what kind of piosolver licence do I need for buying/viewing lets say a sng pack from piocloud?

As of now you need the preflop solver for that (edge license) but it might change in the near future.

ok,
1) do we also need a powerful computer to browse it or it doesnt really matter?
2) would you some how let us know when we dont longer need edge to view it?

No, browsing doesn't require powerful hardware. It can be done on any laptop which runs 64bit Windows.

We never give any ETA's for anything. The reason for that it's impossible to estimate time for any software project so it's never a responsible thing to try.
When it's known it will be announced.

"64GB is already very decent for practical use but for big trees you need 128GB"

Hi Punter. I have the preflop solver and used 24gb RAM for some solves. I play 25bb and below. Is this going to be accurate enough? The poster you responded to didn't specify what games he was playing. If one is using a number of sizings for deep cash I see there could be a RAM limitation.

I can easily add RAM to 32gb and limit sizings to one size per street and also a river shove.

Thanks for the responses in the thread, and for the great software.

How many flops are you able to squeeze in?
If it's 25+ then it should be accurate in terms of EV and general frequencies but may vary a lot when it comes to individual combos.
My view is that you can get a bit of useful practical information with 32GB but trees including limps and raises are not going to be viable.

Again, you can get some information but it's still a bit of a struggle for practical purposes. I would recommend researching cheap dedicated servers with more memory. For something like maybe 30-50$/month you can rent something with decent CPU (nothing powerful but decent) and 64GB of RAM. Do it for 2-3 months and you will have way more useful information. It will not clog your computer either.

Hello,
I am trying to understand how OOP / IP ranges impact on the GTO solution, if my questions are just due to a lack of GT understanding, please point me to the right links, in any case I love you software and I am considering to buy a license so it would be nice to read your answers.

Changing ranges in the same scenario considerably change PioSolver's GTO solution: isn't the equilibrium strategy the one that a player should expect the least possible loss regardless of opponent's strategy?

Example: CO vs BB raised pot in mtt late stages EFStack: 20bb
I've analyzed this hand with PioSolver with:
1. perceived range vs perceived range
2. BB(villain, Ah7s) actual hand vs perceived range
3. CO (hero, JhJd) actual hand vs perceived range
4. actual hand vs actual hand

Board: 2s2hAd9s
Results:
1. perceived vs perceived
OOP ck 100% forcing him to bet as he did in the real hand, than the solver dictates for
IP general strategy FOLD 11.9%, CALL 83.82%, raise 4.27% of the range: the actual hand (JJ) is CALLING 100%

2. actual vs perceived
OOP ck 100% forcing him to bet as he did in the real hand, than the solver dictates for
IP general strategy FOLD 36.9%, CALL 6.24%, raise 56.7% of the range: the actual hand (JJ) is raising 34% and folding 66%

3. perceived vs actual
OOP general strategy CHECK 6.9%, raise 94.1% of the range: the actual hand (A7) is BETTING 98.7% forcing him to bet as he did in the real hand, than the solver dictates for
IP (JJ) calling 82.12% folding 17.8%

4. actual vs actual
OOP donkbets 35%pot 49.76% - donkbets Ai 49.% - ck rest
IP folds 100%

Summing up for hero (IP, JJ) these are the GTO suggestion case by case:
perceived vs perceived: CALL 100%
villain actual vs hero perceived: CALL 0%, RAISE 34% FOLD 66%
villain perceived vs hero actual: CALL 82.12% fold 17.8%
actual vs actual: FOLD 100%

Please challenge my considerations if they don't make sense:

from a GTO Solver perspective (perceived vs perceived) I did the right thing (called)
from a villain contingent perspective (his hand this time) I was absolutely too passive (should have raised, which crossed my mind to put all the Ax hands in a tough spot as I had more stronger hands in my range)
from an hero contingent perspective (my hand this time) apparently I was about right
from a god perspective (full information) I was damn wrong

The thing that is making me think is why the two contingent (actual vs perceived) cases are so different: if I know my perceived range is way stronger shouldn't be GTO to bet all hands in the range at least some portion of the times? why 82% calling 12% folding?

