2 CPUs at the same mother board) as they offer way worse performance/price ratio.

Previous Ryzens 1700 and 1800 have great performance when it comes to Pio although they weren't stable for all users. You needed to be careful with RAM you buy and there were some bugs in them when running at 100% (apparently it was only discovered in Linux compiles but some of our users experienced crashes with them as well; it's very hard to say what the reason was but likely either unstable RAM or the same bug which caused crashes on Linux; on the other hand other Pio users run them basically 24h/day and they are stable for them).

Apparently TRs are free of this issue though. I would wait for some tests and reviews but they will likely be the best deal on the market when it comes to high performance reasonably priced CPU.
Looking at Pio benchmark times Ryzen 1800 produced I think TR will outperform 7980XE by a small margin while being significantly cheaper.

Thank you for the awesome and detailed response punter! I look forward to seeing what tools/ideas you come up with to allow us to more easily see/visualize the blocker effect on the turn and river.


As for the computer, I've spent the entire day fleshing out a build, and I am about to hit the purchase button If you get a chance, let me know what you think about my build. After I setup the rig, I'll be upgrading from basic to edge and utilizing piosolver to further enhance my game in live cash.



CPU: Threadripper 1950x
https://www.amazon.com/AMD-Threadrip...adripper+1950x
$1000

MOTHERBOARD: GIGABYTE X399 AORUS
https://www.amazon.com/gp/product/B0...KIKX0DER&psc=1

RAM: (128gb total): G.SKILL TridentZ 64GB (4 x 16GB) DDR4 3200
https://www.newegg.com/Product/Produ...63&ignorebbr=1

Storage: Samsung 960 Pro 1tb m2
https://www.amazon.com/Samsung-960-P...amsung+960+1tb
$580

Graphics: ASUS ROG STRIX GTX 1080 TI
https://www.amazon.com/gp/product/B0...KIKX0DER&psc=1

Cooling: Nzxt Kraken x62 all in one Cpu liquid cooling
https://www.amazon.com/gp/product/B0...PN1BX9I1&psc=1
$160


Power Supply: EVGA SuperNOVA 850 P2
https://www.amazon.com/gp/product/B0...KIKX0DER&psc=1

Optical Drive: LG Super Multi Blu-ray Disc Rewriter
https://www.amazon.com/LG-WH16NS40-I...ds=LG+WH16NS40
$64

Case: Corsair Crystal Series 570X
https://www.amazon.com/gp/product/B0...KIKX0DER&psc=1



Total cost ~$4800

I don't really know enough about hardware to give you meaningful feedback. The specs look great. Things like a graphic card/storage options are a judgement call though (depends if you are going to use that computer for gaming for example, if not then you definitely don't need a 800$ graphic card. Similarly with SSD, you may not need 1TB, depends what kind of software you are planning to run.

I think it's a good idea to ask on some hardware related forums if the above RAM + mother board + CPU go well together. My knowledge is very limited here.

Btw, I am thinking about buying TR and 128GB RAM setup as well so I am definitely interested in how things go

Ill let you know how it goes a week from now

The ranges (oop,ip) you define when inputting the flop are pre-flop ranges or ranges at conclusion of preflop action, meaning ranges going into flop. Similiarly, if i define the board to turn and do the ev calculations, are the ranges the starting ranges for turn action or still preflop? sorry, about my confusion.

yes, the ranges you need to input are the starting ranges at the node you want to solve. if you were to set up a preflop sim for instance the starting ranges are simply all hands.

Hello,

I would like to run some scripts and get the results already in rounded strategies. Is that possible?

Will the upcoming patch fix the timeout bug in pioViewer? Right now if you set the turn/river accuracy too low, it can take forever to recalc strategy on small/very small trees (i.e. with forgotten streets) on specific cards. Would be nice to have this timeout number of seconds available in GUI somewhere.

Make sure to share Pio benchmark time you get on this.
(https://www.youtube.com/watch?v=5Ozjfnatwms (question number 2))

Yes, please take a look at this post:
http://forumserver.twoplustwo.com/sh...postcount=2714

This is not a bug and it's not in PioViewer.
This feature is on the to-do for the engine itself though as I understand the way it's implemented now is pretty annoying (there is a timeout but it's based on number of iterations and it's set way too high).

Is it possible to include the number of iterations in GUI though? Even as a temporary fix, that would be nice for those who need it (those who don't need it shouldn't change it at all, you might want to consider emphasising that in GUI in case you end up adding that variable there).

Unfortunately our current system sucks for shipping small updates in the engine. This is also one of the reasons our releases are delayed. Fortunately we fixed that and things should go more smoothly from there. Changing our previous one to a new one for our main product will have to wait for a while though.
Adding that to GUI would be easy, unfortunately the engine itself doesn't provide this functionality at this point.

Are you running into it when using the preflop or the postflop solver btw? I was hoping the implementation is good enough for the postflop solver (which is more precise) and for the preflop one it's better to take the ranges and solve subtrees from scratch anyway. If you are running into that problem with the postflop solver I would appreciated sending me the tree config so I can run some tests.

Before purchasing, does piosolver use any internet connectivity while running? for example, checking licenses etc. that would prevent usage when internet not available?

It gives me the following error message when the script runs: ERROR: round_up_to incorrect or missing argument

Yes it checks the license every few days and gives you a grace period if you are not connected (for a total of 14days). This is implemented so we can recover the license in cases like lost/formatted/stolen computers. This happens very often and if there wasn't a check we would be unable to reset it remotely.

It's hard to comment on "incorrect argument" error without seeing the argument first. We need either the whole script or at least the part responsible for inserting round_up_command.

do i need to remove free version before installing purchased version? thank you

Hi,

Can you recommend anyone who can offer skype or screen share coaching at an hourly rate on how to use Pio and/or run HH's for me (not looking for any coaching beyond using the software and how to get the most out of it)?

There is no need to do that. On the other hand you don't need the free version either. You can copy your ranges (from the Ranges folder) if you've already created some.

It's a good question to ask on our Skype group. I could give you some names (after verifying they are still looking for students) of people I talked to enough to be sure they know what they are doing but maybe it's a better idea to collect offers first

There is an output of 3b utg vs BB online. Where can i see the ranges for that scenario so I can replicate the results on my piosolver opt?

It was a long time ago and configs shipped with Pio changed several times but 100bb->3bpot-full in default configs should have very similar (I am not sure if the same ones) ranges.