I'm not sure to be honest i just assumed. Might have been because you can only have 3 players max postflop, is that only for Beta?

The current betsizing UI doesn't make it possible to get into Postflop spots with more than 3 active players that are not all-in, that's why there's no 4+ options for now. You can only allow max 2 flat calls for any raise. Once there have been 2 flats for a raise, the next player would need to raise or fold.

The engine can handle 4+ way just fine though, so this limit may be removed once we re-work the betsizing UI. Obviously the trees get very large if there are more than 3 players still active by the Flop, so I'm not sure how useful this will be in practice.

Is there any way I can copy the results from the quick analyse grid of imported hands to excel? It looks like there should be as there all in columns and rows, but there doesn't appear to be any copy or export options that I'm aware of.

Also, it looks like there were one or two errors in the quick analysis results. How do I go about reporting these possible bugs/errors?

Is there any way of simulating shoving from the button shallow, where the blinds just flat call a percentage of the time? A number of my 'errors' have occurred because it assumes that the small blind either shoves or folds, but when I might have < 2 bbs on button, both blinds can flat call some of the time, so some of my fold might have been ok.

There's currently no export option for the QA results, but I'll put that on our todo list for the QA-rework.

If hands were simply not analyzed then select the hand and you will see additional information about the reason it was skipped in the lower left status bar. If you notice any other issues then please send me the hands & settings in question to support@holdemresources.net and I'll take a closer look.

Hello,

I was wondering about Monte Carlo calcs. Let's compare this example hand. First link is Basic calc and second Monte Carlo.

https://hands.holdemresources.net/?id=1r6p7x8bucilz

https://hands.holdemresources.net/?id=90jecnquf7rs

I don't understand why the results are so vastly different. Especially for Kx hands which don't even block Ax. If take a hand like K7s and get its equity vs that range from Power-Equilab (which also can use Monte Carlo) I only have ~43% and that's nowhere near a call with those pot odds. What am I missing

Results for HU should be identical for Math vs Monte Carlo. It looks like you manually changed that opening range?

For Monte Carlo mode you need to run new sampling after changing ranges, the EVs do NOT update automatically. Basically, the EVs in the calculation are still against the original range, not the 13.6% one. Use the green "Run Nash Calculation" from the toolbar, ideally with "Reset" checked in that case, that should fix it.

Ah ok, I'm laughing at myself now. I hadn't checked the Reset option while doing Monte Carlo. Thanks!

Thanks, hands were analysed, but incorrectly marked as errors I believe, I will send them in a few days as a bit occupied at the moment.

By the way, could you make that checked by default? Now every time when you do a recalc with modified ranges you need to check it again. Else HRC doesn't adjust to the ranges you change.

How to analyze spots with four people? I try to run it via advance monte carlo mode. It was a regular tournament, non ICM stage, these are my settings




and HRC gave me this output



What did I do wrong? How to analyze spots like this? I want to know what range SB can rejam in situation when one player limp, another min iso him, third did 3bet and hero on SB sitting with TT

And could you please make a hotkey or some button for quick select by edge? I tweak ranges every single time, so right now my routine looks like this, i click on a range-> Editor mode > Edit >Select by edge. Might be
there is such an option, but i haven't found it.

So, a few things to notice here:
1) You are using the regular stable version, this does NOT use any advanced Postflop realization. Pots are simply checked down Postflop if they are not all-in. You should never use limps/flat calls in this version without manually adjusting the limping/opening ranges.

2) As a conseqence of above, the EP only limps with a 0.1% range. When the line is entered this rarely, it will take a TON of sampling for the other ranges in this line to converge. That's why you get that weird looking EV grid, you simply don't have nearly enough sampling for a line like that. If you keep running the sim the limping range will likely go to 0% entirely.

As I mentioned above, if you want to run this in the old version then you absolutely need to manually override that limping range.(!)

Alternatively, check out the latest Beta version: This includes Postflop equity realization and you will get much better opening ranges for limping/flat calling scenarios automatically. You also need to run a lot more sampling than in a Preflop-only calculation though. I can post a quick run-through for that spot in the beta version if you are interested.

Edge Selection hotkey: I'll look into this.

Please post it, I'm really interested.

Alright, I'll post one after the weekend. It's a bit of an awkward line though, I wouldn't expect to see a lot of min iso's against EP limps in GTO, that's probably a spot where you should manually adjust ranges based on reads.

I can still explain the basics with that hand, but feel free to post a different example with a more standard line if you want.

Can I install Beta version without uninstalling the stable?

