Two Plus Two Older Archives  

Go Back   Two Plus Two Older Archives > Tournament Poker > One-table Tournaments

Reply
 
Thread Tools Display Modes
  #11  
Old 09-07-2005, 04:42 PM
ZeroPointMachine ZeroPointMachine is offline
Senior Member
 
Join Date: Jul 2005
Location: Las Vegas
Posts: 136
Default Re: ICM/SNGPT rambling thoughts(long)

[ QUOTE ]
This is very incorrect. +EV is +EV is +EV. The one point that you hinge on that's correct is that a push could be +EV on a specific hand but then widen calling ranges on later hands, which would lessen your ability to push. However, as most correct bubble pushes are very blatantly +EV, given that you're not playing against a complete maniac, you're unlikely to run into a situation in which one close bubble decision closely follows another. Thus, slighty widening your opponents' calling ranges is probably a small enough factor to ignore.



Again, arguments that say that +EV plays are incorrect are inherently wrong by definition of EV, with the notable but hopefully atopical exception of gambler's ruin. Don't play above your bankroll, and squeeze as much EV out of your play as possible.

[/ QUOTE ]

So you say push every +0.01%$EV situation every time. Why wait till it's at least >0.5%?

This is the whole point of my post. These are not strict +EV individual trials (i.e. blackjack).
Reply With Quote
  #12  
Old 09-07-2005, 04:55 PM
Nicholasp27 Nicholasp27 is offline
Member
 
Join Date: Aug 2004
Posts: 93
Default Re: ICM/SNGPT rambling thoughts(long)

this isn't what we are talking about

not talking about random hand selections, but ones that are +ev according to ICM based on chip stacks, blinds, villian's calling range and your hand
Reply With Quote
  #13  
Old 09-07-2005, 05:05 PM
Guest
 
Posts: n/a
Default Re: ICM/SNGPT rambling thoughts(long)

Understood but villians calling range has a significant impact on +ev calculations. Assigning a calling range with any degree of accuracy to an unknown opponent based on play observation at a SNL is a crapshoot. What I was pointing out was even though everyone at the table clearly knew the pushes were complete BS hands they only called in effect 10-15% until they were forced in by their own chip stacks.
Reply With Quote
  #14  
Old 09-07-2005, 05:11 PM
ewing55 ewing55 is offline
Senior Member
 
Join Date: Feb 2005
Location: Tucson, AZ, USA
Posts: 126
Default Re: ICM/SNGPT rambling thoughts(long)

One thing I think that is being overlooked is that if you are pushing a lot (every +.0001%$EV) your opponents calling range is going to increase which will have an effect on the the future calculations. (ie. The more often I push, the more often I expect to be called %-wise and I will tighten up my pushing hand range.)

The opposite is also true. If I am being dealt crap cards on the bubble for a while, I'll push anything because I know everyone else is thinking "OMG, he plays soooo scared and if he's pushing he must have AA!" At times I can almost see my opponets looking at their JJ or AKs and still folding.

If your input of hand ranges is correct, you should always take any decent (+.05%%EV) situation.

Of course I could be wrong.

--------------Jeff
Reply With Quote
  #15  
Old 09-07-2005, 05:14 PM
ZeroPointMachine ZeroPointMachine is offline
Senior Member
 
Join Date: Jul 2005
Location: Las Vegas
Posts: 136
Default Re: ICM/SNGPT rambling thoughts(long)

[ QUOTE ]
[ QUOTE ]
Again, arguments that say that +EV plays are incorrect are inherently wrong by definition of EV, with the notable but hopefully atopical exception of gambler's ruin. Don't play above your bankroll, and squeeze as much EV out of your play as possible.

[/ QUOTE ]

What about situations where there is an expected greater +EV event in the near future, like the proverbial coinflip example in TPFAP?

