"Even then though it might be hard to guess the reason, especially if the differences are miniscule."
For my own better understanding, if for example, the EVs of suited hand is
9.1800 9.1899 9.1785 9.17, is it meaningful?
What level of precision will show the blocker affect vs noise from the calculation. Can you help quantify the level of EV difference between seeing the blocker affect and just seeing the model inaccuracies? Thank you.

Refer to this post:
http://forumserver.twoplustwo.com/sh...postcount=2714

if you are using script generation tool you will need to choose small saves as well.

It's probably not meaningful but it's a judgement call. Depends what stacks/blinds are as well.

I've never attempted to quantify what is and what isn't "meaningful". It sounds like a bit of a research project. I don't think I am better equipped to do that than you or any other of our users

hi,

I study with a group of players.

When we study a spot we use Pio and in order to avoid having to wait for our PC solving each spot, our plan is to list the type of spots we will probably need to study. And get them solved by a fast server, in order to use them later.

Now my question :
Given that what we need is to make a rough estimate of what a reasonable play is (especially on flops), and given that we'd like to get it done in a reasonable amount of time, after several tests, I think that using 5% of the pot as the expected accuracy is fine. Am I wrong?

I have tested with several value for accuracy% and found that the results may only vary from a few % (for example, CB% in a given spot might only change from 50% to 55% if the accuracy is 5% or 0.5%), but globally, even with 5% accuracy, we could get the idea of what the strategy should be with or range (and solving is much faster than with 0.5% accuracy) : do you think i am wrong?

also, is there a rule of thumbs to evaluate more or less how much longer it is to solve a spot with 0.5% accuracy vs 5% accuracy? hard to answer I guess, but who knows...

PS : if that matters, the spots we study are simple ones : shortstack spots / end of MTT spots (5 to 25 bb deep).

I read everywhere that the accuracy shopuld be between 0.3 and 0.5%...
I am currently doing tests to compare results with 0.5% accuracy vs 5% accuracy, and so far, the results show a small difference, of course, but not a big one.

so a question comes to my mind : WHO needs to get 0.3% accuracy?
I mean for what kind of use is it so necessary to get such an accuracy?

As a human being, I think I'd play much better than I do now if i could play as Pio does with 5% accuracy. Once again, am I missing something?

I think you mean 0.5%, 5% is quite far off still.

You need more detailed solutions if you analyze river spots deeply (thinking exactly which bluffing combo is better than another one due to small differences in card removal etc.). It's true that overall 0.5% solutions are very good already.

You can think about it this way: 0.5% solution in 5/10 game, postflop spot (starting pot 65$) is exploitable by perfectly mind-reading and adjusting villain for 32c/hand. When you play 5$/10$ online you pay 3$/hand rake (once it's postflop). This means that the 0.5% solution is perfect for all practical purposes although it's not perfect if you are after purity and subtleties

No, I meant 5%.

As I said earlier : I am currently checking Pio's results with 5% accuracy and as a human being, I think my decisions are much much worse than that (and I am a winning online player), so if I could play as Pio does with 5% accuracy, I would be very very good... am i wrong?

You are likely not wrong as long as you adjust obvious blunders which haven't converged to 0 yet. The way Pio works is it's still going to call some obvious folds (or fold some obvious calls) some of the time at 5% accuracy. As long a you look at EVs and avoid those blunders you should indeed be fine. I am reluctant to recommend 5% accuracy though, it shouldn't cost that much more time to go to 1% or something

(I'm a total beginner when it comes to PIO).

So PIO creates a mixed and very complex strategy.. Is it possible to set up PIO in a way that it creates a simplified strategy?
Random example: Instead of having it say, "bet 9h8s 33%, 8h7s 33% and 9h7s 33%", can it choose certain combos to bet 100% of the time and other combos to check 100% of the time?
In other words, I want to use PIO without having it show me a mixed strategy.
I want it to show me "bet 100%" or "check 0%".

It's not possible to solve with those assumptions. You can however round the strategies after solving the tree. To do that:

1)Solve the tree and stop the solver
2)(maybe make a save here to be able to go back)
3)Go to Tree->Round Strategies and choose which streets you want to round and what rounding you are happy with

Rounded solutoins will be nowhere as good as mixed ones. That's the reality of the game that unexploitable strategies are very mixed at many decision points.

