No, you can only edit the entire decision. This is because if you were to edit any action, you're affecting all actions within the decision.

Can you provide a screenshot of some kind?
I'd just like to make sure that I understand you correctly.

I can't lock to show you ))

I was right, in combo section it is showed as 0/1/0 but in percentages section it is showed as 00.01%/99.99%/0 therefore that hand is displayed in hand matrix as hand for raising

I can't really judge without seeing a screenshot, but it is indeed possible for a hand to be present in a range with a near-zero number of combos.

Basically this is what I'm currently doing but it's not all random either. I usually add two bet sizes IP for double barrels and use the smaller one on pairing cards and when a lot of draws complete. OOP on the turn I use a different XR size against cbet and delayed cbet (2x as large in terms of %pot). The fastest way to build these trees in the other solvers has been to add two raise/bet sizes and all in when relevant.

I crEV I've spent 20min building a template tree with relevant bet sizes and then use that for future simulations.

I do by no means think either of the approaches are optimal but they've been more effective than building new trees every time I run a sim. One of the biggest weakness with the treebuilder in GTO+ at the moment is that you have to use the same bet size for the first bet in all lines and same flop raise size, double barrel size and BXB size. Using 1/3 pot flop cbet size will force OOP to use a very small turn probe size and allow IP to check back more capped than if you give OOP the option to probe with 2/3 pot+.

I have faith you'll find a good solution. If you at the very least give the option to manually edit bet sizes you allow people to set up trees they can use as templates for future sims.

Thank you for the feedback.

I was mainly referring to what happens at the point where you have gone beyond the play that you're interested in. I realize that many users will have many different wishes as to customized betting, and we will do our best to create a tree builder+editor that will provide users with the tools that they are looking for. However, these trees tend to be huge, and the question here is what to do once the tree has gone beyond the play that the user is interested in.

When the tree is finished up by the wizard, and the final 2 or 3 bets go in, I feel it should be with single targeted bets (or multiple targeted bets) as opposed to multiple static bets. This will ensure that the remaining bets make sense for the stack-to-pot ratio as opposed to using multiple static bets that are not necessarily relevant to the situation.

GTO+ is currently in beta. More extensive tree building options and editing have not been added yet, but are from a technical point of view actually one of the lesser challenges for us. The bigger decision here is to decide how it should actually work, given that there's many different and even conflicting wishes from different users to take into account. For this reason one of the steps is to provide a beta prior to actually building these tools so that users know what they are providing feedback on. So, that's the reason why we are only providing a basic tree builder in the first release.

What happened with the unexploitable shoving tool?? Can't find it anywhere. (https://gyazo.com/242af19db4f3839eb5efd3426f8f708e)

Thanks for your help.

Consider what I suggested a few posts up. Finishing up trees with only 1 sizing will give very inaccurate results because the biggest ev gain of having multiple bet sizes is precisely on the river.

For example, let's say you have a tree where you bet 50% on all streets. That will produce inferior results because you will miss out on a ton of ev on the river by not being able to bet small for thin value or overbet or even go allin with very strong hands. Imagine you're forced on the river to only bet 50% with your fullhouse. It's pretty easy to see why that strategy has a substantially lower ev than being allowed to go allin with your fullhouse and a proper ratio of bluffs.

Also this has ramifications for previous street play, because not being able to play optimally on the river will lead to suboptimal play on the flop and turn (maybe it will lead to the solver incorrectly suggesting to fastplay the nuts since it knows it cannot get all the money in on the river). Since river spots are so fast to solve it's a must that they have multiple bet sizes. Because the game is no limit it's mandatory that every decision point on the river has the option to voluntarily bet both small and very large or go allin.

The unexploitable shoving feature has been removed from the toolbar, given that it is only a special case of our more general GTO solver. That being said, should you want to use the unexploitable shoving tool, then it is still available by pressing Ctrl+Alt+U.

I've read your post, however, I'm about to go to bed, so I'll have to reply in the morning.
I just wanted to give a quick answer to FragDenLehrer's question.