[/ QUOTE ]

Exactly. If you push 10BB because it is +0.6% you risk losing and missing better opportunities later in the tourney. I know I've used the term risk of ruin kind of out of context. I'm not talking about bank roll risk of ruin in the normal context. I'm talking about losing a single tourney buyin risk of ruin.
Reply With Quote
  #16  
Old 09-07-2005, 05:46 PM
Slim Pickens Slim Pickens is offline
Senior Member
 
Join Date: Jan 2005
Location: Las Vegas, NV
Posts: 786
Default Re: ICM/SNGPT rambling thoughts(long)

One of the primary assumptions in the ICM is that all players are equally skilled. Obviously, we know this isn't exactly true. There are ways to correct for this, though I don't think they've been fully developed yet. By setting a necessary edge (0.5% or whatever), you're accounting for some skill difference with a guess, and depending on how skilled you are, your edge may need to be different than whatever eastbay picked.

The second thing you found is also a necessary condition of the ICM. There is no accounting for position or blind increases. Passing up a +$EV opportunity now may allow you to take an even better one later. Adjusting the cutoff based on stack size, future blind increases, and other factors is a good way to correct for this.

I did a simulation of taking random coin flips to build a stack. I found that 62% (I think) was the number I needed to finish ITM 40% of the time in a standard Party format. There's probably a way to back out a necessary $EV edge that implies.
Reply With Quote
  #17  
Old 09-07-2005, 06:04 PM
ZeroPointMachine ZeroPointMachine is offline
Senior Member
 
Join Date: Jul 2005
Location: Las Vegas
Posts: 136
Default Re: ICM/SNGPT rambling thoughts(long)

[ QUOTE ]
[ QUOTE ]

I offer you a dice game. You pay $10 to play. On a roll of 1-5 you win $3. On a roll of six you lose. Good bet? But you have to roll the dice 15 times before you can collect and if you roll a six you lose all your winnings and your $10 entry. Still a good bet? How many required rolls would make this a good game for you? Or if you have to roll the dice 15 times how high does the +EV on each roll have to be to make it a good game?

[/ QUOTE ]

This is not a good example. Sometimes you are called and win, which your example does not factor in. Sometimes you are called and lose, but you cover your opponent so you don't lose everything.

Part of the EV of making pushing things like K5o in the SB is that sometimes you will get called by A8o and suck out, your equity in the pot is not just FE.

Edit: of course, if you don't think that ICM accurately reflects the expected value of your chips, then obviously you can argue with the $EV numbers. But as long as you have a sufficient bankroll, while of course you will have downswings from when you lose showdowns from making +$EV decisions, in the long run you will make money.

[/ QUOTE ]

Yeah, I admit the example over simplifies things. The point was that push/fold decisions are not independent trials.

Say you have a huge series of independant trials. You divide this series into groups of any size. The sum of the results for each group will be equal to the sum of the whole series.

SNG pushes don't work this way. Because one negative result can erase several previous positive ones. The size of the groups matters.

Nobody advocates pushing every +0.1%$EV hand. Why? Because there are so many. You would be involved in so many hands with a small advantage that you almost guarantee losing one of them.

Everybody accepts that +0.5% is good enough to push. But why?

Every tourney involves a string of +$EV push/fold situations. I am trying to quantify the length of the string.

If you have X BB you are likely to see Y number of +$EV situations. How high does +$EV need to be to belong in the string?
Reply With Quote
  #18  
Old 09-07-2005, 06:12 PM
Slim Pickens Slim Pickens is offline
Senior Member
 
Join Date: Jan 2005
Location: Las Vegas, NV
Posts: 786
Default Re: ICM/SNGPT rambling thoughts(long)

I thought the whole point of the ICM was to turn chip EV into $EV so that the hands could be treated as independent trails. You are right in that SNG pushes don't work this way, but I think that's only if you use cEV instead of $EV. Of course, all of this is limited to the validity of the ICM's assumptions, but that's where the necessary $EV edge comes in, and why SNGPT spits out "both are good" for close decisions.
Reply With Quote
  #19  
Old 09-07-2005, 06:51 PM
BadMongo BadMongo is offline
Senior Member
 
Join Date: Jul 2005
Location: swimming with the brown trout
Posts: 190
Default Re: ICM/SNGPT rambling thoughts(long)

The only way I can see folding a +EV push being correct is if one of the assumptions of the model doesn't approximate the situation satisfactorly. So let's look at these assumptions:

1) Skill level.

