You need to make a multifile aggregation report to see those EVs. It produces all kind information additionally: average strategies, order of hands (sorted by EV), frequencies of particular runouts etc. You can find it in the top menu: Analysis->Runouts aggregated frequencies analysis over multiple files

I also described it in one of FAQ videos, here:
https://www.youtube.com/watch?v=5Ozjfnatwms (question number 5)

Same question, but i want to run overnight different flops and ranges (tottaly different situations), havent find how to do it yet, is it possible? If so how?

It's possible. You will need to create separate scripts for the spots you want (generate script as before and save it). Then after you have script1.txt, script2.txt etc. you should crate a master script (just a .txt file) with the following content:

load_script script1.txt
load_script script2.txt
...
load_script scriptn.txt

then run it by going to Tools->run script

While this requires some work it's also quite flexible as you can decide to run various parts at different time or change/modify them if they are not yet run.

My version is 1.8.2.0(in the about section) when I try to update it just says "the downloaded version is not newer than the currently used one"?

Are you sure the path to the solver is correct? It often happens that people have it in many places (for w/e reason) and the updater is updating for example C:\PioSOLVER but you are starting it from C:\My Programs\Pio or some other location (maybe your desktop shortcut leads to that other location as well)

Maybe check this video: https://www.youtube.com/watch?v=1PEmsGaGLvk (point 1, see the video description) and see if it helps.

I would like to do the following pre flop calculation vs an open raise:

- Create a static calling range with some split frequencies in it (e.g. 55 calls 50% of the time)
- Lock the whole calling range and compute the 3B and fold ranges under that constraint, where the remaining weight for the split hands can be distributed freely between 3-betting and folding (e.g. the remaining 50% of 55 can be used as 3-bets or folds, as Pio sees fit)

Is this possible?

I think I asked the same thing earlier and was told no, but this is the feature I would like to see most! Node locking explorations can be inhibited significantly without this capability.

Sadly this is not possible as of now. You can lock the whole range or specific combos but not fractions of specific combos. It is possible to implement though so it has chances to appear at some point

I don't think this is the issue. I start it from C:\Piosolver and that is what it says in the updater, Pio updater is not in the same folder as pio viewer tho, but it is in your video?

It doesn't need to be although it will be after you successfully update to 1.9
Can you email support@piosolver.com with your registered email (or just provide the key). I will check what happened there.

Great! :-) Like tobakudan I think this would be a very useful feature. There is no inherent reason to be constrained to playing every hand in a binary way (either in a range, or not). Because of lack of efficient tools for building and testing strategies accurately, this is mostly the way range building has been done throughout the years, but now it can be done better.

One area where I would like to use it is for adapting HU solutions to scenarios with players left to act. For example, I can compute an IP strategy vs an UTG opener and use that as a starting point for playing BU vs UTG, but I need to tweak the strategy to account for the effect of having SB and BB left to act. One way to do that is to choose a reasonable calling range (with fractions, similar to, but somewhat stronger than the HU solution), lock it, and recalculate the 3B range under that constraint.

It would be nice to have tools to work with range building using specific fractions of combos, and this feature has my vote. :-)

Sorry if this has already been asked a million times before, but what is an "exploitable for %" to consider a solution good enough?

The original reason was because of implementation reasons. If you we allow locking any node in the tree with exact information about locked frequencies that's a lot of additional memory while locked/unlocked information is only one additional bit per combo. What can be done though is limit this detailed locking to say flop/turn and then memory is no longer a problem.

It's a judgement call. We feel 0.35% is very good, 0.25% is close to perfect, 0.5% is a good overview although rivers might be a bit off still.
It's the best to experiment and look at EVs when you are browsing. You will develop a feeling for that. If you just want to see general frequencies/range composition then 0.5% is definitely enough. If you want to go deep into analyzing blocker implications then maybe you want better precision.

Hi guys

For those of you who are interested - I made a video on how to run PioSOLVER on dedicated servers and EC2 instances.

You can view it hear:

https://www.youtube.com/watch?v=ihMc3dCU_Rw

Feel free to leave comments/likes/shares. More PioSOLVER help vids to come.

