Hi!

Im new to the software and im trying to understand why pio gives me some, for me weird numbers.

If i use the predefined profile "6max-BTNvs3bet" and change the ranges to identical ranges for both IP and OOP, the solver returns a higher EV for IP

Example - "6max-BTNvs3bet" Ranges: AA,KK,QQ,JJ for both IP and OOP returns "EV IP: 100.531, EV OOP: 84.469"

Why is IP EV higher?

If i then check the browser, both IP and OOP have 50% equity.
But if OOP chooses to bet, OOP equity rises to 58.495%

Why does equity change due to betting?

Thanks in advance

In position player has an advantage in betting so it's natural they make more money than OOP player when both sides play optimally.

Yes because the ranges are the same. Equity is "all-in"equity that is assuming checkdown to the end and looking who wins.

When you are looking at equity after the bet you are not looking at equity of the initial range against initial range anymore. You are looking at equity of betting range of OOP against current range of IP. Betting range is usually stronger than checking range therefore betting range of the OOP is stronger than IP's range (at least before the call).

It's important to remember that:
1)equity = all-in equity without any further betting
2)EV = expected value in $ for the player at given point assuming optimal play of both players to the end of hand

I hope that makes sense

Thank you!

I want to look at this approach for 6 max. I'm aware that there are variables that may make the conclusions highly questionable, but for this type of analysis, is it better to:

a) Include 100% of hands preflop, simulate a subset (using 184-- is this necessary?), run aggregation report, and look at the EVs of individual hands.

b) Use an standard/estimated starting range that does not include the obvious defends, sim the subset, aggregation report, look at the EVs, and adjust the range from there?

The resulting ranges will depend on open size and amount to call, but I assume we can use the same report for all sizes and build ranges for each respective size using the EVs from the report. Is this correct?

Any additional info about this method is highly appreciated.

Question regarding PLOCalc:

If purchasing PLOCalc today would there be a discount for the solver (another product?) when it's released in the future?

Comment regarding PLOCalc:

I use a Mac with dual boot as well as a VM for when I don't feel like restarting. One license seems a bit restrictive.

How do i specify a bet size of all-in?
for example, if i want flop, turn and river bets to have choice of 50%pot and all-in.

There is nothing magical about 184 number. Smaller subsets are quite good, personally I don't think anything above 74 is necessary but it's a judgement call.

I am sorry but I don't feel I have anything interesting to say on this topic. Approximating multiway ranges is a hard problem, maybe it's not one worth worrying about as equilibrium in multiway pots is very shaky (one player doing something different may mean you are going to be losing money if you follow equilibrium).

The solver will be a product of the same line (same as basic/pro/edge with our holdem solver) so you will be able to upgrade for the difference in prices.

I encourage you to wait a bit, we will make some decision about 2nd activation soon (3 possible solutions are either selling it way cheaper, just adding it or adding it while increasing the price of the product, it depends on our assessment how big a market is there for PLO tools and what kind of price justify working on it).

Use "add-allin" checkbox, like here:
https://gyazo.com/e72effced4b0025bc922476c68871314

All-ins are only added if they are smaller than size specified in "Add allin only if less than % of the pot" field which is set to 500 by default (it's likely a good idea to set it to around 300 to avoid adding pointless overbets which only make the tree bigger and more difficult to solve.

So for customers that only need a single installation, still purchasing it now could be a mistake, because you may decide to to sell it cheaper soon?

No, I mean additional activation once you have the license (we do the same thing with Holdem, once you have the license you can purchase additional activations for it cheaper than buying another license).

Ok, just benched my threadripper build, getting an average of 2.1 seconds time taken.
My build:

Threadripper 1950x @3.9ghz
128gb G.skill Tridentz RGB @14-14-14-30 3060mhz
Gigabyte x399 Aorus Motherboard
EVGA GTX 1080 TI Hybrid
EVGA 1000Watt Platinum Power Supply
NZXT x52 Liquid Cooling
Corsair Crystal 570x Case



build is ripping through piosolver trees insanely fast!

Is going to be a PIO update soon? I am not sure if the "update" that you mentioned some weeks ago was the PIO Calc

Is it stock or did you overclock? AMD website says it's clocked at @3.4Ghz

I post a lot of things and answer a lot of emails, it's really hard to comment without specific quote included. There is going to be 1.10 version of PioSOLVER and it's going to be a free update. We don't give any timelines though.

Anyone into PLOCalc yet?