Thanks for your thoughts

Thanks for you interest in the solver
Answering your questions below:

I am not sure what you mean by "ranges". If you mean starting ranges then there is no conflict between your statements. It's just that in different spots with different ranges the optimal strategies are different but still the optimal strategy ensures maximum possible payoff against all possible strategies of the opponent.

If you by "ranges" you mean strategies and you are using node-locking to do that then the solution calculated by the solver is optimal with the assumption that locked strategy can't be changed and that both players know that.

Feel free to ask questions here if the above doesn't answer your question.

That's what I get for reading from the top.
What you called "perceived" ranges are actually ranges. A range is a set of all possible hands a player could have at given point weighted by probabilities of having that hand.
Ranges influence the optimal solution. This is obvious once you realize that in optimal settings your opponent knows what your starting range is and can adjust accordingly. If it's JJ alone, it's easy to play against it, if it's many possible hands, it's not so easy.

The key is to realize what the range is (a probability distribution across all hands) and that your opponent is aware of it. At the start of the hand at least they know you have every hand with the same weight. On the flop it's more difficult. You either solved from preflop or you have a good guess. It doesn't make sense to solve individual hand vs range because GTO depends on both ranges. Solving with only one hand in range is equivalent of laying your hand face-up on the table.

I hope that makes sense, feel free to ask follow-up questions.

Thanks for your quick reply,

I am actually a bit confused too let's see if we use different wording for the same thing.

I believe that any strategy for player i is a match of one hand and one action (f,x,c,r) weighted by a certain probability between 0 and 1.

therefore the range of hand a player is supposed to be playing is already half of his strategy, the other half is what I think the gto algorithm should tell us: the frequency for each action so that opponent can not improve his ev by playing different from equilibrium himself

now from your answer:

I believe you are talking about "optimal" strategy (not equilibrium, as equilibrium does not ensures maximum payoffs but rather 0 exploitability)

So am I correct PioSolver does not calc equilibria but rather pareto-optimal?

Thanks again for your time

By strategy I mean a distribution of probabilities across all actions. So for example AsKh -> (0% 85% 15%) for fold/call/raise would be a strategy at a given point for a specific hand. Strategy in general is set of all those across the whole tree (strategies for all decision points for all hands).

Starting range is not part of a strategy, it's part of initial assumptions. If we are playing heads-up poker from preflop, the silent assumption is that both players have starting range of 100%. If we start from the flop we need to make some guesses.

Equilibrium by definition is a state at which neither side can improve by changing their strategy alone. By "optimal" I mean exactly that. Equilibrium has a property that it guarantees the highest possible payoff out of all possible strategies if we don't know what opponent is going to do.

I am not sure what you mean by "exploitability" here but I can tell you what we mean. Exploitability is how much we would lose against a perfect opponent (the one knowing our exact strategy and playing in perfectly exploitable way against us) per hand assuming we switch sides every second hand (so in SB vs BB we play SB one hand and BB next hand etc.)

It's true that there is a lot of confusion about the term "optimal strategy" that's why I am explaining what we mean by it specifically. We call a strategy which wins the most against a specific opponent "maximally exploitive strategy" which we abbreviate to MES (and that is shown for every solution along with EV against current approximation).

PioSOLVER finds an equilibrium. To be more precise it finds epsilon-equilibrium (https://en.wikipedia.org/wiki/Epsilon-equilibrium) which is an approximation of Nash equilibrium. The error of that approximation is called "exploitability". If exploitability is 0 then epsilon equilibrium is the same as Nash equilibrium.

I suspect that your doubts come from the fact that initial range is an assumption and we don't know what initial range our opponent is going to have. This is true but there is no way around it other than solving from preflop (where you know the initial range exactly). Notice that solving against all possible starting ranges of the opponent would not be very interesting. Specifically we would always fold if their range is top set only and they go all-in on the flop. There is no point in finding mini-max across all starting ranges but it does make sense to find a mini-max across all possible strategies with a fixed starting range.

Feel free to ask follow up questions. As I wasn't sure what's the source of your doubts I tried to lay out what exactly PioSOLVER does.

