Any VM that runs Windows or Bootcamp (but that means booting into Windows).

We don't support that and I have no idea how that works but I am going to answer your questions assuming you would run it in Pio.

This is taken into account by the solver. For example if you have two flops:
-As7h2d:1.0
-As6h2d:1.0

the starting pot is 100$ the ranges are 77 vs AK with simplistic assumption that 77 always wins on As7h2d and always loses on As6h2d then there are:

3 * 12 = 36 matchups on the first flop (3 combinations of 77 and 12 combinations of AK)
6 * 12 = 72 matchups on the second flop (6 combinations of 77 and 12 combinations of AK)

The total EV for 77 is then:

(36 * 100$ + 72 * 0$) / (36 + 72) = 33.333$

If you introduce other weights than 1 there, the math changes but the principle remains the same: flops which are less frequent when you hold a particular combo are counted with smaller weight when calculating overall EV.

It's already done for you by the solver. It takes that into account when solving flops, when doing aggregation reports as well as when solving preflop trees. Notice that it wouldn't be possible to adjust the subset itself as probability of getting various flops is different for every single matchup (so for example 22 gets a differently weighted subset than QQ). Again - it's all taken care of by the solver's internals.
If you run aggregation reports you will see some details about relative probabilities in them.

One interesting point is that it's a bit complicated when it comes to preflop solving as our tests show it's better to assume an all-in preflop is always on 1755 flops even if the postflop solving is only done on a subset. This means that there is different card removal effect when all-in preflop and when going for postflop. It introduces some challenges when accounting for card removal and it's the reason the preflop solver is a bit more complicated to correctly implement that it looks at the first glance but again - it's worth taking care of because the results are better (by that I mean that you need less flops to approximate real solutions than you would need if you assumed all-in on a subset of flops as well).

Thanks, punter11235, I see that it's more complicated than I thought to use the flop set in a meaningful way outside of PIO itself. So I've decided to go for PIO.

Being Mac user, I am thinking about setting it up and running it in the Amazon cloud, but first I need to know how. Has anyone written up something somewhere about how to set PIO up and run it on an Amazon instance?

I will not be doing pre flop calculations at the moment, but I will do a lot of flop calculations. For example, calculating the postflop EV of each hand in a a 3B flatting range vs a 3B range, using the largest or second largest PIO flop set to approximate the full thing. The largest flop calculations I can see myself doing at the moment would be a big blind flatting range against a small blind open range, using a game tree with four bets/raises, 100bb stacks, and single bet sizing.

If anyone can give me some advice on how to set up and run PIO on Amazon, and what type of instance would be most suitable for me, I would be most grateful.

hi punter,

It will be extremely helpful for an easier/faster learning process (at least for me) if we will have a "strategy" checkbox in the range explorer window! This way we'll be able to see in a click what the solution is for different parts of the range like what "nothing" or different types of draws to bet.

Sorry if this is too basic, but I know nothing about PIO functionality. :-)

I would like to compute the post flop EV for every hand in a range X when it plays an opponent's range Y in some game tree (say, 3B pot with a flatting range vs a 3B range using a game tree with three bets/raises and geometrical bet sizing).

The brute force approach is of course to run a simulation on every possible flop and take the average, which is not feasible. The PIO folks have made reduced flop sets to do this calculation fast and get the EV as (as I understand it) a weighted average.

What I would like to know is if this is a menu option in PIO Basic version. As in, can I set up two ranges, a game tree, and get the total post flop EV for every hand in each range by an automated calculation that simulates over all flops in a predefined weighted set of choice (say, PIO 95)?

Just to be clear, what I was asking in the post above is if option c) in the last quote is available as an automated option in PIO Basic (or in the higher versions).

All you need is to install 64-bit Windows and connect to it via some kind of Remote Desktop interface. I am not sure how that works from a Mac though (that is connecting to Windows). There are many cloud/dedicated solutions and we are unable to support that. You don't need anything Pio specific though, once you have Windows and can use it installing Pio is the same as on your personal computer.
My suggestion is to look into cheap dedicated servers instead of AWS though. You can probably get them for like 40$-50$/month (for something like hexa core modern CPU).

This is already possible. In the gif below I am showing strategies for set, overpair, top pair by selecting categories using checkboxes and then for overcards with a backdoor flush (by manually selecting those hands):

https://giphy.com/gifs/l0HlQ813WJ4DUv2Ao

It's not a simple average as you need to account for card removal (as in my post above).

Yes - those subsets give results very close to all flops. It is in fact feasible to run those simulations on all 1755 flops which is the way we got the results for all flops to compare with result on subsets (some of our users helped us a lot with running the simuls on their hardware).

If you want to do that automatically you need scripts which are supported only in pro/edge versions. This is the only functional difference between pro and basic version (two others are number of activations and support for more threads).

Follow up (I am still a little confused about the weights): To calculate the post flop EV of a hand in a range that I play vs another range using a PIO flop set, I would need to make a script and run a simulation over all those flops. What I still don't understand is how to calculate the hand's total post flop EV from the results of all the individual flop simulations.

We have an aggregation report feature which does it for you. It calculates EVs, EQs, preflop orders and several other things using multiple solved trees with the same structure.
I talk about it here:

https://www.youtube.com/watch?v=5Ozjfnatwms (question number 5 - see the video description).

