Hi,
I tried to use the plugin "multiple files runouts aggregated frequencies analysis".
On the flop, it analyzed all the files which I saved in a separate folder, but on the turn I got the message ERROR: show_node incorrect path. I used fully saved trees for this. Is there anything I can do to prevent this error from occuring?

Make sure you have the newest versions of the solver and the viewer which should be:
-solver 1.5.0
-viewer 1.5.2

There was a bug in the viewer before 1.5.2 which caused this error when using multi-flop aggregation report

If you still get that in the newest version, the best would be if you go to Tools->Configuration->enable logging, repeat the action (so it crashes) and send me log.txt file.

Thanks, it was the old version of the viewer which caused the error. This is a very cool feature .

Hello.
Is it possible to block single card (not the full range) when calculating?

It's not.
Would you like to be able to remove the card from the deck or would you like to be able to lock a strategy for a single hand? (or subset of hands). Adding the latter shouldn't be too difficult if there is demand for it.

Lock a strategy for a single hand. I think it is necessary to build own strategy closest to equilibrium. I am surprised that no one is paying attention

Yeah, this is a useful feature.
We are paying attention, there are just many features to implement and it takes time to do so. I am putting it in the to-do.

+1

The node locking interface in general needs an overhaul. We will do this first and maybe add functionality later. There are two known bugs there right now:

1)The interface doesn't weigh hands correctly if the starting weight is less than 1
2)The interface sets nonsense if you put 0's for all the actions (for a specific combo)

1) is not that bad (you see what you get so you can adjust it). 2) may lead to nonsense results. While 2) could be qualified as a user error it's hard to catch and some verification mechanism is needed.

This is also why I am avoiding making a tutorial about it, some fixes/re-design is needed. It's a very rough part of the program right now.

Locking specific hands isn't very conceptually difficult and if limited to a flop (or maybe flop + turn) won't cost much memory/performance but we need some smarter way to input strategies to make it usable.
Anyway, the suggestion is noted and I think it's a very cool functionality to have.

hi all,

When considering bet sizes I know what is 'standard' in my games but those sizes have come about most probably though none analytical reasons. I'm trying to reset and relearn poker with Pio as my baseline.

I really interested in the dynamic nature of how range, stack size, bet size, and texture interact. If I had a PC that was infinetly powerful and set the sizing at 1,2,3,4 etc to 100 as my stack size and the raise size the same and did the same for turn and river would an equilibrium strategy favor a certain size? Or would there be a ton of mixing? I feel there has to be a threshold with portions of our range that past a certain point we would start to loose EV

Would be great to hear people thoughts.

We don't know that but my best guess is that there wouldn't be many sizes. There is a cost associated to splitting ranges (it's harder to balance and you lose EV) so it only makes sense to split if there is some tangible benefits. It seems most toy games seem to converge to not that many bet sizes used even if you insert more, that is as long as the opponent has some kind of threat available (preferably ability to raise/check-raise and some strong hands in their range).

Another problem is finding "best sizing" defined as best assuming we only use this one. This is one of the most mentioned problems and we are likely to add some tools to make such analysis easier (although it will be quite demanding when it comes to CPU time).

Most of the discussion takes place on our Skype group these days although it's getting more difficult to follow because of the volume. We are likely to add a forum in near future. It seems that this thread sadly didn't get much traction when it comes to strategy discussion, maybe people prefer not to share publicly. Make sure to join the group if you haven't already

Sweet !!

Thanks for the great reply!

I have and its really good. But thought posting here would be better suited for theory discussion. A forum would be really good.

In general it's good to post here because then the answers can be seen by more people and asking here forces me to formulate my thoughts more carefully as well so yeah, please do
As to the forum, there are specific plans and it's likely to happen in near future.

Sorry if this has already been asked - I know there was some similar discussion but couldn't find a specific answer to this. Also I understand that it's a question regarding a crossover between crEV and Pio so I have posted in both.

I am creating ranges for turns in crEV using the postflop conditions menu and then importing those ranges into PioSolver. What I'd like to be able to do is import ranges together with weights. For example I say IP cbets 2nd pair 30% of the time. If I then export the range from crEV as a text string it doesn't include this weight. It just seems to appear in the text as though each combo is either there 100% or not there. Is there any good way that either piece of software can handle the weighting of postflop ranges?

It should just work if you copy-paste into range selector. If it doesn't maybe we missed something about the format. Can you paste a sample range string which doesn't work?

