ranges are not same

Suggestion: A "Set rake structure" option in the tree builder tab. Makes it faster to set rake, and it's good for the user to see the rake used for a loaded tree (at least I can not see any rake info anywhere).

Well spotted. Again, the way to share configs is using "copy to clipboard" button not making screenshot as those are very cryptic.
The way to copy a config from one computer to another or from one instance of Pio to another is to use "copy to clipboard" button and then Tools->paste tree building config from clipboard.
Alternatively you can save the config and share the file itself (they are in TreeBuilding folder).

It's there, look:
http://imgur.com/a/tCkUq

Is it possible to set different betsizes for the same street, but on different branches of the tree? For example I want to have the betsizes A and B for the turn after the flop has been bet, but betsizes B and C for the turn after the flop was checked through.

Yes. There are two ways to do it:

1)include all three (A, B, C) sizes in both branches and then remove bet size C in the first branch and bet size A in the second branch (that's probably a simple option)

2)include only bet size B in both branches and then add bet size A in the first one and C in the second one.

I talk about how to use add/remove line interface here:
https://www.youtube.com/watch?v=zLGpcZavxeQ (first point - see the video description)

Make sure to confirm the tree is what you want after building it (but before spending time to solve it).

Ha, sorry, that's what happens when not seeing the forest for all the trees. :-)

A question about script jobs: Can you (within hardware limitations) launch as many script jobs at the same time as you like? The way I understand it is that once a script job launches, one instance of the solver runs (console window pops up) and we are free to go back to the script interface and set up and run another script job that will launch another solver instance and so on?

You can do it but it's not recommended to run several scripts at the same time because:
-you need more RAM
-you are unlikely to see any performance gains (maybe with 2 instead of 1 but even that isn't very likely)

It's better to combine scipts into one instead. There are many ways to do that, one is described in FAQ:

and then run that "master script".

I have been playing around with heads up games, and I ran simulations using continuation bet 100% IP and then I have done it where there are no constraints. I was assuming that with no constraints, the IP has to yield more gaining the option of being able to check back and split the ranges as it pleases to be more lucrative. However, when I do so, the continuation bet all yields the same mes EV IP as it does when it can also check. Am I overlooking something?

Depends on the spot but in fact ability to check behind is usually worth something. Maybe share the config by clicking "copy to clipboard" button in the tree building tab so I can take a look.

Newbie question about the preflop solver (I just bought Pro, considering Edge):

I'm interested in calculating responses to a 3-bet (100bb starting stacks), but without calculating the optimal 3-bet range in the process. Is it possible to set up the solver with my open-range, a predefined 3-bet range from Villain, and then solve for optimal play at the point after the 3-bet has happened?

The solver would then produce my ranges for 4B/call/fold, Villain's response (jam/call/fold) to my 4B, and my call/fold ranges vs his jam. But how to set that up?

"The "show_all_freqs local pp" command doesn't seem to return any frequencies on turn and river for 3bet pots..

I'm also confused about the structure in which the information is presented, since some of the numbers that describe the line of the tree don't seem to correlate with neither the bet/raise sizing or the current pot size.

Yes, this is possible and will be faster (and require less memory) than solving the whole thing.
Here is a sample config like that:
http://pastebin.com/esrAdkRa

(to load this config copy it from the link above and then use Tools->paste tree building config from clipboard)

You may verify that's the tree you want by clicking "Build pure preflop tree" button and browsing the preflop structure. Notice that this is quite small tree, even using 62 flop subset it's still needs less than 10GB of RAM.

Yes. Be careful about interpreting EVs though. As starting ranges are different than 100% and there are no folds/calls for 2 first actions the total EVs are average across starting range (the one for open/3bet) not across the full range (as many hands are not included in the calculations at all). I hope that makes sense

You are probably running it on a very small save (flop only or preflop + flop). It would produce (preflop) + flop + turn on small saves (which include turns) and all the streets on the full saves.
The reasons it doesn't re-calculate the whole tree from a small save to show you frequencies across all streets are:

1)the tree is potentially humongous but you might be browsing it on a laptop (for example if you bought PioCloud packs, some trees there are 200GB or bigger but you can browse them on a 2GB laptop)
2)it would take a very long time in some cases

It is possible to recover the whole tree from a small save by recalculating everything but it's not faster in practice than just solving from scratch.

(Just in case others don't know what this question is about, it's about a new feature in 1.9 which allows printing all line frequencies across the whole tree in one file)

The solver always reports cumulative sizings, so for example a line:
r:0:b94:c:3d:b296

means bet 94 on the flop and then increase the total invested to 296 (that is bet 296 - 94 = 202) on the turn.

The viewer translates solver's output to what people are used to. In this case it would be bet 94 on the flop and bet 202 on the turn. In the future it's likely that the viewer will have ability to display those frequency reports in more digestible form but for now they are just internal solver commands so the output might be a bit crude. I just wanted to add it for advanced users as this was an often requested feature.

We have released another batch of fixes/small features - PioSOLVER 1.9.2.
Highlights:

-Free version is updated to contain most 1.9 features and ability to solve one flop (Qs Jh 2h).
-Preflop subsets we use for a preflop solver are now available for download.
-Overall results are back to aggregation reports - it was an oversight and it's now fixed.

More, here:
http://piosolver.myshopify.com/blogs...iosolver-1-9-2

This is now fixed

Done!

Damn, did decide too early to pay 250$ for the basic version. Grrrrr!