Thanks for the reply and to be honest, I wasn't considering the exponential complexity and size of the tree. I understand in the future there will be an option to edit the tree and rerun the solver, so studying multiple bet sizes on CREV an bringing one size to the solver makes a lot of sense to me.

Also it makes a lot of sense to me. Maybe the river could have some extra option allowing overbets.

Thanks!

Can you perhaps show me an example of where this is true?
I have tried for myself, but have not succeeded so far.
For example, let's use the CREV file /equilibrium_sample_files/flop/flop_3bet.stx.
Open it and solve to 0.25%.
This will set up ranges for the river.

Now, I'll use a turn that is 4d and a river that is 3h.
We'll use the flop:check-check and turn:check-check lines to get to the river.
Set a checkpoint at the entry decision* and run the solver. The EV will be 13.23.



Now I will add two more bets for SB to consider, namely a probe bet of 10 and an overbet of 50**.



After that, we let the wizard create play for BB in his newly created decision. For this, right-click the decision node and, in the wizard, select the option that will place BB's bet closest to 60% of the pot. In the Bet 50 line this will be two bets of 0.37 times the pot and in the Bet 10 line this will be three bets of 0.5 times the pot.



If we now re-run the solver, we will see that SB's EV has increased from 13.23 to 13.28, which comes down to an increase of 0.4% in EV.



Now, I'm sure that we can agree that this is not a significant enough increase to justify the added complexity in the tree, as well as, should this be a flop tree, the additional requirements in resources.

So, can you or anyone else perhaps show me a spot where using multiple bet sizes makes a significant difference?
One important rule: In this "game" you can do whatever you like for hero, however for villain you can only use the wizard with the bet size closest to 0.6xpot.



*=To set a checkpoint, press F10 and click on the entry decision.
**=To put, for example, the "Bet 50" on top after having added it, left-click it and then left-click on the "Bet 20" action to insert it there; the order of actions is not relevant here though, so you can also just leave them in any order.


If you're looking at a river spot in a hand that starts at the river, then speed is indeed not an issue. However, for hands that start on the flop, pretty much 99% of the calculation time is spent on the river. So, this statement is not correct for flop hands.

However, even if it's a river hand, then a reservation that I have here is that the human brain (at least mine) is not capable of understanding such enormously complex situations. There's very little point in performing analysis on a tree that has a hundred decision nodes, all of which interact with each other. Simplification is needed (at least for me). I'd rather understand a simple situation as opposed to having a complex solution that I don't understand nearly as well.

I tend to agree with this, but the ability to change/add bet sizes on turn/river after flop is solved would kind of give the best of both worlds (fast flop sims and more details on later streets if needed).

Not the best example for why not to use more than one river size because it's a spot you don't need more than one size at equilibrium, but against passive players it can be good to add in overbet or small bets and mostly XF after you check. The spot you chose is a spot you have plenty of bluffs and villain should bet quite often (around 37% of the time) when checked to (so small bet isn't super important), you usually put sets and 2pair into XR range (so overbet isn't very important) and bet the medium-good hands for a standard size (the most important bet size).

Adding in all-in raise matters though in terms of EV, and it reduces IP river stab by almost 25% and makes IP check back more strong hands OTT:


Some other feedback about the beta is that I think it would be really useful to have no draw as a category in the draw area. Especially on draw heavy turns it would make it much easier to see how many hands like 2nd/3rd pairs with no draws are called/folded.

The problem is that in order to get accurate EVs you need to consider all lines where you'd want the ability to bet multiple sizings. For this I think it's easier to use Pio to show the difference in EV.

I used the free version of Pio which lets you solve the QsJh2h flop. The settings I used are as follows



I also removed the ability for OOP to donk bet. I solved to 1% exploitability. This game me an EV for IP of 4.262

After removing the multiple bet sizings and restricting IP to only a 50% bet on the river the EV for IP went down to 4.153. That's 2.6% decrease in EV on this specific flop.

Multiple Bet sizes and a dont 3bet and dont donk bet function would be nice in GTO+.
For example if I think that my opponent would never donk bet on flop then I can remove the donk bet tree with CREV but with GTO+ for now its not possible. Would be nice to see these functions in later updates.