After looking into it a bit more I think the issue is on the crEV side. The text string that gets output doesn't include weights. For example if I say to bet AA 10% of the time on the flop, the text string is still given as [AcAd,AhAd,AdAs,AhAc,AhAs,AcAs]

Yeah, looks like a bug in CREV. We accept CREV format (something like "[30]AA, KK[/30]") as it was a common request but yeah... the information about weights needs to be there for us to be able to do anything about it.

Thanks, I'll ask in the crEV forum.

Another couple of quick questions that have probably already been answered...

a) what does Pio mean?
b) Why does Pio want to check quads on the river 0.002% of the time (when there is no straight flush available)?

It's a name it had long before it became a public project.
P is for poker, I is for Interface O is Optimal. Don't ask me why they are in that order.

Currently the solver doesn't have any poker knowledge hard coded into it.
This means that every tree starts with uniform probability distribution over all actions (so if you have bet and a check available it's 50%-50%). At every step the solver adjusts a bit in direction of the best action. While this process would reach 100% - 0% in ideal world in the real one it sometimes takes a long time.
Another reason might be that sometimes there is no incentive to bet (because the opponent is folding everything anyway). in such case the EVs of betting and checking would be the same and thus the solver won't adjust.

Educating the solver so it always bets the nuts in position on the river, never folds nuts, never check-calls nuts on the river etc. is an interesting problem but probably not worth worrying about at this stage - there is lower hanging fruit in performance department available and there is also the fact that there is currently more value in developing new functionality to make analyzing results easier/more productive than optimizing the solving process itself.

I agree however that it's not very elegant when it takes a lot of iterations for the solver to stop folding very strong hands (even if the remaining folding % is miniscule) so it will be taken care of at some point.

gto is for fish

The way the program displays percentages is confusing. It seems to do it as a percentage of starting combos rather than a percentage of combos that arrive at a certain node.

For example, I was looking at this spot where it bets top set on the flop with .33 weight (all 3 set combos in range, so in combo mode this number was .99). On the turn it is betting with ~.2 weight, but in combo mode it is ~.66, which means it is betting about 2/3 of the sets that get to the turn bet node. So seeing this .2 number is confusing because it does not mean bet 20% of combos that bet the flop and now are at turn, it means bet 20% of combos that started from before the flop.

Is there a way to toggle this setting, or am I just confusing myself, or what?

There are two different values: weight at given point and strategy at given point.
Strategy at given point tells you how often an action should be taken (so for example bet 0.65 check 0.35) and weight means how often we have this combo in range.

From this:

It seems you are a bit confused between the two so let me explain.
There is a strategy view which tells you how often an action is taken assuming we are at given point, for example bet 0.65, check 0.35, like here:



There is also range view which is a different thing, it tells us how often we have given hand in our current range:



There are also clickable rectangles which shows you range for every action, like here:



Notice that it doesn't make sense to show strategy for bet and check separately because that's what "Strategy" already shows.

So now, specifically for your case:

This doesn't make sense. It either bets with 0.33 frequency or the betting range is 0.33 (you can switch it to combos).

In the combo mode it would be 3 (Tools->Configuration->"show total combos in range boxes on the 13x13 grid, rather than average for category). In weight mode it's 0.99 but the strategy (frequency) for the bet was 0.33.

It seems like you are looking at wrong numbers. I hope it's already clear from what I've written above. It doesn't bet with .2 weight, the betting range is .2 and this is useful information. If you would like to see strategy (which would be bet 0.66 in that case) click "Strategy" or "Strategy + EV" button.

Also notice that there is "square size proportional to weight" checkbox on the right which scales all the squares to current weight. This way you can combine Strategy and Range views in one (having colors/numbers for the strategy and square size representing current range).

I hope that helps a bit.

Any thoughts on adding rake impact?

CardrunnersEV have just added a function to output weighted ranges. Would you consider creating the ability to import those weighted ranges?

Also I have a theory/strategy related question - I ran a spot yesterday where I'd set ranges for getting to the turn, and Pio wanted OOP to check 80% and then fold to a bet 60% of the time (60% pot bet). While I realise that my ranges for getting to the turn aren't going to be perfectly optimal, they're not going to be too far off, and I can't see how a strategy of check 80%, fold 60% is ever going to be a valid, unexploitable strategy. Am I missing something?