Thanks, I have purchased the pro version just now and will get to work. :-)

Is it normal that I dont have option for "include oop bet" ?
https://gyazo.com/d2a34b9e3fbdb721fe425046cc2a448c

nvm got it

Yes, the interface was changed in 1.9 to be a bit more powerful.
Some of the things not 100% compatible with the old way will be fixed in 1.9.2 (for example "use one bet if there was a raise before" option mentioned by one of the users in this thread)

yes, i know this, what I ment was to be able to see both strategies ( check/bet) at once like It is displayed on browse tab. Its not very helpfull If i donț have a visual representation of both in the same box with two different colors that toghether fills the box to 100%

Hi, if I write JT:0.6,JTs:0.3 in select range window JT and JTs are both selected with 0.6.

Yeah, this is a good suggestion.

This is correct. "JT" represents all JT combos both suited and offsuit.
"JTs:0.3, JTo:0.6" is what you need.

and one email question from today:

This is not a bug. The solver always reports cumulative amounts (that means total invested so far) while the viewer translates it to what people expect (that is amount of bet not counting previous ones).
This is similar with EV: the solver always reports absolute EV internally so for example:

bet 100, call, bet 300, fold has EV of -100 for IP player while the viewer translates it to what people are used to (that is fold has EV of 0$ no matter how much was invested so far). How EV is translated is controlled by Tools->Configuration->Data presentation - other. The formulas there can be used to display EV as % of the pot for example.

The reason for both is that it's easier for programmers (parsing/interpreting the lines/results) when you have cumulative bet amounts as well as absolute EVs (you don't need to implement logic to remember how much was invested so far to know what was won/lost at the end of the hand). On the other hand the viewer shows what is easier for players and people used to reading hand histories.

Thanks

1. Does Piosolver use blockingbets? A very reduced riverbet for example?
2. If you want to see the most preferred bettingsize on a street (you only want the most prefered). Is it accurate to apply many bettingsizes and see which it prefers. Or do you have to try and run it with only one betsize at the time. (does the first option "destoy/change" the real answer or it will basically be the same?

I am not sure what you mean by "uses". You can add them to the tree and see if they are part of optimal solution. If you mean to ask if they are in fact part of optimal solutions in practical cases then unfortunately I don't know as I've never tested that

You can run it with many bet sizes, choose the most frequent one and then hope that process results in the best bet size (assuming you can pick one) but this doesn't guarantee to be correct in all cases. The only way to be sure is to run the same tree many times with different bet sizes. This is not specific to Pio, this is just how math of it is.

I am running script jobs on the Pio flop sets and have a couple of questions:

1) Do I need to do full saves to run aggregation reports on the output files after? I'm currently only working with HU situations where I want to know a) EV and R for all the OOP player's hands after he checks and b) EV and R for all the IP player's hands after OOP's check
2) Is there an option for setting rake for script jobs to get aggregation report results with rake effect included?

Definitely not, in fact it will take much longer to run a report on full saves (because it takes a lot of time to load them). That is unless you want an aggregation report for a river spot but that's a very rare usecase (and still doable on small saves although it will take a very long time).

Those are flop spots so even micro saves would do but I recommend making flop+turn just in case you want some turn analysis faster.

Rake will be included in the script if you include it in the tree config before clicking "generate script".

Sorry if this is in FAQ, but I just fire away my beginner's questions. :-)

1) Is it possible to run aggregation reports for a calculated flop set and test results for various rake structures by loading one of the trees in the set, setting a new rake structure, and rerunning the aggregation report for all the files?

2) When flatting 3-bets with low pairs I am interested in knowing how much of the pair's EV comes from flopping sets (as in, can they profitably flat 3-bets if they c/f all flops where they don't spike). What would be the easiest way to do this?

It isn't. The reason for this is that you don't have solutions for a different rake. It would theoretically be possible to use the solution you have with a different rake but that only works on full saves (you need the whole tree to recalculate EVs with different rake) but making full saves is in general a bad idea (slow to load, take a lot of space etc.)

This is not supported at the moment.
Multiple file aggregation reports give you EVs for a range on every flop and EVs for every combo on all flops but they don't produce EVs of every combo on every flop.
You could theoretically work around it by running a separate report on every flop and then calculate a weighted averages from flops with sets and without it but that's a lot of work.
Maybe better approach is to just look at EVs for 4-5 flops with sets and then for 4-5 without it and approximate from that.

We may add this feature in the future if there is interest and some more useful things that could be done with it.

hey punter, kinda offtopic but can you recomend a subset with a minimum number of flops for a meaningful study of postflop play? Or some guidelines on how to choose them?

We get this request a lot but sadly we can't offer anything here. The task of finding the best subset to approximate EV (or preflop solving) is a different one than finding the best subset to study. I think that's something for coaches or experience players to tackle.

New version of pio
https://gyazo.com/ed9cd3ff338f20f4d1c2a0400c82011f
Older version of pio:
https://scr.hu/40b4/7lxb3

same range 2 diferent results, where is my mistake?

I don't know. You haven't shared what you did, the results nor even configs (screenshot are not very useful as they don't contain all the information). Here is how to report those things:

1)use the "copy to clipboard" button to get the config and send it to us
2)make sure you are solving to the same accuracy
3)describe where exactly the results differ and what versions you are comparing

Preferably send all the information by email. Point 3) is very important for us to be able to verify if there is a problem.
The most likely reason if the solutions are different is that you are using different configs between version and/or are solving to different accuracy.