You are right I have my doubt on this, but I understand also when you say "there is no way around it rather than solve from preflop starting ranges 100%" which might not be feasible.

I still have some doubts on this, sorry to be pedant, are you sure about it? I thought equilibrium is just ensuring to minimize opponent expectancy rather than maximise hero's one (your statement is only true if opponent plays nash himself, otherwise isn't rather common to have strategies with higher expectancy outside equilibrium?)

If you agree with it shouldn't your "maximally exploitive strategy" MES rather be called MINIMUM exploitable strategy?


Ok I thought this was actually the point but I just wanted to double check with you. As an approximation, do you think we are better off to give opponent a wider range? (ie. Can we calculate if it generate a greater distance from Nash to exclude an hand that might actually be in the range or to include it when it isn't?)

Thanks

It's feasible for HU situations and in fact our edge version does that although it solves it on a subset of flops as trees with all 1755 are just too big (even using 40-100 flop subsets you are looking for at least 64GB RAM for practical cases and 128GB or more for big trees with many options).

It's ok to be pedantic about definitions

We are maximizing our expectancy assuming the opponent is going to exploit us and the opponent is maximizing their expectancy assuming we are going to exploit them. In a perfect equilibrium our strategy is also MES vs their strategy and their strategy is MES vs ours. It has to be the case because if there was some other strategy for us to choose which higher EV against his strategy it wouldn't be equilibrium by definition (the definition is that no side can improve changing their strategy alone).

My statement reads: "assuming we don't know what opponent is going to do". This means they are free to play max exploit vs us. Consider this process:

We try every possible strategy s1, s2, .... , sn (there are bazilion of them but don't worry, just assume we can try every single one). We get the opponent playing max exploit every time and we see what our EV is. We choose the one which guarantees us the highest EV. Our opponent does the same and the resulting pair of strategies is equilibrium.

It's easy to prove as well:
1)if we have chosen a different strategy that would mean that either the opponent can improve (by choosing max exploit vs us) or we could improve by choosing a different strategy (to guarantee a higher payoff)
2)same reasoning goes for the opponent

Again, it's important to remember what the definition of equilibrium is. It then becomes obvious that both sides must play max exploit vs each other in that state.

Max exploitive strategy wins the most against a specific fixed strategy. The name is just fine. It's true that in equilibrium both strategies are MES'es vs each other although they are one of many possible MES'es.
It would make sense to call an equilibrium strategy a minimum exploit so to speak but there is already a name for it (equilibrium or optimal strategy). The term minimum exploitive strategy was recently claimed by one of our competitors to mean something else as well.

It's hard to say if it's better to err on the tighter or looser side. I guess it depends on how actual games are. It seems people are way too passive comparing to the solver so my wild guess is that we can play more hands than equilibrium suggests.

hi, I recently upgraded to windows 10 from win 7 and have lost all my data. When trying to activate my license it gives me an error message about activation. I already used my 2nd key on my laptop so I would like to know how to reset my key. I found a faq on your website but I can't find any of the things mentionned in it. I suppose you should make it somewhat clearer , add screenshots. there's alot of noobs like myself out there.
ty ty in advance

It is sadly a bug in licensing system (they expected the license should survive Win10 upgrade but it doesn't in practice, at least not most of the time).
Drop us an email with your key and ballpark registration date and we will take care of it.

Hi punter,
with a i7 6700k with 64gb ddr4 can i solve and roughly how long it will take for:

1) - interested for accurate preflop ranges
- 50bb deep, 82 flopsets
- preflop: limp, one raise size, one 3bet size, shove
- posflopt: one betsize/street, one raisesize/street, one 3bet/street, shove, oop bets on all streets.

2) - 100 bb deep
- only posflop with three bet sizes/street, one raisesize/street, one 3betsize/street, shove

Also how much hdd space will need each of those?

This not a well defined question (depends on the exact structure of the tree) but I will try to give you ballpark estimations.

As a rule trees with limps are rather big. Assuming you do reasonable cuts in postflop play you will be able to fit around 40-45 flop subset under 64GB.