Just to double check, if I want to run this spot in the stable version, do I need to repeat all the steps that I did, plus add an additional step with setting limping range (only limping range?) or do I need to assign all the ranges manually?



I tried to add limping range manually and :

Advanced hand doesn't show SB action (as i understood it has limitation to maximum 3 players)

Advanced monte carlo - i can't open page with ranges for some reason, my next button is inactive




E.g. EP min raise, HJ flat call, CO Sqz and hero sittning on SB, which hands hero could rejam?

Do you plan to include a bubble factor display option in the software in the near future? I prefer HRC over ICMizer but I'm afraid that if you guys dont include that option I'll have to change to ICMIzer as it's a very important tool.

Yes, there's a separate installer:
https://www.holdemresources.net/blog.../#installation

In the old version: Simply calculate the hand with the Monte Carlo mode like you did before. After running the initial sampling, double click the limping range and edit/lock it to some range that you think is realistic for your games. Then use the green "Run Nash Calculation" button from the toolbar and check the "Reset" option, run additional samples until the ranges you are interested in converge.

A little tip to speed things up: After editing the EP range, select the HJ opening range and then use "Run Nash Calculation" with the "Update selected subtree" mode. This will only process samples if they actually fall in that selected line (EP limps, MP folds) and will vastly increase the processed samples / second. Keep running additional sampling (without Reset checked now) until the SB range converges.

Looks good, I'll post that tomorrow.

We already added that with the update from February 16th. Select the calculated hand and use Hand: Show Bubble Factors from the main menu.

Didn't notice the awkward stacks yesterday. EP probably just plays push/fold if you give him the option, but lets give it a go anyway. We are somewhat forcing the players into these weird lines by removing any other options. For instance CO barely squeezes here at all if we give him the option to flat call.

Make sure you are on the latest update via Help: Check for Updates.

I left the setup mostly unchanged, just using 40% commitment threshold to get rid of some redundant lines (e.g. we don't want SB to have a flat calling option against the 3b here). You can leave the Postflop abstraction on the default 256 buckets or even reduce them to 64/128, that's not going to change much for the line we are looking at.



After the initial calculation, select the green "Run Nash Calculation" icon and add ~50m additional iterations. We want to run this until the EP opening range remains somewhat stable. This runs at about 1 mio hands per second, so that's roughly a minute or so depending on your hardware. You should get something like below.



Now that the main ranges look reasonable, continue sampling the subtree you are interested in. We can select the HJ flatting range after EP's limp and run sampling for that part of the tree only. Note the "Update selected subtree" mode. The EP range will no longer be updated now. This runs at about 5 mio hands per sec.



Wait at the very least until the bold part of the SB range matches the +EV part. This is quite a few actions deep into the tree, so that's going to take a few minutes again.



Optionally you can just sample the CO 3-betting subtree to speed things up even more after the HJ range converges. It will take even longer until the call-off ranges against the 4bet converge. You need some patience with Monte Carlo mode if you look deep into the tree

plexiq thank you very much

How do you increase heap memory? I'm stuck at 4096. I tried increasing system jave heap size but that didn't make any changes in HRC. I'm on 64gigs of RAM but HRC gives me a limit of 4gb

Open the file hrc.ini from the installation directory in a text editor and search for "-Xmx", the default for the beta is -Xmx4G.

Save/load is still a couple of weeks away though, just in case you plan on running something that large.

Great, thanks a lot!

Hi Plexiq,
I have more of a theoretical kind of question.

Is there any information available regarding the evaluation of a) the ICM model and b) the bounty model you use in HRC? I read on the HRC blog, that a summary of the evaluation of the bounty models would be made available, but I couldnt find it.

Thank you very much!

Regarding an evaluation of FGS/ICM, you can find details in my thesis:
https://www.holdemresources.net/misc...ity_models.pdf

In short, this builds the entire push/fold tournament tree for a SNG and calculates the actual full-tournament strategies/equities. We used the same method later to evaluate 3 different bounty models, before settling on the proportional model.

The blog post didn't happen as there wasn't a lot of follow-up interest for the details, but the data should still be around somewhere. The gist of it is that the proportional model performed remarkably well and there really wasn't any reason to keep the alternative models around. I'm a bit swamped with the beta development work right now, but if there's interest I can still do a write-up of the bounty evaluation once things settle down a bit with the beta version.

Thank you very much!

Yeah, I assumed you are quite busy with the new beta currently.
If you find some time in the upcoming month I would appreciate it.
For now I am busy with your thesis I guess. Thanks again.