Peace

6seven8

Wow thanks! This is awesome

Any chance you could somehow give estimates as to how much we could expect to solve with each service using the plans that you used for trees of whatever complexity you solved for?

It's a very nice video. I haven't used EC2 myself and it's easier than I thought it would be.
Some remarks:
-Pio licenses (other than edge) are not intended as floating licenses but as per computer licenses (with reasonable limit of computer changes/travelling/etc.) so you will run into limit if you try to do that with basic/pro license (constant deactivating/activating on another computer); dedicated servers are better for that as you always get the same machine

-If you are browsing offerings of various providers make sure you get Intel CPUs, some offer impressive number of AMD cores but they are just very slow and will offer worse performance than even 3x less Xeon cores (I don't know about the newest Zen generation of AMD though)

-you can run a bench command https://www.youtube.com/watch?v=5Ozjfnatwms (question number 2) to see how your instance is performing; preferably run it a few times because there is slight variance (the time depends on what else running at the computer at the same time)

Overall cloud instances and dedicated servers are great options for the preflop solver as building big hardware at home is expensive and for many people not very practical.

We all cannot wait to have this feature.

I think this is what you are asking for;

Solution: Preflop, 100bb, 6max heads-up subgame
Subsets: 91
Average Time: 1 day
Exploitability: 1bb/100
Server Hardware: Intel Xeon 6core CPU/128GB ram
1 month Rental agreement @ OVH.com
$179.99+$25 for license


Thanks

hi punter,
what are the differences between "72_THREE_EQ_2__814.txt" and "94_FOUR_EQ_2__690.txt" files from the 487 subset?
Im interested in solving and analize them for postflop play with predefined preflop ranges, which do you recommend?

They are just 2 the best ones out of thousands generated. None of them is better (one was better in some set of metrics and another was better in another set) so it doesn't matter which one you use.

Yes. Just one more thing - what was the estimated complexity of the subgame?

You can run memory estimations using basic/pro version without actually having the RAM to run those trees.

I'm looking at the EV difference between PIO's mixed solutions (Bet AQ and AJ half the time on an A82r flop, check the other half) and how I think most players spread their range (bet AQ all the time and check AJ all the time).

With a pot of 75, the EV in the first case is 39.7. In the second it's 36.9. I guess this is due to PIO knowing exactly what our strategy is? So, for instance, if we never bet AJ (or JJ) on that flop, then a J on the turn can't improve our hand and PIO knows that. The reason solvers use a lot of mixed strategies for most of their hand combos then is exactly for this reason - board coverage on later streets and keeping their clairvoyant opponent in the dark?

Having said that, would the shown EV difference matter in actual play? Unless your breathing opponent knows exactly which hands you bet and which you check - and knows exactly how to use that knowledge - it shouldn't matter, right? So as long as we're roughly using the GTO ratios of betting/checking it won't matter with which exact hands we do it?

I'm asking mainly because I'm looking for some pratical shortcuts to implement the solutions into my game. If I know I need 3 combos of sets in each of my check/calling. betting and check/raising ranges, I'd rather work with a rule like "check/call the highest flop card (here A, so X/call all AAs), bet the middle card and check/raise the low card" than one that says to use one combo of each set in all three ranges.

Right, but it depends on how they set up the game, which I don't know.

Solvers are clairvoyant in a sense that they know a starting range (at the beginning of the tree) and they need to protect against opponent doing something like trying to exploit AQ not being in a checking range. They don't know what the opponent is actually doing but they need to protect against all the things the opponent could hypothetically do.

It's a judgement call. It probably won't matter much, even in theory if you lock AQ to being a bet and AJ to being a check (you can do that with combo locking) the EV difference is going to be very small so there isn't that much to exploit to begin with.

Well, if your opponent is clever enough to notice that you are playing ranges that simplified (stronger hands go into a bet, weaker go into a check) then maybe they are able to adjust to it, it's hard to be sure.

My intuition is that people won't be able to exploit you. I am not an active player though and some people are really good at poker and getting better so who knows. I don't think I am equipped to answer that with 100% certainty