It will be added, just not right now.

If the format didn't change copy-paste into range selector will just work.

It is very possible to be optimal. It depends on exact line, ranges and runout. Overall "1-a" and "prevent automatic profit" nonsense did a lot of damage to general understanding of optimal play. Real optimal play has nothing to do with it and is, surprisingly more in line with "human" play (fold more on bad runouts, call more on good ones etc.).

Here is an example from 3bet pot in 100bb play, OOP cbets the flop and then checks the turn; IP then bets and it's OOP to act. Using aggregation report feature we can produce the following table:



I sorted it by fold frequency. You can see that on some cards OOP folds 64% while on others they fold 36% and this is facing 55% pot bet.

I don't have much more to say other than repeating that naive "prevent instant profit" and "1-a calling frequency" don't apply to multistreet games with draws like Holdem.
This is especially true for lines where one player already called and is then betting (or raising) but even in simple bet-call/bet-call type of lines 1-a is still significantly off on many runouts (although it is a better guideline there).

What was the X/R %? (X/R picks up a bigger payoff than X/C so you have to look at combination of X/C and X/R)

At the moment it seems the postflop ranges are being output in a different format to the way they output the weighted preflop ranges. I've asked in the crEV thread to see what they can do.

I would be very interested to learn more about this and why the 1-a/prevent instant profit is not relevant. Do you have any links to info about this?

I couldn't find the exact example but on a very similar one it wants to check 65%, then x/r about 7%, and fold 70%.

Hi,

I try run a script for know how many % of cbet is optimal , on diferents flops,
I set up like this;
https://gyazo.com/70d8b9f2c34f2a6c36395bb85d1f1895

But it is not working,

I tried seach info here and in the website, but cant find it

Thank you very much



set_range hero xxxxx
set_range villain xxxx
set_board jh 7h 2c
set_bet_sizes 50 75
set_pot 0 0 20
set_donk_bet 1
build_tree
go 600 seconds
wait_for_solver
forget rivers
dump_tree C:\Users\jeremy\Desktop\PioSolver free-version150\treees\test.cfr
set_board 9h 2d 2c
rebuild_tree
go 600 seconds
wait_for_solver
forget rivers
dump_tree C:\Users\Desktop\PioSolver free-version150\treees\test2.cfr

There is just no reason why it should be. I get this question a lot and it's frustrating because there is 0 reason why preventing instant profit or calling with 1-a frequency should be optimal in any game more complicated than the most simplistic river one (and even then only with numbers not real cards as card removal ruins it).

Maybe one intuitive thing:
Let's say the pot is 100 and OOP cbets 70 on the flop, IP calls and it's 240 in the pot on the turn. Now it goes check-bet 160.
It may seem that IP invested 160 to win 240 but it's not true - they invested 230 to win 170 this means that even if OOP folds a lot here IP still can't make money with nothing. They can make money calling with something which has some equity but it's very difficult to guess where the optimal calling frequency is - it depends on exact ranges and runout and how they match.

"Opponent shouldn't profit automatically" or "we should call 1-a" are basically worthless when thinking about those situations. Optimal play is about finding a strategy which guarantees maximal minimum payoff (that is guarantees the highest possible payoff assuming the opponent exploits us maximally) not about preventing instant profit.

There is no mathematical reason to believe we should prevent instant profit or call with 1-a frequency. Those are shortcuts which work in one street toy games with numbers instead of cards (and even then there are caveats). There are situations in Holdem when we won't allow the opponent to make automatic profit with junk, for example when we have stronger range and are in position but there are so many restrictions for it to apply it's not a useful tool.

Again, it would be useful to know what ranges you took for this example. If it was bet-call on the flop and it's the caller betting now it's the most natural thing in the world to fold a lot as explained above.

I am not sure what you want to achieve. You can run a tree with 2 (or more) sizes but to do that just fill in the tree config and click "generate script", don't fill this up yourself.
The concept of "optimal cbet size" isn't well defined, you may want to find one of the two:

1)solution for a tree with many cbet sizes (here just build a tree with many available)
2)best c-bet size assuming you can only pick one

2) is not solvable without a lot of trial and error (solve a lot of trees with one size but different one every time and pick one with the highest EV).

This is an old now obsolete tree building interface.
The way to use tree generation window is to first fill tree building form and then click "generate script" in there. The script will appear automatically.

Your betsize is probably too big for this runout else intuitively opponent would continue more so betsize can probably be reduced and get better bang for the buck. My intuition is that betsize needs to vary by runout. You can't oversimplify by using the same betsize for all runouts and still be optimal. We still have a lot to learn about GTO.

Your earlier post helped me a lot, thanks punter.

Hi there,

Would be good to get some IP from folks here bet-sizing.

Same ranges, stacks and board.



The only difference is that OOP can bet 25% pot or 50% in one run out of Pio solver and then 50, 100% pot in another, Turn and river are the same in terms of sizing available. It produces a really different strategy I guess its not surprising, but when trying to apply it in game which sizing is more optimal? unfortunately I can't have 25%, 50%, 100% as my comp can't handle it....

different strategy:

20% +50%



50% + 100%

It would be cool to have like a little window box that we could see retrospectively how a hand was played, then if we set up spots like this we could compare/contrast the evolution of the strategy basis different bet sizing..

