There are many more besides that, we ship them in preflop_subsets folder. There is nothing special about 184 flop one.

One thing you can do is to use the preflop ranges and solve n flops in a script (generate script button) then run multifile aggregation report on the results to have all the results in one .csv file. See here FAQs here:
https://www.piosolver.com/pages/faq-videos

It's the 6th question in the first video and 5th in the second one.

Hi. I was wondering what the difference is in PLOCalc between 'PLOCalc.exe' and 'PLOCalc-olderCPU.exe'. I couldn't find a setting to select between the two. I simply removed 'PLOCalc-olderCPU.exe' from the folder, and that prompted PioViewerOmaha to use the other version. I'm assuming I should be using the non-olderCPU version seeing as I'm using a Ryzen 5 CPU.

If PLOCalc.exe works on your computer it means you don't need olderCPU one. It wouldn't start at all on a system that needs the olderCPU version.

Hi - I know the hardware conversation has been had to death, but just wanted some advice before pulling the trigger on a new rig.

I know you guys don't recommend laptops but if this was a non-negotiable, would something like a i7-9750H (6 core, 12 thread, 2.6ghz base, 4.5ghz turbo) with 32gb RAM be sufficient to solve largish trees (100bb, wide blind vs blind or BTN vs BB) with 2 bet sizes and raise options across streets? Also do you have any experience with how much stress / CPU heat is involved letting a script with a bunch of flops run in a laptop form factor?

Many thanks.

ty punter11235 for your answer !
Regarding preflop action, the intuitive thought used to be that the deeper the stacks, the bigger the preflop sizings, especially regarding 3bets, 4bets, and so on.
But I recently heard Sauce123 saying in a stream that he had run sims with Pio and the 3bet sizings were the biggest with 100bb stacks. I have not understood if he meant the 3bet sizings remained the same when the stacks get bigger or if they even decrease... Please, can you give your thoughts on this?

Thank you

I saw some posts about people getting between 3 and 3.2 in benchmark for i9 9900k build.

I'm getting 5.2?

I am on high performance power settings and no anti virus or anything like that. Any ideas on why I'd be so much higher?

Do you know any benchmarks for the 9900k off the top of your head? Does 5.2 sound too high?

Yes, 32GB of RAM is plenty for postflop trees even with many bet sizes. If you already have a Pio license you can use "estimate tree" function to see how big trees are without actually building them.

That depends a lot on manufacturer. In general slicker/thinner laptops are much worse at cooling and gaming/bigger laptops are better at it. I think it's a very good idea to buy a cooling pad along with the laptop. It's likely not necessary when you are in front of the computer solving things but if you run long scripts it might be very helpful.
CPU won't overheat (at least things are really messed up) but it may start throttling bringing the performance down.

I haven't run enough simulation to have any informed view about unfortunately.
I remember Pio wants bigger 3bet sizes in 100bb than people used to employ but I haven't run many deep stack trees to see if there is any difference.

5.2 sounds too high unless it's not an edge version.
3 sounds too good but maybe it can get it on some setups. I don't remember any benchmark times for this CPU unfortunately.

Can Piosolver allow to calculate :
- optimal preflop Open raising range for every position of 6max NLHE
- optimal preflop 3betting range for every position vs each OPR range of 6max NLHE
?

If Pio can do that, and given that my computer can't run Piosolver Preflop solver, do you know where I can by these ranges? Maybe in the Piosolver discord?

thank you

Hey I recently purchased pio and have come across a few issues. I’m a huge fish when it comes to how computers/software etc works.

I recently bought a gaming laptop for the sole purpose of running pio. It was suggested by someone who also uses pio. Here are th issues

1. When I click estimate tree size it says I have 14-20k physical memory available out of 32k total. When I first got it I had the full 32mb available. How am I able to free this up? I don’t have any other applications open.

2. When I ran a sim today that estimated 10kmb and had 18k available it had issues as well. Over the course of an hour and half running it went from being exploitable for 18%-8%-%13-6%-11%-3% and never got past that before I stopped it. This never happened when I ran sims earlier this week that worked fine. Is this something I should worry about?

For what it’s worth I’m running a sim on bu vs bb sep 50bb effective using upswings tournament ranges.

Thank you for any help I can get.

Just tried to run the same sim at 30bb stack sizes and didn’t have any issues at all.

Hi !

I'm new user. I'm trying to use free licence but I don't know how to connect to the server... Can you tell me how can I do that? : )

How much is the difference edge vs pro?

No. It's a solver for spots which are already HU.

For the same reason as above it can't.
You can see in task manager what takes the memory (open it and sort processes by memory). A good first step is performing a full system restart (not sleep/unsleep cycle).

It can happen on some trees. If you still have a config please send it to us and I will run some tests. You are solving a very big trees (10GB+ in Pio is humongous as we use very little memory in comparison to popular algorithms).

The free version download is a .zip file. You need to unpack everything to one folder (for example C:\PioSOLVER-free) and run it from there. Make sure all the files are there and it will connect automatically.
If it doesn't, double click on PioSOLVER-free and see what happens. One possible reason for it not working is using legacy 32bit Windows version.

