Two Plus Two Publishing LLC Two Plus Two Publishing LLC
 

Go Back   Two Plus Two Poker Forums > Internet Poker > Commercial Software

Notices

Commercial Software Discussion of commercial gambling-related / poker software & commercial graphics modifications

Closed Thread
 
Thread Tools Display Modes
Old 09-15-2011, 12:52 AM   #1
Q
Pooh-Bah
 
Q's Avatar
 
Join Date: Sep 2007
Posts: 3,835
New software: ICMIZER - advanced online icm calculator (Windows & Mac)

I am introducing a new poker software - advanced ICM calculator for STT, MTT and HU SNG players ICMIZER.

Here is a short features list:

  • Calculates ICM EV correctly unlike most similar icm calculator software (taking into account results of tie headsup)
  • Provides ability to manually edit any hand range. Literally any range can be entered, unlike most similar software.
  • Analyze modes : Push, Call, Resteal (with possible limper)
  • Parse hands from poker clients (poker stars, fulltilt, cereus, party currently)
  • Available online in browser, works on both Windows and Mac
  • Two types of useful interactive charts (range charts and hand EV charts)
  • Ability to easily share results with others by means of permanent links
  • Free and fast basic ICM calculator with rich and responsive UI
  • Easy way to create new payout structures or blind structures


Here is a screenshot from it so you have an idea what I am presenting.


SNG Power Tools was one of the first poker ICM analyzing programs. It was very successfull and popular.

The program quickly got very widespread and became a main tool for players, who wanted to understand ICM better. Computing ICM for poker hand is not trivial and is a very laborious task. As I can see, even next generation ICM based tools developers had difficulties with computing ICM correctly, and it appears that they sometimes took results coming from SNG PT as “correct” and during their development compared their own results with SNG PT ones for the same situation. It would be fine of-course if SNG PT computed answers correctly… Since it was used as a reference by further developers, this mistake can be found in all other ICM based tools that followed SNG PT path, including for example famous SNG Wizard.

The mistake appears so noticeable that I can say, it is pretty amazing, that it wasn’t found by anyone in the course of over 5 years. I am the first person who noticed it and created the first tool that computes ICM based EVs for decisions correctly – ICMIZER, without repeating that old and stupid mistake.

Now about the mistake: as most of tournament players know – “we don’t need a coin flip on the bubble”. The price of losing is too high, and doubling up chip count doesn’t really increase our $ EV by far. Because of that, when number of players’ decreases, and especially at the moment when the next busted player becomes a bubble boy and gets nothing, ICM recommends to play very tight. But I have to say that very different situations can be seen as “coin flip”. This difference wasn’t considered in SNG PT, and it still is not considered by SNG Wizard in August 2011 (at the time I am writing this).

When we say a “coin flop” or 50%/50% one should still consider, what outcome the hand is preparing for us. A typical coin is a pair against a couple of overcards, especially suited overcards. For example 33 vs 45s, have equity that is very close to 50%. Here is the pokerstove result for such matchup:

Equity win tie pots won pots tied
33: 49.801% 48.36% 01.44% 19874556 591510.00
54s: 50.199% 48.76% 01.44% 20037720 591510.00

Notice that even though equity of hands is very close, there is a very little chance for a tie. In this case it is close to 3%. This means that either one player or the other will win the hand, and double up. This is the type of coin we don’t on the bubble in the SNG.
However another type of coin exists, that guarantees both sides exactly 50% equity. It happens when 2 players have the same hand.

Equity win tie pots won pots tied
AA: 50.000% 02.17% 47.83% 223260 4913652.00
AA: 50.000% 02.17% 47.83% 223260 4913652.00

As we can see equity of players is exactly 50%, and a total chance for a tie and a split is over 95%, so in such situation player 1 will lose to player 2 not in 50% of cases, but only in 2.17% cases, which is a very rare event! In the most other cases players will split the pot (usually big blind, small blind and antes).

Needless to say, situations are totally different, and we need to consider Tie odds, when we perform ICM calculations. This is what none of existing ICM programs on the market currently do, expect for the featured program – ICMIZER.

Next you will see an example of mistake, when other ICM tools compute EV in such situations (using SNG Wizard, and a single hand in SNG PT, because it appears that program has been abandoned by its developer).

The first example is the most explicit example of this mistake one can get. Let’s assume that our opponent has got two aces and pushed all-in with them on the bubble of usual 9 max SNG.

We hold Aces too, and here is the SNG PT suggestion for such situation, when blinds are quite high.



As we can see, our chance to win here is 50% (%Win column) and difference between a call and a fold according to SNG PT is -5.6% of total prize pool, so this is always a fold of aces.

