I don't know what I'm talking about - I just saw the other thread so I was asking. I'm trying to learn.

Me too! That's why I am asking Scylla. He is the oracle.

your right about same ev and mixed freq, but also it mixing them because wo mix strategy will be potential more exploitable. So its mixing up the hands w same ev to make dEV smaller (or maybe better will be to say, to make max exploit show less profit vs GTO strategy).

No, I don't see how it's possible to build the same tree if the effective stacks or pot have been changed. If there's higher/lower effective stacks, as compared to the pot, then the players will be all-in at an earlier/later point in the hand. So one way or another, it can not be the same tree.

If a tree is in a perfect equilibrium, then, if a mix is applied, the EVs for the actions in the mix will be exactly identical and the player will be indifferent between which action to choose.

In practical use of a GTO solver, the Nash distance (or dEV) will not be 0%, but for exampe 0.5%. The EVs between the actions in a mix will be slightly different, given that the solver is still trying to figure out the exact equilibrium.

The values to look at are the frequencies in the mix; that is the best guess that the solver has for you at that particular moment. The fact that one action has a slightly higher EV than another is something that has to be ignored.


No, it has nothing to do with the EV. It's just as easily possible for the action with 10% to have a higher EV. It just means that if you had run the solver for a tiny bit longer, it's next iteration in its learning algorithm would have been to try a slightly higher frequency for the action with the higher EV. Then again, in the next "step" after that it may easily have reverted back. There's just no way of knowing, so the values to look at are the current mix; the EVs should not be used for the decision process.

The EV of a hand will depend on which other hands they are mixed with. If you change a single combo, then the other combos will be adapted to correct for this, a different range will reach the turn, and it will not really be possible to compare the two situations.

Thank you Scylla!

So in practical terms how do I continue when both betting and checking have mixed frequencies? If one has 90% bet and one has 10% check - do you just RNG these?

Also - why aren't mixed frequencies always just 50% bet/50% check if they have the same EV?

Ideally, they should be randomly chosen.

Basically you're trying to set up your ranges so that you have a certain mix of strong vs weak hands everywhere (relative to villain's range). Just having strong hands in strong (bet,raise,reraise,etc) lines is a wasted opportunity, given that you can easily include a few bluffs and get value from your weak hands. Similarly, you can't just always go check-check with only your weak hands; strong hands need to be included in these lines as well. Setting up these frequencies requires complex betting strategies.

Scylla can you settle this debate once and for all and just tell me which statement is correct and which isn't.

A



B



A or B?

That's true.GTO is,by default, maxEV.People usually don't have a clue that when u find a leak in villain's play and u take exploitative line that can be countered (because is not GTO standard move if u were playing against optimal player)as long as he don't fight back,that's GTO.

That's the real explanation for mixing strategies. When solver tells u AA should be played 90%bet,10%check in a certain spot that is basically protecting of your checking continuation range.AA combo stand alone maybe will produce higher EV if bet 100%.For example, preflop AA should never be limped or checked GTO wise.

So you think A is more accurate than B when comparing these two statements?

A or B?

A) In GTO you take the highest EV line with your range but some hands in your range may not be max EV individually

B) You take the highest EV line with each individual hand, always, regardless of what the best strategy is.

I think that neither is true statement.It is A but u cannot turn a losing hand into a winning one because of randomness of flops.You may think of sacrificing Ev of a particular combo(let's say AA checking 10% just to have it in your checking range as a protection for whole range)but as long as I understand GTO,up to the river,it's all about frequencies and not about EV or equity.Up to the river,is guessing game in some way.

after I solved for 163 flops I cant seem to add more. ive tried both ways ( the button in database) as well as the button below the "run solver" button. I know it can be done but neither one of those are allowing me to. the numbers should say 163 solved 164 flops or we and I cannot for the life of me figure it out.

Neither is correct. You need to look at the weights; not the EVs. The weights are designed in such a way to prevent villain from counter-exploiting you. If you try to get a higher EV by deviating from the recommended weights, then you will become exploitable yourself.

Ah, I see what you mean. When using a subset, the buttons that allow you to add more trees are disabled. I would strongly recommend not adding trees to your subsets. The subsets are designed to reflect the total 1755 possible flops as closely as possible. Adding more trees to them will not make them more accurate; they will become (much) less accurate. This is because suddenly there's disproportionally more/less 2flush/paired/etc board, or A,K,Q,etc hands. Adding more trees simply throws off the entire balance. Adding more trees will only improve databases that were built with the "Add X random flops" feature.

Hello Scylla,

got another one from node-locking.

I have 100+ flops, I pick one then nodelock something on the flop. After I solve it, the flop shown is different. I have to choose again the flop from the list that I node-locked in order to view the solution.

Can you have fix for this?

So when we solve for a subset, it solves for that subset as a whole? If I add another flop it would have to recalculate all flops to re-balance the strategy? Is that correct?

How can I do a hand review then if the flop isnt present? For example..
Hero opens bu, bb flats hu to flop
Ah9h3c is flop in game
The subset has AT5r,A86 two tone

You see how the texture is different?

I dont know if A93 twotone was the exact flop, but it was something very similar. I was trying to see how a certain hand was played and ended up using flop strength ( ie. A pair below TP but above mid pair) and guessing or estimating I guess would be a better word on how/if I played a hand correctly.

That's why I wanted to add in a flop texture that I seem to find tricky.

*edit*
As of now, I solve those spots individually when I do session reviews but wanted to addin some of the more common spots to my db.

Should I continue solving each of those individually if the texture isnt represented well in the subset?

Or

Should I use the estimation process I mentioned above using the subset ?

Or

Both lol

Or something completely different?
Sent from my SM-G975U using Tapatalk

will it take a long time for the v120 version to be released?

Ok, I will see what I can do here.

No, the subset is perfectly balanced.
Adding more flops will throw off that balance and make the subset less accurate.
So flops should not be added to a subset.


I suppose that if you want to use this approach, then you should either create a database for the full 1755 flops, or just solve specifically for the situation that you're interested in. I would recommend the latter, given that planning in advance for possible flops, pot+stack ratio and tree composition would be a very difficult task to accomplish.

It will be released tomorrow, so please check back then.
I think you'll like it; it's quite a lot of fun.

Thank you!

Sent from my SM-G975U using Tapatalk

Hey Scylla,

I encountered some trouble.

I didn't notice I was running v.1.1.7 for a long time. When I tried v.1.1.8 I noticed that there was this RAM allocation feature. I only left it blank because it says that the program will automatically allocate the system's resources.

Now for some reason when I solved my BTN vs BB spot that only needs around 5GB RAM, in some boards it cannot solve because it says it needs more RAM.

I have a laptop which only has 8GB and it's solving the sim fine on v.1.1.7

I'm wondering why is this happening?

I currently have 16GB of ram in my system.

My bad, I noticed that each flop has different estimates.

Hi Scylla,

I’m having memory issues on sim sizes that I never had issues with before. I’ve checked my system out and can’t see an issue.

Looking here I’ve noticed a few issues since the update. My system can’t have ram added, is there a workaround?