is using decimals possible? i m getting incorrect pot/stacks error, obv its not that big of a deal but its annoying

This works but for now only if you drag&drop to an icon, not opened app.

It's not possible and there are good reasons for that.
Using floating point math for bet sizes would make comparison between lines impossible and thus would make it impossible to have features like aggregation/changing river cards/line comparison impossible. Implementing fixed point decimal math just for the purpose is a bit of a pain and then again - you need to make an arbitrary choice about how many decimal places you want and how much space you want to use to keep them. Limiting it to integer values has nice property of simple and unambiguous node identifiers which are a base for many very nice features.

I have a few suggestions for improving the speed and usability:

- More hotkeys in general

- Hotkeys for: range explorer in current spot, node locking, show strategy and show EV, close pop-up (with ESC?)

- Hit enter to apply range in range selector

- A way to choose a sub tree of a flop simulation and make a new turn/ river sim of it (so you can easier tweak bet sizes and ranges and compare EV)
Can this be done by simply copy/pasting ranges now?

Is there an option to buy the basic version and later upgrade to higher versions?

thanks for the quick and thorough replies!

one point in particular i was not clear enough on, and two of my separate questions sort of got mixed into one:

What I intended the quoted part to be was a visual display of additional buttons, so that that "cf" "cc" "cr" etc would each be a different button, applying only to the current street but across multiple actions of that street. so instead of the 3 buttons we currently have, we'd have a total of 9, perhaps with the new 6 ones being hideable with a checkbox or something. i think that was the idea you said you played around with yourself in 2013.

As for multistreet views, I agree that the details of this feature -- both the subtleties of the calculation and the UI -- would need thinking through. And I agree there will always be specific hands that, as you said, "aren't that interesting" from the view we're looking at. This is essentially the result of data-collapsing: by necessity we are averaging across turn and river cards, so context-specific situations get lost. But....

1. This same criticism applies to nearly all views. Eg, when I currently look at the "check" view, some of those hands are c/raising, some c/calling, some c/folding, and yet looking at the averaged over "check" is still valuable.

2. In the same way, I think getting a sense of my "check calldown" range (and similar ranges) would be valuable. And keep in mind, if we are averaging over all cards, some hand which was going to c/r and barrel on most turn cards but c/callsdown instead on 1 really bad turn card doesn't contribute much overall, so it would will be very lightly shaded. That is, the averaging should work pretty well to produce a useful metric.

Seems like a cool idea, indeed. But one problem you have to think through clearly is that there are several different bet sizes - so, do you x/r a certain hand vs 1/2 pot bet, but x/c vs full pot, and x/f vs 1.5 pot overbet etc. ? One hand might thus be in all the sub-ranges of "check" at the same time ...

personally i've been doing LHE analysis, so that's not an issue. but yeah, to make this a general feature applicable to NL too would require more thinking.

Hello,

We are relasing PiOSOLVER 1.5 today. This is mainly a bug-fix release about which you can read here:
http://piosolver.com/blogs/news/43559301-piosolver-1-5

That being said:

Here you go.

Hopefully this will no longer happen. Please report if it does.

There are now graphs!
You can even reverse them as commonly requested in response to pictures from early version posted on Skype.

Yes

Done.

This should be fixed or at least greatly improved.

Like this?

That's it for now. The updates start rolling now and should be available tomorrow (edge version is already updated).
The free version is going to get an update pretty soon as well.


Yes.

Great update! Like any good feature, it's already making me want more. For the equity graphs, the current graph is the cumulative distribution function. I'd like the option to see the probability density function, which better shows information like a flat or polarized range ("U" shape graph).

I am not sure if I understand what you mean. Just in case, on the graph X axis has all combos starting from the weakest while Y dimension is equity and EV against opponent's range.
The utmost right upper corner represents equity of our strongest hand against the whole opponent's range.

yes, sorry cdf was technically the wrong term for that. the view i'd like to see is the following histogram:

X axis is equity. leftmost point could be either 0 or the equity of your worst hand. rightmost point could be 100 or equity of your best hand. Y axis would be the % of all hands with that equity. of course, you'd probably want a simple histogram algorithm to autobucket the equity into sensibly ranges, to make the graph reasonable smooth.

Thinking about this a bit more, an even better feature might be the ability to filter your bet range to show, eg, the bottom 20% equity hands of your bet range. Or top 20%. Or mid X%. Range selection style sliders would work well for this. Essentially, this would be a simpler proxy for the multistreet feature I was discussing above. You could use low equity or EV as a way of seeing which hands were being used as part of a bluff range. Not perfect, but good enough for many purposes.


Also, after stopping and restarting a calc, I got this error:



I also tried closing and reopening the saved tree, but am still getting the error. Maybe a corrupt .cfr file? Is there way any to continue with the calc and not lose the previous results?

it looks like you've saved the tree small (meaning it forgot rivers), and you can't solve an incomplete tree. you can type rebuild_all into the solver command window and solve again afterwards if you want a better solution.

thanks! i actually had to do "rebuild_forgotten_streets", but it worked.