By the way, the tree I built in PIO with 5 river bet sizes was only 2GB after removing donk bets. I imagine in CREV or GTO+ it would be even smaller because your software has the memory compression option and it's much faster to solve in general.

This in itself does unfortunately not really prove much. If villain is forced to always bet 50% of the pot then hero has a huge advantage in putting him in spots where this is a really ineffective bet size. In the system that we are considering bet sizes would not be fixed, but would be chosen depending on the stack-to-pot ratio. Other than that, hero has a similar handicap in also being limited in his choices of bet size. So I would consider any play with this system where the players are getting close to having 2 or 3 bets left to be an awkward and unrealistic spot for both players. I don't think there's a whole lot to be concluded from it. I suppose we will just build our own system and see how it works with non-static bet sizes.

GTO+ is currently in beta and one of the features that is not present yet is a tree builder/editor. This will most certainly be added though. We have however chosen to first come out with only a very basic tree builder in order to first get an impression what people are looking for in a tree builder. In designing interfaces it tends to be preferable to first have a clear picture of what features will be used. This prevents realizing at a later point that additional features are needed and then having to work them into the interface after the fact.



Ok, it's definitely bigger than the improvements I was able to find, but, in all fairness, it's still only a 3% increase in EV. The main question here is whether or not providing both players with a whole bag of tricks on the turn and river will actually significantly change the flop solution. My suspicion is that an almost identical result can be obtained by just using single targeted bets (or multiple targeted bets), thus saving a lot of memory and time. We will find out when we provide more extensive tree building and editing options. Just to be clear, extensive tree building+editing options will be available; we are just trying to figure out exactly which form they will take.


This is rather tricky from a definition perspective. This is because almost any hand has a chance to improve/degrade when a next board card comes off. Even 4 high is a draw to one pair. Are gutshots considered a draw? Are overcards a draw? Is K high a draw?

This will be added.
The current form of GTO+ is just the first beta.
More extensive tree building+editing options will be available in future releases.

The tree size would depend on the properties of the tree. Trees where the starting stack as compared to the starting pot are small will not require too much memory, even when using multiple bet sizes. This is because it simply does not take too many bets to get all the money in, no matter how the hand is played.

Hey scylla, you are probably aware of this but sometimes when running the solver the dEV actually starts increasing.

This can indeed happen temporarily while the solver is figuring out which lines belong where, however, if the solver is run long enough, in the end, it should always converge. I assume this is the case? If not, then please send a savefile to support.

Scylla ,
About multiply betsizes, yes EV increasment is not much , but it is because solver knows perfect play vs it , while in theory we have a little bit bigger EV with multiply betsizes rather than one betsize , in practice it won't be only 1-2% as our villains are not solvers and will make a lot of mistakes which will lead even to 20% of ev increasment , poker is not a game for simplification , the more complex our poker strategy is , the harder it will be to play against. You perfectly understand that GTO is mixed strategy , and there was a post where someone asked something like "why you can't make an update which rounds 0.69 combos to 0.70 for simplification and easier for brain", you answered him that it is wrong approach as it won't be same GTO anymore , same goes for betsizes.
Overall this feature has been asked for a lot of months already, it shouldn't be ignored for so long.

Any thoughts yet on discount amount?

The way I see it, most detail is lost in translation when using results from a GTO solver and turn it into something (tables for example) that can be used in live play. A lot of simplification is required, given that the human brain does not have perfect recall and understanding of millions of different situations that each deal with hundreds of hands. If our brains worked like machines, then all this extra detail would indeed have a significant influence, however, this is just not the case. All of this detail is only relevant if the user is able to replicate GTO play in live spots to a degree that would be super-human. That being said, this is just my point of view; we will provide options for creating extensive play on the unknown turns/rivers, so if you feel differently, you will be able to approach it in any way that you like.

No, the reason why I don't like this feature is because it does not accomplish anything in making GTO solutions easier to understand.
If it were, then I would gladly sacrifice accuracy.

I feel that with GTO+ we're going above and beyond what anyone was expecting of us, but I suppose your point of view is a different way of looking at it.