But like I said before, in reality we lose here only in 2.17% cases, and taking that in account ICMIZER produces the following result for the same situation:



After computing result, and hovering mouse over icon with letter “i” against our opponent, we can see detailed data used for computation. As we can see ICMIZER is taking Tie chances into account and because of that deduces that all-in is better than a fold by 0.4%. Compared to result from SNG PT difference is 0.4%-(-5.6%) = 6% of total prize pool, which in 9 max SNG for 110$ would roughly mean 60$ difference, or over half of total buy in.

ICMIZER is cross platform program that is available to users on web. In order for it to work you need to install Microsoft© Silverlight plugin.

Since ICMIZER is a web application, it has big advantages when compared to traditional programs that need to be downloaded, installed and then reinstalled when an error is found or something is changed in poker client format and when update is available. For example, ICMIZER can save all parameters that were used when computation of decisions EV was performed like player stacks, payout structure, blinds and supposed opponents poker hand ranges. Then by means of a short link you can instantly share computed results with other poker players. For example to get result, that is equal to the one I got above you can simply follow this link: http://www.pokericmcalculator.com/icmizer/#wTNw and press [Compute AA] when page finishes loading.

Now the same error in SNG Wizard:


Usual column Win% is showing: 50%. I am wondering: where is the Tie%? I also have to mention, that unlike ICMIZER that provides total control over opponent range, SNG Wizard greatly limits user ability to modify opponent range. For example the ability to add or remove a single hand for opponent range is missing, along with ability to create a range containing random hands – for example ‘22,AK,89s” range is unavailable in SNG Wizard. Also it is super difficult to select a single hand for opponent; I was able to do that only for AA and KK. Besides SNG Wizard range slider is very slow and buggy, when compared to same slider for range editing in ICMIZER. All those limitations are caused by SNG Wizard attempt to simplify ICM computation task, by replacing real situation with various assumptions and hypotheses, and to allow it’s praised 3 way computations. For example in this case SNG Wizard says that it correctly computes expected values even for situations, when a player from blinds overcalls our push. Do I need to say that those simplifications lead to serious mistakes in EV calculations?

ICMIZER allows to manually setup any possible hand range to opponent, even a single hand. Also it has convenient sliders that can be used to quickly setup opponent ranges.

Here are computation results for such situation in ICMIZER: http://www.pokericmcalculator.com/icmizer/#xVRE



As we can see SNG Wizard mistake is: +0.9% - (-5.4%) = 6.3% of total prize pool.

But well, those results can be seen as pretty experimental, even though mistake is very high in above cases, in real poker we rarely end up in situation, when opponent range is narrowed down to a single hand. In real game we assign our opponent a possible range of hands, and based on that assumption we compute our range or EV of our hand for call, push and resteal situations.

When I first found this mistake, I wasn’t able to calculate its impact on typical cases (because of complexity of computation). But now, when ICMIZER is compete for Bet Test, I can say: influence of this error is very serious even in typical situations.

Let us consider a simple situation, we are facing a fairly tight shove on the bubble, and we are holding AKo.

Here is SNG Wizard take on this situation:


As we can see, Wizard thinks that difference between EV Push – EV fold is equal to -0.38% , and it’s an easy fold for us.

Now the SAME situation in ICMIZER: http://www.pokericmcalculator.com/icmizer/#qHoE



Suddenly, when tie % is taken into account, our answer for AKo is +0.1% already, and now we can consider the call, and given blinds size, even decide that it is time to shove. Anyway, now it is clearly not an obvious error, like SNG Wizard sees it. And I have to say – difference for a typical situation was nearly 0.5%, almost 5$ for a 110$ 9 max SNG!

It is often recommended to set value of Min Diff% (EV Push - EV Fold) to 0.1% - 0.3%. In this case error is over two times more, than typical edge value!

Okay last example, in this case we got AJo, SNG Wizard:



Call is an error worth -3.19% of total prize pool.

ICMIZER computes the result as follows: http://www.pokericmcalculator.com/icmizer/#hoDf



Now the wrong action costs us -2.5% of prize pool. When compared to -3.19% for same situation from Wizard its almost 0.7% of prize pool!

It has to be mentioned that currently ICMIZER is unable to perform calculations for 3 players unlike Wizard. It is closely related to exception difficulty of those calculations – they involve a lot of computations, and when performed precisely and correctly they will take pretty long time. I don’t think that it is correct to say that your program can perform 3 way ICM based computations, when in reality it is not computing for 2 way situations correctly and completely ignores tie situations. Because for 3 way computations situation becomes even more difficult, now two players can split pot, and the third one will be eliminated. Wizard completely ignores such tie situations too because it totally forgets that tie and split actually do exist in poker.