I was gonna suggest this too, I think it will be good visually.

Like this?

Only if you save the whole tree (the big save).
The reasons for that I hope are obvious: while we can cache results for exact strategies used in a small save the moment they change all the rivers are needed to calculate new correct EVs. That's also the reason you can't change strategies and lock nodes on small saves without re-building the rivers first.

rebuild_forgotten_streets rebuilds the missing parts of the tree but doesn't solve them. The moment you change any strategy anywhere in the tree the cache is going to be invalidated and your results will become nonsense (because the rivers are rebuilt but initialized to equal frequencies for every action).
The way to get the whole valid tree from a small one is this:

rebuild_forgotten_streets
solve_all_splits rivers

You can adjust required accuracy solve_all_splits uses (it's the same one which is used when you browse and the solver recalculates rives on the fly). Just be careful with it because if you choose something too good it will take a very long time to rebuild.
I wouldn't recommend doing that with a limit tree as they are huge and the process isn't very efficient.

Yeah, noted although it will take a while till it becomes a priority.

Exactly. Except, I can't figure our how your selecting the weakest portion. Whenever I click anywhere on the graph, it selects the top hands up to the point I click. What magical thing are you doing in the gif?

Thanks for clarifying. I tried this and it was slow. In general, will it be better to just resolve from scratch (or solve the tree to the full accuracy I want the first time, before saving)?

General question: Is there a way to lock all nodes, and then find the perfect counterstrategy? Something that could be used for:

1. Seeing what the maximum exploitive strategy looks like?
2. Setting up a villain strategy by hand (probably through the scripting interface) and then seeing the best way to play against that?

When locking a node, can you set the frequency and the program calculates the best range for that frequency (Bet flop = 70%) or its necessary to insert the range manually? I want to model a player that c-bets 70% on flop.

Right click.
Maybe it's not the best interface but it wasn't easy to fit sliders in there. One good thing about current interface is that you can also click bars representing specific categories (so like top pair or a flush draw etc.).

A lot depends what recalc_accuracy you choose but yeah the gain (if any) is probably not worth it unless you really want to revive the tree. The nature of the game is that 95%+ of information is on the rivers and solving the whole thing is mainly about solving rivers so forgetting them discards most of the information which is nice for storage but then you can't have the information back miraculously.
For example in the report I posted below you can see that there are 3969 ways to reach turn and then one needs to solve all of those for 48 rivers.

Yes, you can lock as many nodes as you like, the problem is that there are a lot of nodes in a tree. For example calling "node_count" on a full limit trees produces:

H_DEC: 954770
FLOP 5
TURN 2205
RIVER 952560
V_DEC: 954770
FLOP 5
TURN 2205
RIVER 952560
SPLIT: 3978
FLOP 9
TURN 3969
RIVER 0
END: 3242240
FLOP 8
TURN 3528
RIVER 3238704
ROOT: 1
FLOP 1
TURN 0
RIVER 0
TOTAL: 5155759

This means 954770 decision nodes per player. Locking specific strategies in all of them would take a while.

You can call set_mes. That sets a strategy of chosen player to max exploitive one in the whole tree. Be careful though because calling it means you won't be able to go back to your solution (so maybe it's good to save the full tree first).

If you lock some nodes and resume solving the solver will solve without touching your locked assumptions.
The way it works is that both strategy changing function as well as max exploitive calculations are forced to assume strategies in locked nodes are part of the rules of the game so to speak (so can't be changed). That's why when you resume solving with lock nodes you can get a very good accuracy - it's calculated assuming locked nodes are untouchable.
By the way, set_mes also respects locked nodes and is not allowed to change the strategies in them.

No. Such constraints aren't easy to program and would require developing a new algorithm just for them. It may happen one day but this is a difficult problem without an obvious solution.

Use mouse right and left buttons

How does one get into this skype group?

You get added by someone already there or by me. See PM.

Hi there,

set a tree as follows in the trial version, situation is river play:



Results are strange when villain cks back we are only betting AA QQ A9 (the nutts) 50% of the time:



thats seems very odd. What have I done wrong?

thx

As visible on the first screen-shot you discarded your tree and built a new one after solving.
A new tree always starts with equal strategies everywhere.
The proper way is this:

1)build a tree
2)click go
3)click stop when you are satisfied with the accuracy
4)don't click "build tree" again
5)go to the browser

Also you are looking at the "range view" not a strategy view. Range view shows you current range at the current point. To see strategies click "Strategy" or "Strategy + EV" buttons

EDIT:
Also I see this is 1.3.2 version. Please downoad the newest one (1.5) from here:

https://cdn.shopify.com/s/files/1/07...04600411993506

It has more features and is significantly faster.

legend thx mate