The maximum rake in milli smallblind (irrelevant if rake is 0%).

Save the tree, then load it again. In the load prompt, check "Edit stacksizes".

Hi,

I just downloaded the free version. I tried to run the monkersolverfree.exe
I got error message
"Could not create the Java Virtual Machine"

So, I followed the link in "getting started" and installed Java. I still get same error message again. What have I done wrong?

Check your browser version. If you're using firefox 32bit for example it will download the 32b version. So you'll have to install both versions if you're running chrome 64b and firefox 32b for instance. Anyway the Java64 is a prerequisite.

Can somebody explain the exact meaning and difference of buckets and texture?

Also as a feature:

I would like to be able to lock a certain strategy as it is, and adjust certain specific hands/filters for the frequency's I like. For a simple example: Right now if I want to see what happens if I let hero cbet any 90th percentile hand+, solver will adjust for this and start checking back every 80th-90th percentile hand so he ends up fairly balanced on turns. I want to make sure this option gets locked, so he actually ends up unbalanced on turns.

thx JackBurton

I tried a simple example for plo in the free version to get familar with it:
heads up on the river, 1 potsize bet left, btn has QsJsTh9h,Ts9s8h7h and bb has AAAA,AsKsJcTc on board 8s6s2s3dKd. I did not touch the standard settings.
I enabled all actions but not in the second row "auto" and "bet %"

Now, monkersolver is running with billions of iterations already. Does it stop at some time and shows me what range bb should be betting with which betsize, checkcalling, checkraising aso

or is it solved already (at least the frequencies)?

I am shown
Call 25% and All in 75%
AAAA 50% 0 1 AsKsTcJc 50% 150244 1
AAAA 50% -393 1


so, call means checking AAAA here as bb is first to act and there is no bet to call, right?

and I am not 100% sure what it tells me when I lick on the responses-buttons.
Clicking on All in shows the button calling and folding 50% each (makes sense).
Clicking on "call" which again means button is facing a check not a call(???) I am shown

Call 100% and All in 0%
T987 100% 100000 1
QJT9 100% 100000 1

so, button is checking behind 100% not calling(?)

tyvm

Hands are bucketed based on public information (texture buckets), and private information (strength buckets, which is a combination of hand strength/potential). If we have perfect information about the texture, no hand on a given board will be merged with a hand on a different board. If we don't have perfect texture, then for example AsTs on Ks5s3s 9h may be merged with AsTs on Ks5s3h 9s, even though the boards are not strategically equivalent.

I will look into the feature request.

I assume you mean you checked the check boxes of these actions, which does nothing until you use "Add" on some node. You will have to add them to the tree.

Yes, it is done. Note however that it only had the option to use all in, since no other bet size was added. Luckily, no other bet size would have been used in that scenario
Yes, call means check, and button is checking behind 100%.

Hello, I have a bunch of questions if you would please kindly take the time to answer as thoroughly as possible:

http://www.monkerware.com/faq.html
You say here that these are ε-Nash strategies. What does the solver define as ε? Or rather, by how much does the solver permit a player to maximize its EV over another player beyond equilibrium?

In multiway situations, can you offer some kind of proof that the solver's results actually yield strategies which are said ε distance from equilibrium?

Are full ranges, hand combo frequencies, and hand combo EVs viewable at every node of the tree? Is the root EV for every player viewable in all types of simulations?

What kind of system would be needed to run a full preflop 6max simulation? How large of a file would this be if compressed as much as possible? How long would it take to load?

How well can the solver deal with rake?

Regarding the chart mode. When running a solution and the call is 25% and i open the chart, the chart is all yellow. It would be more convinient if the chart has only the calling range (on this example). It doesnt make much sense to has all the hands from the beggining

How have you fixed this? If you change up the decimal weighing, the strategies get messed up?

I have a laptop with 8gb ram.

Could I solve HU spots with a single size bet size from the flop giving both players a range? (Pot limit omaha)

And how long exactly would it take?

Thanks

It makes no definition of ε. That should not be interpreted as anything other than the intention of the solver, which is to reduce said ε to the best of it's ability.

In multiway games it becomes particularly uninteresting to talk about exploitability, as any strategy can be exploited through involuntary or voluntary collusion.
The algorithm will however iteratively remove dominated actions which produces strong stategies. There are tons of papers on this, for example Gibson's thesis, chapter 4. http://richardggibson.appspot.com/st...-phd-paper.pdf
Yes.
If "full" means one postflop betsize, with donk bets removed, and various flat calls removed, you should fit it in a 128gb RAM machine. This should save to about 10-15gb, and load (note: not solve) in a few minutes (improvements will likely be done here).

It deals with rake as well as any other solver I would imagine, in that it sort of messes up theoretical guarantees due to the game no longer being zero-sum, but seems to work well practically.

Not sure what you mean. Yellow means no lock, so in the beginning the chart has no hands, rather than all hands.

By rounding to the closest percentage. Ranges will change a little, but not too much.


Generally, 8gb will not be enough. Required RAM will vary on type of flop (monotone takes least, rainbow most), size of ranges, and the tree size (which will naturally be lower if stack-to-pot ratio is high). Some trees will fit in 8gb, but most will not.

The time will vary on the above and on your CPU.