82 flop subset is a big one. You may be able to use it on trees without a limp.

Assuming:
-100% ranges
-3 sizes everywhere (40% 70% 100%) + allins if they are smalle than 5x pot
-all the donk bets allowed (also 3 sizes for them)
-one raise size
-starting pot 55, stacks 975
-rainbow board

The tree would be 48GB. It would take a very long time to solve it (at least several hours).
Factors like reducing number of sizes for donk bets (possible in the next release), paired or suited board, smaller ranges, cutting some lines make the tree smaller.

Im assuming this is RAM? So with my 64gb I have it covered?

..and thank you for your time!

Yes, it would be 48GB in RAM.
If you want to save the tree for later it's highly recommended to use either flop + turn saves (about 100x smaller) or flop only saves (about 1000x smaller). Flop + turn saves are better in general because it's faster to recalculate rivers only when you are browsing (it usually takes about 200ms and happens automatically when you are browsing so for practical purposes small saves are almost as good as full ones).

Does Piotr still support edge users over skype 1 on 1? I have some issues with a solver and he don't respond in skype anymore longer than usual.

Very rarely. It's not sustainable anymore as I get 20+ pms every time I log in.
When you have a problem the best idea is to either send an email or leave detailed pm on Skype (by detailed I mean exact problem description with exact errors etc.). I will answer when I log in. I try to limit my Skype usage to minimum. I wouldn't get any work done otherwise.

I help people directly when they have problems requiring such help but unfortunately 95%+ pms I get is "hey, how do I do X, it's easier to ask than read the FAQ, thanks!".

All general questions should be asked here as well. It's way more efficient as more people can see the answer and it's easier to link to it in the future. Email is 2nd option when the problem is too specific and/or is about licensing.
You can ask on Skype if you think it's the best way but it's likely to be lost among many pms.

are there any guides out regarding how to install PIO edge on a dedicated server to run pre-flop solver?

There are two things you need to do:

1)setup your server (you need Windows there) and connect to it via remote desktop
2)install PioSOLVER there

We don't provide guides nor help with step 1) as there are too many different solutions and there are people way more qualified than us in this area

Step 2) is done exactly the same way as on your home computer. Just run the updater and follow the installation video.

Let me tease one feature from coming 1.9 release. It will be possible to dump all line frequencies from the tree both postflop and preflop. It's especially useful for preflop solutons as one will be able to see what's average cbet frequency, 2nd barrel frequency etc.
For example I solved BTN (about 47% stealing range, 100bb, 6max) vs BB spot from preflop with someone simplistic postflop play (55%/75%/75% bet sizings for flop/turn/river) Those are the resulting frequencies:

http://pastebin.com/taZDS1xD
You can see that:
-r:0:b25:c:c:b55 - 58.740% (cbet is made 58.7% of the time)
-r:0:b25:c:c:b55:c:c:b141 - 47.227% (2nd barrel is made 47.2% of the time)
-r:0:b25:c:c:b55:c:c:b141:c:c:b356 - 51.634% (3rd barrel is made 51.6% of the time)

r:0:b25:c:c:b55:c:c:c:b141:f - 55.669%
r:0:b25:c:c:b55:c:c:c:b141:c - 36.435%
r:0:b25:c:c:b55:c:c:c:b141:b313 - 7.895%

If IP cbets but checks behind turn then they fold 55.7% to a stab on the river, call 36.4%, check-raise 7.9%

I think it's pretty cool because you will now be able to compare those optimal stats to what people have in your database
Initial assumptions as well as bet sizings will affect those of course but still it may give good idea how optimal play looks like across the whole tree.

hey there,

I bought and am using pio solver pro on my Mac thru parallels. my Mac has started acting weird lately, altho I think it's related to something else. I backed up all my stuff to external hard drive and am prolly going to have to reformat. is everything going to be alright with my license of pio or am I going to have to reinstall with new key or something?

thanks

Preferably you would deactivate it first, like here:

https://www.youtube.com/watch?v=5Ozjfnatwms (question number 3).

and after all reinstalling is done just activate it again.
You can also email us after the fact if it doesn't work but it's safer to deactivate now.