To beat a dead horse ... what % of pot accuracy needed to distinguish blocker impact on river betting?

Hey, I was wondering if this was already implemented. I just noticed that I can't node lock any of my saved script trees bc I chose for the small tree option.

It's implemented: you can make a small save keeping the whole tree in memory. Of course when you close the solver it's no longer in memory so you won't be able to node-lock on it once you load the save. That won't change as it's a hard limitation of small saves caused by:

It shouldn't be a problem though. The tree requires recalculation after node-locking anyway so you can make a new tree, lock there and then solve.

We are pleased to release our first PLO product PLOCalc which is equity calculator/range explorer for PLO which calculates range vs range equity (which includes equity for all combos in range) without approximations on preflop/flop/turn and river.

There are many other features of it as well, more here:
https://www.piosolver.com/blogs/news...d-tool-plocalc

A few notes:
-Omaha is much bigger game than Holdem, hardware requirements and calculation speed are going to be affected although PLOCalc itself should run on any relatively modern computer
-the program requires a big (1GB) file to calculate preflop equities
-it requires about 2GB of RAM to work
-PLO tools will be more expensive than Holdem ones as the market for them is significantly smaller
-There will be PLO solver in the near future although we think PLOCalc might actually be more useful for many practical things as the full GTO solutions will be difficult to make sense of (as PLO as much bigger game than Holdem)
-This is a very early release, if you want a free version to test it first or are unwilling to deal with unavoidable problems in the early stages of the life of the product we recommend waiting a few weeks before making your purchase. That being said the product is extensively tested by us and our testers (some of them world class PLO players)

We believe PLOCalc is the most advanced tool for Omaha range and equity analysis available. While there is going to be unavoidable learning curve it allows you to see things which were not possible to visualize before.

it would be nice if there was some warning after clicking "build tree" after locking strategies. it is pretty frustrating to create a strat and lose it all because i clicked "build tree" instead of "go"

There is a warning like that. Most likely you have disabled it though at some point in the past. To get it back:

1)Open pioviewer.settings file in some text editor
2)Search for (ctrl + f) IgnoreWarnings in that file, the relevant part will look like this:

Remove:
BuildTree=true

line, save, close, restart the viewer.
To test it build some simple tree and try to build another one, the viewer will warn you that you are about to lose your previous tree.

Getting around 2seconds on pio bench!

Wow, Thread Ripper is impressive. Is it stock or overclocked?

Hi,

I am trying to configure a custom PC build for the new AMD TR 1950x and am unsure on how much the speed of the RAM will affect performance in PIO. I have a few choices ranging from 2666MHz up to 3600MHz.

Also:
I've tried googling the dual box setups but cannot find anything of use, how expensive would the setup need to be to attain a ~1.5s benchmark in PIO? And can you give me any links to these dual box setups?

Thank you!

Small suggestion: for show total strategy as real frequency, add check box option to display the other frequency too, maybe smaller font underneath/ on the side of the main one being used. This would be convenient instead of having to go to config every time you want to compare.

It shouldn't matter much although those things are very difficult to test and be 100% sure.

Those are setups with 2 CPUs on one motherboard.

A poster above mentioned they've got a 2 seconds benchmark with Thread Ripper. To beat it you would likely need 2 very fast Xeons, for example two 16-core ones (for a total of 32 physical cores). This is going to be prohibitively expensive. Alternatively you can try going for new Epyc CPUs (from AMD), again you will need a lot of cores to beat 2 seconds time.

http://www.amd.com/en/products/epyc-7000-series
http://ark.intel.com/products/81060/...Cache-2_30-GHz

are two examples. I recommend against getting those. Something like Thread Ripper offers great value and very good single core performance. You can beat that but you will have to spend tons of money and you will get a machine which is going to be worse for everyday things (as single core performance is going to be worse) and only marginally faster for Pio.

Noted. As a quick work around you may try opening the viewer two times and viewing the same tree with different settings.

Suggestion: add scroll ability in add/remove lines boxes in case you have many line changes.

While the scroll doesn't appear you can still scroll the text using down/up arrows or home/end keys.

What is the meaning of the subsets titles? 72_THREE_EQ_2__469

There isn't anything meaningful you can take from the names. We included 2 subsets for every size so there is more room for experiments. Recently we've run more comprehensive benchmarks on those subsets and now we are recommending the best ones listed here:

https://pastebin.com/AtyU84qC