Another request:

I would like to be able to open trees in monkersolver which exceed my internal memory. So for example, let's say I have a tree solved of 40GB (solved on another PC), but my PC has 32GB, I still would like to be able to watch the solution, without doing any computations on it.

Given it can be quite easily imported into monkerviewer this should be possible I think? Reason being is that monkerviewer gives me wrong ranges, and it's very chaotic, I like the monkersolver UI a lot better.

What i mean is i ran a sim. This is the calling range



when i click lock call from chart i get this screen



Which is inconvenient because it has all possible hands while i should see only the hands from the calling range with the percentages. Like this



and then fix the range as i need to run the sim. Doesnt make much sense to have all hands

any plans to implement a scripting interface(similar to what piosolver has done) down the road?

** bug report **

when saving ranges to pio format if you uncheck 'weigh by reach probability' it does not work

When loading a run, choose for example streets=2 to only load preflop+flop, which will require less memory. This should work better in the latest version.

Well, what if you want to lock a hand which is not in the calling range, to call?
I guess it would be convenient to see the current frequencies, but it should not be a big inconvenience to open up the range chart next to the lock chart.

Very possible.

Could you provide some more information on this? I can't find any issue with it, unless you mean it saves the ranges without 'weigh by reach probability', which is working as intended.

would love the scripting if you decide to add it, i understand there are probably way more important things to work on atm though.



heres an example of what i'm talking about, as far as i can tell its saving the ranges with 'weigh by reach probability' when i would like to save them without. the image shows the same range with the box checked and unchecked and the output in piosolver.

Bug report

I have this call frequency of 13.5% with K2s calling 54%



Whether i have checked weigh by reach propability or not when i try to create a pio chart i get this



Not only the percentage of the hand is changing but the total range frequency. This should be fixed asap

Thank you

Both of these are due to normalization. For example, a range of 0.2 AA and 0.1 KK will normalize to 1 AA, 0.5 KK as generally, only relative weights are interesting, and it makes things clearer when ranges get very small.
It also solves a number of issues, such as a 1% AA, 0.5% KK range not being representable in PPT due to full percentage limitation, but normalized to 100% AA and 50% KK works fine.

It does not make much sense to normalize when exporting the strategy ('weigh by reach' unchecked), so that will be fixed. Also, feel free to provide a reason for exporting non-normalized ranges and I'll add an option for it.

Because then the charts will have wrong frequencies. its not obvious?

I think when you bring the chart up only the call/fold/raise hands with frequencies should be the only hands included because it will save much time if you want to add hands using the existing frequencies

I'm testing the free version of MonkerSolver. I ran until 'Iterations' equaled approximately 10x the 'Nodes' and then hit 'Stop', but it seems like the software is now frozen. Status is 'Running' and I can't click anything within the software. It's been like this for about 45 minutes. My CPU has been running Java at about 46% load continuously. Here is a screenshot: https://gyazo.com/44448bb94bd23fc492b97777f8b46eeb. I'm assuming this is not by design and it's simply crashed right?

Update: I noticed PokerJuice was running so I closed it, restarted Monker Solver and it then worked. Do you think it could have been connected to that? Can I not run both programs simultaneously?

Piosolver does not use these normalized ranges and also supports full decimal weights so it doesn't make much sense to use them with pio. If I want to export a range to pio for further use I need the non-normalized starting range for it to work correctly. Right now I need to enter the ranges manually from monker in order to use them in Pio.

+1 on normalization.
If you have a result with CALL 1,9% and RAISE 25%
and want to get this into 1 chart in Piosolver it will result in a chart where 30% of hands are played given the normalization increases the 1,9% to about 5% in piosolver.

-->unchecked weightbox has to save the original ranges imo without normalization.

agree on Post #163:
Locking by chart has to be changed/improve
solving like in post 163 seems a fine solution that you first see the range played as CALL when you click "lock CALL from chart" instead of an empty range(all hands yellow).
"Well, what if you want to lock a hand which is not in the calling range, to call?"
Should work by just clicking on that specific hand so its now locked for calling

Also if you want to edit further and click "lock CALL from chart" again then the already locked hands have to show locked in that vision.
Always starting with an empty chart is very inconvenient.

Is there a significant difference between the free and paid version?

I'd be inclined to purchase the full version but there are couple of such annoying things like:
- Ranges don't get saved when saving a tree, also they start off at 100% and there's no way to make it 0% unless you unclick EVERY single one of them
- Locking the entire range leaves off many combos out anyway, and the frequency (eg. stabbing OOP frequency) goes back to default even though I wanted it to be higher
- When you double click on a combo in the Tree Viewer thing, it dissapears forever and you get some 1@123.12 text
- No way to continue running the same tree
- Adding ALL-IN to every possible node - this is by far most annoying, what's the point of having ALL-IN vs 25% bet when there's 100bb behind!? agan, you have to remove all of them manually

Options for normalization has been added.

That is not by design, no. I don't see how another program could have caused it, unless possibly if it's using alot of RAM which would otherwise be available.

1) Ranges are not saved with the tree. The most convenient way to input ranges is by drag-and-dropping a file.
2) First time I hear about this bug. Pretty sure it's not in the full version, but I can't reproduce it in the free version either.
3) Fixed.
4) Save it, then load it to continue running. If you want to pause a run, you should reduce threads to 0 by pressing the - button instead of pressing stop.
5) Settings for when to auto add all ins have been added.