While your at, it 3D topographical image of evolution with bullet time cam that narrates the journey of A9o across multiple streets. **** would be right yo

Having my mind blow by betsizing ATM. Its such an important factor and yet seemingly so difficult to implement. Was watching the Janda series on CR the other day. First two videos were very well done from a novices perspective, however the third was essentially 'ahhh here's a formula with limitations and 'it depends GL' Totally fair enough as you can only cover so much in an hour series but amazing how complex it is

It's hard to tell how to apply it. My view on it is that it's more productive to analyze general frequencies and range composition (how many strong hands it has in a betting range, how many draws, what it does with air etc.). If you use strategy+ev view you will see that in a lot of places EVs are almost the same (especially for bet/check decisions, less so for call-fold ones) so it's quite possible to construct a strategy with similar characteristics but with different hands included.
As you can see though it bets more with small sizes and less with bigger ones which makes sense, good thing to do is to compare EVs of whole strategies to see which size/pair of sizings it thinks is better (this sadly varies from board to board so not that easy to predict).


Some of those trees have a million (or more) decision nodes. I am not sure what "how the hand was played" could be represented if you mean the play in the whole tree.
If you just want a flop strategy you can make a small saves (either flop+turn or flop only). Then solve with different sizings and then open two viewers, load the save in 2nd one and compare. It's some clicking but it should work.

Yeah but millions of nodes...
Btw, if you build a tree you can try this:

1)Tools->Solver->arbitrary command
2)type: node_count
3)press [enter]

It will show you how many nodes of various kind the tree has.

What is difficult is to make it faster or use information from one solution to arrive faster at another. I am not sure about important part, if you compare EVs it often turns out that changing the bet sizes or removing one of the option almost doesn't influence your EV.

I think over time people will come up with shortcuts. Again, I think the way to go is to analyze range composition for bets and seeing how they are constructed depending on size.
We plan to extend aggregation reports in near future so it will be a bit easier to compare strategies across multiple trees.

much wisdom sir, many thanks

This will be amazing!

piosolver free has some 6max preflop ranges. Does the basic version come with more ranges like 3b / 4b etc?

It comes with a few more but not very detailed ones. Those ranges are me eye-balling averages over some good regulars from popular MSNL+ game on popular site. They aren't random but aren't necessary very good either.
We also include Cepheus opening/defending/3betting ranges for limit.

You can create your own btw and save them.

Any timeframe on forums for licenced users as I don't want to be part of a Skype group because of security risk?

(Skype is easy to abuse to cause DDOS when players are in a HUSNG).

What is the difference between basic/ pro / edge?

I am very busy right now but we are considering supporting a forum on:
http://apexstaking.com/forums/forum/piosolver/

Alec (b3rstare) is one of our first users and big fan of the software we think it might be a good place. I just need to talk to them about details and find some time to make an announcement. I think we will have a decision about official support from us in few days.

Please read this:
http://forumserver.twoplustwo.com/sh...&postcount=332

FWIW I think having it as part of your site would be better. Keep it in house where all different game types and players can mingle. Being associate with one particular group of people may make open and free discussion difficult.

We are too small a team for that. Maintaining a forum is a time commitment. We don't have enough man power for everything we want to do anyway.
We will see but it's not as simple as "make a forum and forget about it" as it adds n hours per week to maintain it and we don't have those available.

Hey, so the HU3B range and HUBBdefend ranges contradict each other as some of the hands the range suggest to flat most of the time are 3b in HU3b range 100% of the time? Any reason for this and how are the ranges constructed?

They were constructed by me eyeballing the stats. It wasn't any scientific effort nor GTO approximation. You can construct your own by using "Inverse range" options at the top and then removing the weakest hands.

FWIW, I support the suggestion of a trusted partner running your discussion forums. I'd rather see your team focus solely on software development. (I used to work in software development for ten years as an architecture director/marketing director).

Also, re: pre-flop hand range. Pio's are obviously illustrative samples only. Again, I'd rather Pio focus on software development and let others focus on the results and meaning of analysis.

Very much so.

Is there a solver command like "solve_rivers" - e.g. I did a small save (forget rivers), and open the file in the viewer again and want to round strategies now: is there a way to get all rivers into RAM so that rounding is possible again ?

Yes, like this:

1)rebuild_forgotten_streets
2)solve_all_splits rivers

You can round (or node lock) after doing 1) only but the rivers aren't solved so the results are worthless (the cache will be invalidated the moment you change strategies by either rounding or node locking so you will lose information about EVs which can no longer be retrieved without solving all rivers).

Step 2) solves all rivers to accuracy given by recalc_accuracy which you can set from GUI (Tools->Confguration). If you are going to do step 2) I would recommend solving rivers to maybe 0.005 or 0.002 (or 0.001) default value of 0.0005 is good for browsing as it takes about 200ms per river in normal trees but it's too low for solving all of them in one go (because it will take forever).
Running step 2) will take some time so be patient and start with not too low recalc_accuracy for rivers.