I try to syntax a range that consists of two high cards: A, [K-T] and 2 lower cards which have either 0 or 1 gap(s). The result should look like this:
In PJ or Odds Oracle you would simply syntax: A[K-T]: (23-89,24-79)

I tried over an hour but have no idea how to write this with PLOCalc, any idea?

This is difficult right now (various patterns of connectors/gappers are easy if we are talking about 4 cards but not with one/two high cards as well). We already talked on Skype about the solution for this problem. Macros are going to be available as well.
Thanks for helping us. We try to get those usability things right before proceeding with more advanced tools.

Hello, im recent to pio and have a question that is probably easy to answer.

For example, im analysing a SB vs BTN 3 bet pot situation and i want to see the results in at least 50 different flops, how can i do this?
Do i need to create a script? ( i dont know how to create them )

Yes, of course overclocked , 3.9ghz stable, ram is also overclocked to 3066mhz stable at 14-14-14-30 timings.


Cant really push the cpu any higher, requires too much voltage to stabilize 4ghz, and my current liquid cooling system quickly becomes insufficient to deal with the increased heat output.

I think the current speeds are more than enough for piosolver now however

Yes, scripts are made for this purpose. I go through generating and running a simple one, here:

https://www.youtube.com/watch?v=VhG_YsEvfwU (point 6, see the video description)

RAM speed likely doesn't matter much. It's a very impressive benchmark

Thanks punter

Anyways, I have noticed an issue with my computer and piosolver. It seems that whenever I try to build a new large tree and wipe the old, it takes ages to clear the ram of the old tree, and then equally long to build the new tree. For example, building a 80gb tree (and clearing the prior 80gb tree) takes around 20 minutes or longer. Is this normal?

On the other hand, when I build, say a fresh 80gb tree from scratch, it is MUCH faster from a fresh load of pio with no ram/prior tree in use, like less than 15 seconds to build the tree. However, after the tree is done, my system only shows ~55gb of my ram in use. That seems a bit odd. The estimated tree size of 80gb seems quite a bit off!

I know my cpu/core usage is capped, because I only have the basic version, which I intend to upgrade. But I would like to figure out these issues first, if possible. It shouldn't take so long to deconstruct the tree and then build a new one, its forcing me to close pio and open a new one because it's faster. And as I said before, the estimated tree size is giving me much larger numbers than the actual tree ends up being, about 40% larger.

Thanks for the frequent responses/awesome customer service You will have a new edge subscription very soon!

Hey,

I am trying to use the aggregation report function in piosolver. but when I open the report file it just shows all the day with comma's in the first column. Any idea how could I format the information properly so i could analyse the data?

thanks

Unfortunately this is normal as we use Windows memory allocator instead of a custom one. This doesn't matter much with small/medium sized trees but is painful with big ones. I recommend killing the solver (Solver -> disconnect) and then starting it again (Solver -> connect) to get around that (alternatively you can just restart the viewer as well as you mentioned)

The estimation is done assuming rainbow unpaired board. If you are solving suited ones or paired ones they will take less memory (and solve faster) due to various optimizations. This is done so maximum size of the tree is known for purposes of scripts (where there are many flops and some are likely to be rainbow unpaired ones).

This is a bug in Excel. Please refer to this post:
http://forumserver.twoplustwo.com/sh...postcount=1666

In aggregation reports, does it make sense to multiply raw equity by equity realization to get a more accurate or "true" value?

Very nice video, thanks a lot!

I was only solving one flop at the time (JcTs4s). So even if i dont intend to use a script it will overestimate anytime i build a tree around a paired and/or suited flop by about 40%?

It doesn't because the "true value" is just EV which is already displayed.
Equity realization is a curiosity but it doesn't have much practical value because the EV is already calculated by the solver.

Yes. This shouldn't be a problem though once you are aware of it. If you are running into RAM limits with the estimation there is a way to disable the check as well (it's highly recommended not to that). This is described at the bottom of this post:

https://forumserver.twoplustwo.com/s...postcount=2000

The reasons for this design are twofold:
1)In the earlier version of the solver we haven't had optimizations based on ismorphisms implemented so the trees took the same amount of space regardless of the board.
2)Showing true estimate would be problematic with scripts, it's easier to explain the way it is once someone notice some trees takes less memory than deal with "why the script wasn't able to build 5th tree, I tried estimating before I generated the script and the tree was within RAM limits".

I hope that makes sense

Can I suggest a new thread for questions regarding PLOCalc?