There isn't anything to regret because:

1)if you like it and want a higher version you can upgrade for the difference +24$ (or if the transaction is from less than 2 weeks ago just drop us an email and I will make an offer to pay exactly the difference)

2)if you don't like it and the transaction is from less than 2 weeks ago you can request a refund

Hello Hello

How EV Rescalated works, I clicked check it showed me that 42 has 5.1 EV BUT it doesn't win never ever as a check, literally EVERYTHING beats it. How come?

http://prnt.sc/ccz3zw

Edit: OFC I am talking about Range Explorer

Pre flop server newb question: How to decide whether OOP or IP acts first? It's a little unclear to me.

Running a pre flop sim for OOP (2.5x) calling a 3B (8x) by IP (i.e. both blinds are out and leaving behind 15 in dead money). When running the sim I observe that setting accuracy threshold to 1% of pot gave an accuracy of 0.15 bb/hand. That must mean that the pot is counted as only the pre flop dead money, not the pot size when the flop is reached ($80 + $80 + $15 = $175).

Is this (i.e. pot size used for accuracy threshold calculation purposes = sum of dead money) always the case for pre flop simulations, no matter what the pot size is when a flop is reached?

make it to support more core, 12 cores is not enough, i7 cpu is for gaming not for scientific calculation

If I remember correctly it shows EV for chosen subrange. Some hands still get EV from betting even if you chose the checking range (because some of them are mixed). It works like this:

choose subrange -> check what EV is for those hands -> display it

(notice that 2nd step doesn't know how the subrange was chosen, maybe you clicked it yourself or maybe you have chosen the checking hands, either way it still takes EV from all actions in the 2nd step)
If you would like to see EV for an action just go one node further in the tree (in this example check) and check overall EV there.

It depends on "OOP Player money in the pot" and "IP Player money in the pot" fields. If one has more invested (posted a bigger blind) then the other starts the action. If it's equal then OOP starts the action.

As usual, the best way is to play around with various configs and use "Built pure preflop tree (no flop exits)" button. This way you can very quickly build various configs and verify they are what you want (by going to the Browser tab and browsing the tree).

It's unclear what the starting pot should be for the purpose of those calculation in the preflop trees. That's why it's recommended to use bb/100 instead. One way or another the accuracy is always given in chips so if you convert it to bb/100 or % of the pot matters little.

If I remember correctly: yes but again - just don't use it.

Edge version supports up to 128 cores.

One from an email:

Yes but those trees are humongous. You can set the postflop play using c postfix for bet sizes, for example "100c" means "bet 100" or "raise 100 more". Here is the full limit holdem preflop tree with full (cap of 4) postflop play:

http://pastebin.com/uGSVh1D4

To load this config into your Pio:

1)copy the text from the link above
2)Tools->Paste tree building config from clipboard

Notice that this tree is really humongous, it has 7 preflop exits and it would be ~200GB big on 5 flop subset.
You would need to trim it a lot to make it practical on even very big machines. Some suggestions:

1)get rid of limp preflop (that cuts more than half the tree)
2)get rid of 4 betting preflop
3)reduce cap on the flop to 3
4)remove donkbets on the turn and river (most controversial, not likely to cost much EV)
5)still get a very big machine to run it

I came to the same conclusion after playing around with it, and landed on 10bb/100 for loose screening calculations and 5 bb/100 for tight reference calculations. Does this seem like reasonable criteria to you?

The most important factor is obviously flop set size. I ran some sims yesterday on 30 and 60 flops using loose and tight criteria, and found that going from loose --> tight changed very little for frequencies and EVs, but doubling the flop set gave significant (as in, non-negligible) changes. A loose calculation on 60 flops took the same time as a tight one on 30 flops, so the best bang for the buck if you invest X amount of time seems to be to run loose convergence on the biggest flop set you can fit.

How loose you can get away with is not yet clear to me, but it seems 10bb/100 is fine for approximating frequencies and EVs. To decide qualitatively whether a hand definitely should be in a range or definitely should not be, it seems more than good enough.

If you or anyone else here have some advice on how to best choose accuracy settings for a good balance between accuracy and computing time, I would be happy to get some advice.

While preflop play should be quite accurate at this point I would want more like 2-4bb/100 in SPR pots and 5-7 in 3bet pots (starting pot is bigger there is % wise it's similar).

Yes, 30 flop subset is still a bit small while 60 is already quite big. The question is also how do you measure changes. You may see significant strategy changes on specific combos even if you go from 60 to 90 but you are very unlikely to see significant changes in overall frequencies, EVs etc. For all practical purposes it matters very little of say 52s is 3betting 100% folding 0% or folding 100% and 3betting 0% if both solutions have the EV of both actions as very close. Probably you need to look at total frequencies and EV of specifc hands to see if solutions are close.

I feel that 60 is plenty and going higher just gives smoother and a bit more precise solution but it's not something you would benefit from in practical play. The best way is to run some simple trees and make a judgement yourself.

It's not clear to me either
The thing is that time is a scarce resource for us. When I have time to potentially run those tests I prefer to just work on the engine or on some new feature so I will probably never test those things myself in detailed way. A lot of people run a lot of trees though so it's probably the best to find other users to talk to (feel free to look for them on our Skype group).

The only thing I feel strongly about is that it's not worth it to input detailed postflop play (many bet sizes, turn donkbets etc.) as they have little to no influence on preflop ranges, at least from what I have seen.

Hi,my laptop seems to be heating a bit too much when running PIO,I heard it'd be good if I bought a cooling pad,would that really improve my problem?

If so,do you recommend any?

Thanks!

It's likely although it depends how bad the situation is (if your cooling doesn't function at all then it won't help much).

I haven't used one personally so I have no idea.

Another thing you can do to mitigate a problem a bit is to go to Tools->Configuration->Behavior and change Max number of threads to something smaller than amount of logical cores you have (so to 3 in case of quad i5 and to say 4 or 5 in case of an i7, 0 means OS default there so it's uses all of them)

Is it possible to use PioCloud packs with the free version of PioSOLVER?