Any experienced SNG player on the other hand will say that he had experienced following situation at least some times: its bubble and we aren’t smallest stack, it’s a 3 way all-in against a big stack and a short stack, they both tie, and we are eliminated, on the bubble. It’s a disastrous outcome for a middle stacked player on bubble, and it is fairly strange to completely ignore this outcome. We cannot predict impact of such situations on our EV yet but it is obvious that in such case our EV will be bigger than it actually is in Wizard, while in other it can be smaller. Anyway as I have shown in example with 2 way computations, tie makes big change on resulting outcome Expected Values.

ICMIZER is in public beta phase now, and I am not ruling out situation, that after some time it will be able to correctly compute 3 way all-ins.

I hope you enjoy using it. I am using it and find it very convenient. If you need some help you can find additional info about features available here:http://www.pokericmcalculator.com/en-us/help/

One of main features of ICMIZER is ability to create links that lead directly to results of ICM calculation to a specifc situation. No need to take any screenshots and hosting them somewhere, and then manually entering stuff (which is quite a pain) afterwards in order to get same results. You saw some examples of this in this text, but here is another sample: http://www.pokericmcalculator.com/icmizer/#dgmD

If you like ICMIZER you are welcome to [+1] it on main page, since its currently totally free I am not asking for more.
I also recommend registerting and authorizing, in that case you will get access to more of ICMIZER features. benefits of registration

Q

Last edited by jukofyork; 10-06-2011 at 11:24 AM. Reason: Edited at request of OP.
Q is offline  
Old 09-15-2011, 12:58 AM   #2
Carpal \'Tunnel
 
jurrasstoil's Avatar
 
Join Date: Sep 2007
Location: SHEOL
Posts: 8,746
Re: ICM computations done right - first time in history (online and with Mac support)

This looks pretty sick.

OT: first time i have seen ms silverlight on a website
jurrasstoil is offline  
Old 09-15-2011, 01:07 AM   #3
Carpal \'Tunnel
 
kowboys4's Avatar
 
Join Date: Jul 2009
Posts: 9,723
Re: ICM computations done right - first time in history (online and with Mac support)

this looks really good. Thank you for this.
kowboys4 is offline  
Old 09-15-2011, 01:11 AM   #4
Q
Pooh-Bah
 
Q's Avatar
 
Join Date: Sep 2007
Posts: 3,835
Re: ICM computations done right - first time in history (online and with Mac support)

Thanks for kind words guys.

Another very useful feature I believe are charts. Here is a screenshot of range chart graph.

Notice that you can hover over chart and see your computed range for various EVs.



You can read more about them here : http://www.tenbb.com/en-us/help/#Charts

Last edited by Q; 09-15-2011 at 01:16 AM.
Q is offline  
Old 09-15-2011, 01:45 AM   #5
old hand
 
Join Date: May 2009
Posts: 1,903
Re: ICM computations done right - first time in history (online and with Mac support)

Sweet. I have thought of this but I always assumed SNGwiz accounted for ties......
paul rizzo is offline  
Old 09-15-2011, 01:49 AM   #6
Q
Pooh-Bah
 
Q's Avatar
 
Join Date: Sep 2007
Posts: 3,835
Re: ICM computations done right - first time in history (online and with Mac support)

Yeah well its pretty hard to spot in wiz because they made range input so limited
Q is offline  
Old 09-15-2011, 02:22 AM   #7
burnt out
 
Regret$'s Avatar
 
Join Date: Mar 2010
Posts: 9,388
Re: ICM computations done right - first time in history (online and with Mac support)

Thanks for putting this together. Nice to see the tools are coming along. Resteal stuff looks very nice.

What hand formats are supported?

To make sure I understand the error you have identified in previous software. They are accounting for ties by splitting preflop win% equity instead of computing the separate icm possibility?
Regret$ is offline  
Old 09-15-2011, 02:28 AM   #8
Blowin it all at HK
 
jdawg91's Avatar
 
Join Date: Oct 2009
Location: Batting cleanup TeamOI
Posts: 15,878
Re: ICM computations done right - first time in history (online and with Mac support)

Looks pretty sweet, already joined.

Thx for putting this together

Last edited by jdawg91; 09-15-2011 at 02:36 AM.
jdawg91 is offline  
Old 09-15-2011, 02:29 AM   #9
Q
Pooh-Bah
 
Q's Avatar
 
Join Date: Sep 2007
Posts: 3,835
Re: ICM computations done right - first time in history (online and with Mac support)

Regret$, yes you are correct, sample with aces is very clear I think. If you replace aces to 44 and 56s you will get very similar result in Wizard, while situations are completely different. PS: damn forgot that you cannot do that in Wizard. well you could do that in SNG PT. anyway you got my point, its ignoring tie and only considers equity.