ICM assumes all players are of equal skill. This is a big assumption, and is rarely applicable in a literal sense. Most of the time (hopefully) we will be more skilled than the people we play against. Thus, passing on a slightly +EV situation can be correct since we are minimizing our chances of going broke and affording our opponents more oppertunities to make mistakes. This can result in greater +EV situations later on. Although ICM can't directly quantify this effect, we can approximate it by using rules like eastbay's >0.5%.

2) Position and blind size.

ICM ignores who will be in the blinds on later hands, and when the blinds are going up. As a result, calculations will overstate your EV of pushing the further you are from the blinds, and it will understate it the closer you are. Similarly, if the blinds are increasing soon, ICM will understate the EV of a push. These factors only really become a problem when there are one or more very short stacks (i.e. only a couple BBs or less). If there are no very short stacks, this effect is negligible and can be ignored. Even if there are very short stacks, you can estimate your actual EV by discounting your equity appropriately.

3) Hand Ranges.

Well, you have to put your opponent(s) on a range to calculate your EV, and this is where most of the debate arises. Part of the difficulty of doing this somes from the fact that hand ranges are dynamic. As the game progresses, hand ranges change based on many factors like blind size, push frequency, tilting opponents, etc. Most of the time it's impossible to know your opponents' true range in a particular situation, but this is where skill and experience come in. Your read of an opponent's hand range is just part of the data ICM uses to tell you your EV. If your assumption about an opponent's range doesn't approximate the situation satisfactorly, that isn't a problem with the model, it's a problem with your reads.

I think if you're going to say "well, ICM says this is +EV, but you should fold anyway", you need to justify your position by showing which of these assumptions has been violated, and why it might in fact be -EV to push even though ICM says otherwise.
Reply With Quote
  #20  
Old 09-07-2005, 07:59 PM
ZeroPointMachine ZeroPointMachine is offline
Senior Member
 
Join Date: Jul 2005
Location: Las Vegas
Posts: 136
Default Re: ICM/SNGPT rambling thoughts(long)

Thanks for the well thought out post.

[ QUOTE ]
Thus, passing on a slightly +EV situation can be correct since we are minimizing our chances of going broke and affording our opponents more oppertunities to make mistakes. This can result in greater +EV situations later on. Although ICM can't directly quantify this effect, we can approximate it by using rules like eastbay's >0.5%.



[/ QUOTE ]

Don't misunderstand this post as an "ICM doesn't work and that's why I can't win" post. ICM has been a tremendous tool and I have been very successful using it. I have been playing at the 20s much longer than my bankroll dictates and have a 25% ROI over 1000 games.

At this level I feel that the factor you listed above is much greater (on average) than the 0.5% fudge factor. This is especially true when you are on the high end of the 10BB range and are risking elimination. I think many people know this intuitivly and don't go into "pure ICM mode" until they are below 8BB. At the higher buy-ins, as the skill levels become closer, this discrepancy probably diminishes.

This is what I am trying to quantify.

I know it will be buy-in dependant. But I think some analysis of how many ICM pushes are made in a typical SNG and the chance of being eliminated somewhere along the way should yield a better guideline than +0.5% is goot enough.

I believe a sliding scale based on stacksize could be developed.

Wish my programming skills weren't 15 years out of date.
Reply With Quote
Reply

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

Forum Jump


All times are GMT -4. The time now is 11:12 AM.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.