Depends on the system, edge has a bit faster compiles due to optimizations for modern CPUs. Probably around 15%-20% on your CPU.

I've just browsed a few videos on Pio - looks very interesting.

Let me see if my conclusions are right:

Bet sizing is an input, not an output. I.e you tell Pio bet amounts relative to pot on all streets and Pio returns the optimal distribution of checking, betting, raising, etc.

Analysis can just be done for 2 players seeing a flop.

The results assume opponent is playing optimally with the hand range you give him.

Yes.

Yes, although we also have a preflop solver in the edge version. That requires a very serious hardware and a lot of time to run though.

No. The resulting strategies are optimal (+/- accuracy). No matter what the opponent does they won't increase their EV over the solution and they might lose significantly if they deviate.

How much has the software evolved? Is watching hand analysis from 2015 for example still worth it, or would the later versions give different results?

The results are still the same. The solver is faster and more precise, it uses way less RAM.
There are many more features to help with analysis and more features to build more complicated trees (the first version was really simplistic in this regard). The core results though are still the same.

Hello

I am trying to install PIO on a second PC. I have a edge subscription however I cannot find the key as I upgraded from Pro version.
I managed to install the Pro version using the key I got for that, however I cannot connect to solver as it says access denied when I try.

Could you tell me how I could find my edge key?
I have checked all my e mails and I have the confirmation of upgrade but not any other details. Or how I can get the pro version to work as I don't need preflop solving on the second pc.

Thanks in advance

Send us an email including:
1)the email address where you got the confirmation of the upgrade
2)your name used for the purchase

We don't collect any personal data but we have access to the store history which we can browse by name. If I am not able to find your edge license using the email address I can look for the transaction using your name.

Hi,

I'm wondering about the weightings in the flop subsets and how it relates to the probability of a given flop occuring.

Is the following correct?

1. The probability of a given flop occuring within a subset, is the flop's weight divided by the sum of all flops weights.

E.g. in the 25 flop subset, the sum of all weights is 37.67. The weight of flop Ks9s2d is 1.86. So the probability of Ks9s2d occuring is 1.86 / 37.67 = 0.0494.

In other words, 4.94% of the subset is represented by the board Ks9s2d?

2. The relative probability of a given flop occuring compared to all possible flops, is that flop's weight divided by the average weight of all flops.

E.g. in the 25 flop subset, the sum of all weights is 37.67. The amount of flops is 25. So the average weight is 37.67 / 25 = 1.5068.

The flop Ks9s2d has weight of 1.86. So the relative probability of Ks9s2d occuring compared to all flop is 1.86 / 1.5068 = 0.2344. It's 23.44% more likely than the average flop.

(Disregard the exact numbers of course, more concerned about the method).

Thanks!

Yes.

I am not sure if that what "relative probability" means but your calculation seem correct for what you want (knowing how more frequent a given flop is than an average flop).
This number can be misleading though. In a toy example with 3 flops with weights: 100, 10, 1 "the average flop" has a weight of 37 so now the first one is almost 3x as likely as average one and the last one 37x less likely than "average one". The problem is it doesn't really tell you anything useful.

Thanks for the quick reply.

I'm not sure "relative probability" is the correct term but just tried to find something to explain what I was looking for, which is exactly what you said (knowing how more frequent a given flop is than an average flop).

My intent was to use the subsets as some kind of baseline for how probable certain types of boards are. But I'm not sure it works. Looking in my database over a few million hands, 9-high and lower two-tone boards are the most common. (At least within the constraits of how HM2 groups textures). In the 184 flop subset, for example, this doesn't seem to be reflected. The flops that are more frequent there don't include many 9-high (or lower) two-tone.

Is there any correlation between the weight of certain types of boards in the subsets, and the probability of those boards appearing in actual play?

The subsets are constructed with sole goal in mind: approximate overall equity/EV well. They may not be a good representative for flop frequency. The reason our subsets are so good at the task in comparison to subsets created in the past by others (Will Tipton for example has done some work on it) is focus on the goal. We didn't care if 9xx is as probable in the subset as in all 22100 (1755 strategically distinct ones) flops or if KQx occurs as often so KQ can flop 2 pair or if flopping a flush draw probability is the same as in the real game. Approaches based on those notions proved to be inferior when it comes to approximating EV/EQ in the whole game.

There probably is correlation but it wasn't the goal so it's likely way off.

Should the displayed strategy update in a node locked tree, or will the EVs be the only thing that changes?

For example, here's a node locked tree:



If you consider the 6s4s combo in the node locked solution, b275 is much higher so we'd purely take this action under the node locked solution. But the color coding of the strategy that's actually displayed is the original strategy (which is to mostly check this combo at equilibrium).

I guess a simpler way to ask this is: after we node lock, is it possible to have the color-codings update in addition to the EVs? That would make a lot of sense when you consider the purpose of node locking.

Also, a question about the ICM implementation:

If I want to study the impact of several different ICM configurations on a tree, do I need to rerun the tree every time I change the ICM configuration? This doesn't work like node locking where things just get updated on the fly, right? I'm assuming I actually have to rerun the whole tree each time.