To answer your question PokerStars, Fulltilt, Cereus and Party are supported, in sequence of stability. Its a public release so Ill be adding new formats, depending on what appear most priority.

I didnt put all the information here, but its all available in help. hand pasting and parsing
Q is offline  
Old 09-15-2011, 02:41 AM   #10
adept
 
Join Date: Dec 2007
Posts: 1,020
Re: ICM computations done right - first time in history (online and with Mac support)

Does this Program account for situations when Hero folds? Because this is very important.For example if we are at the bubble and are first in from the button our Push Range depends not only on the callingranges of the blinds but also on the Push Range from the SB and how wide the BB calls the SB Push. Is this integrated into the Program?

Last edited by Cpt.Hero; 09-15-2011 at 02:48 AM.
Cpt.Hero is offline  
Old 09-15-2011, 02:44 AM   #11
adept
 
CJSaunders's Avatar
 
Join Date: Apr 2008
Location: The real money games, but wait
Posts: 948
Re: ICM computations done right - first time in history (online and with Mac support)

Quote:
Originally Posted by Q.. View Post
The mistake appears so noticeable that I can say, it is pretty amazing, that it wasn’t found by anyone in the course of over 5 years. I am the first person who noticed it and created the first tool that computes ICM based EVs for decisions correctly – ICMIZER, without repeating that old and stupid mistake.
Brag : I had found it a while ago.
Beat : I thought the error was significant only when two very tight ranges clashed, I am surprised that the actual error in a realistic case is so large.

But to be fair I think Holdem Resource's Nash algorithm handles split pots correctly, or so its maker pretends in the FAQ. I never really checked since I thought this was not very significant. But anyway thank you very much for this.
CJSaunders is offline  
Old 09-15-2011, 02:55 AM   #12
Q
Pooh-Bah
 
Q's Avatar
 
Join Date: Sep 2007
Posts: 3,835
Re: ICM computations done right - first time in history (online and with Mac support)

Cpt.Hero I agree that its very important, and no current beta release doesn't include that.

Hey sng_jason: I looked into your program, looks like it was released earlier this summer. Well I saw this error in december 2010 but never had time to finish the program. Let me think about what I need to do about my statement. I dont think you presented description of this error like I did with samples etc? Or did you?

I like your 400mb size, gets me thinking Ill try to download it and see how quickly you solve 10 way push from utg with overcalls and manually edited ranges. Should be interesting.

CJSaunders: I also thought that result might be unimportant, when I saw actual results with real hands like AX i was pretty amazed. as for Holdem Resource Nash.. Did anyone ever took time to prove they are correct or not?

Last edited by Q; 09-15-2011 at 03:01 AM.
Q is offline  
Old 09-15-2011, 03:09 AM   #13
adept
 
CJSaunders's Avatar
 
Join Date: Apr 2008
Location: The real money games, but wait
Posts: 948
Re: ICM computations done right - first time in history (online and with Mac support)

Quote:
Originally Posted by Q.. View Post
as for Holdem Resource Nash.. Did anyone ever took time to prove they are correct or not?
I have written a very rough prototype of SNG Wizard like software, that includes Nash Equilibrium and the results I get are absolutely identical to Holdem Resources. I have given up on it for now because it is wayyyy too ambitious overall to make it into commercial grade software with all the features I had in mind, and the fact I prefer using my poker time playing rather than writing this. Also it seems that sng_jason is now attempting to do pretty much the same anyway, with probably much more skill at writing code than me.
CJSaunders is offline  
Old 09-15-2011, 05:00 AM   #14
Pooh-Bah
 
SkinnyStax's Avatar
 
Join Date: May 2010
Location: mom's basement
Posts: 3,608
Re: ICM computations done right - first time in history (online and with Mac support)

This is so ****ing awesome thank you!
SkinnyStax is offline  
Old 09-15-2011, 05:05 AM   #15
journeyman
 
cedr!c11's Avatar
 
Join Date: Jan 2010
Posts: 209
Re: ICM computations done right - first time in history (online and with Mac support)

Hi i tried paste a hand from PS.fr and got error:" Unable to parse provided text as valid poker hand". Its problem, that its in €?
cedr!c11 is offline  

Closed Thread
      

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off



All times are GMT -4. The time now is 05:19 PM.


Powered by vBulletin®
Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Content Relevant URLs by vBSEO 3.6.0 ©2011, Crawlability, Inc.
Copyright © 2008-2010, Two Plus Two Interactive