Two Plus Two Newer Archives  

Go Back   Two Plus Two Newer Archives > Other Poker > Heads Up Poker

Reply
 
Thread Tools Display Modes
  #1  
Old 09-01-2007, 04:55 PM
TNixon TNixon is offline
Senior Member
 
Join Date: Apr 2007
Posts: 616
Default Finally, a useful conclusion drawn in the variance thread! NL

Leader, delete this if you feel it's wholly inappropriate, but there's been so much useless argument going on in there, that I feel there are likely to be a lot of people who have given up on it entirely.

But this post:

http://forumserver.twoplustwo.com/sh...age=0&vc=1

Contains what I believe are very useful (and at least partially surprising) conclusions, and I've invited anybody who is willing from the probability forum to come look and comment (with a specific invitation to one of the mods over there who helped me get the final duck in line with the others).

It is very long and verbose, but anybody who has been following the thread will immediately understand what it was necessary to be as complete as possible. [img]/images/graemlins/smile.gif[/img]

Summary of the conclusions:

It seems highly likely, given the evidence so far, that cash games are *lower* variance than HUSNGs at equivalent buyins (meaning that playing $100NL would be lower variance than playing $100 HUSNGs).

Which would make perfect sense if a typical "good" winrate at HUCASH is 10BB/hundred hands, because you can make a much better hourly rate than that playing SNGs. In this case, the tradeoff would be lower hourly rate for a lower variance.

I would especially like to invite omgwtfnoway and jay_shark to read and comment, although I would completely understand if they want to be disassociated from the topic entirely.
Reply With Quote
  #2  
Old 09-01-2007, 05:42 PM
Guruman Guruman is offline
Senior Member
 
Join Date: Jan 2005
Location: still a NL fish - so lay off!
Posts: 3,704
Default Re: Finally, a useful conclusion drawn in the variance thread!

your linky no worky nix.
Reply With Quote
  #3  
Old 09-01-2007, 05:48 PM
TNixon TNixon is offline
Senior Member
 
Join Date: Apr 2007
Posts: 616
Default Re: Finally, a useful conclusion drawn in the variance thread!

Doh. It's in the "re variance blowup" thread. I'll try to get a working link. [img]/images/graemlins/smile.gif[/img]
Reply With Quote
  #4  
Old 09-01-2007, 05:49 PM
TNixon TNixon is offline
Senior Member
 
Join Date: Apr 2007
Posts: 616
Default Re: Finally, a useful conclusion drawn in the variance thread!

I can't edit, try this one:

http://forumserver.twoplustwo.com/sh...age=0&vc=1
Reply With Quote
  #5  
Old 09-02-2007, 06:33 PM
jay_shark jay_shark is offline
Senior Member
 
Join Date: Sep 2006
Posts: 2,277
Default Re: Finally, a useful conclusion drawn in the variance thread!

So do you agree with me ?
Reply With Quote
  #6  
Old 09-02-2007, 10:00 PM
Leader Leader is offline
Senior Member
 
Join Date: Jan 2005
Location: Excellence: Learn, Play, Win.
Posts: 7,682
Default Re: Finally, a useful conclusion drawn in the variance thread!

[img]/images/graemlins/smirk.gif[/img] last thread about this kthx
Reply With Quote
  #7  
Old 09-03-2007, 12:56 AM
jay_shark jay_shark is offline
Senior Member
 
Join Date: Sep 2006
Posts: 2,277
Default Re: Finally, a useful conclusion drawn in the variance thread!

Tnixon , it looks like you've smartened up [img]/images/graemlins/smile.gif[/img]

Your long term variance in an SNG is a function of your win rate . Since your buy-in is always constant (hypothetical assumption) your variance or standard deviation is going to be constant . This means that two players who have the same win rate , will share the same variance . Also , variance in an SNG is expressed in terms of buy-ins .
+1 = one win
-1 = one loss


In a cash game , your variance need not be the same since the size of the pot and stacks may vary . One thing is certain which is that your $/h will be greater playing in a cash game . This leads us to believe that your variance should be lower . However it's also true that the deeper you're stacked , the higher your variance becomes . At some point , the game will become higher variance than an SNG .

I will repeat this again . A $100 sng is higher variance than a cash game with a $100 buy in and blinds at 4/3 ,2/3 . I've chosen those specific numbers to illustrate my point since your buy in is 75* the BB amount which is consistent with an sng starting with 1500 chips at 10-20 .
Reply With Quote
  #8  
Old 09-03-2007, 01:20 AM
xSCWx xSCWx is offline
Senior Member
 
Join Date: Oct 2006
Location: Texas A&M / Teaching HU SNGs
Posts: 1,776
Default Re: Finally, a useful conclusion drawn in the variance thread!

3 $33 tables definitely has less variance than 1 $100 buy in cash game.
Reply With Quote
  #9  
Old 09-03-2007, 04:15 AM
TNixon TNixon is offline
Senior Member
 
Join Date: Apr 2007
Posts: 616
Default Re: Finally, a useful conclusion drawn in the variance thread!

[ QUOTE ]
Tnixon , it looks like you've smartened up

[/ QUOTE ]
How so? There's nothing in that post that I hadn't already tried to say a million times by appealing to logic rather than laying out the math. The only difference is that I finally added the math to prove it. (and finally understood the math well enough to lay it out) So in a way, I did smarten up, but not in the way you seem to be suggesting. If you were in disagreement with me before, then you should still be in disagreement, because there's not a single point on which I've changed my position. In fact, I did specifically say a number of times that any amount of confusion that was going on would be cleared up instantly if you guys simply did the variance calculations in dollars instead of big blinds, which would have immediately pointed out that there really was a unit conversion problem going on when trying to directly compare variance calculation results for the $100 10BB stack against the $100 100BB stack.

If you believe we are now in agreement, then either you have changed your position, or you did not understand what I was attempting to say before. Which would be exceedingly surprising, since I tried about a million and a half different ways of saying the same thing.

But do you at least understand now why you can't "simplify" the 10BB vs 100BB problem by shuffling around the values of the blinds, and saying that a $10 10BB stack is the same as a $100 10BB stack when comparing to a $100 100BB stack, just to make it easier to figure out what's going on? That the simplification changes the results so drastically that it's impossible to draw any conclusions?

[ QUOTE ]
One thing is certain which is that your $/h will be greater playing in a cash game

[/ QUOTE ]
I'm not sure that's really at all certain. Depending on whether 'h' means hand or hour in this context, it could be very difficult to compare, but I've heard a 10bb/100h figure being tossed about as a "good" winrate at cash, which would lead to somewhere around $20/hr at a .5/1 cash game, buying in for $100. I don't know if 10bb/100 hands is actually good or not, but that's what's been tossed about.

The "good winrate" number that gets thrown around for husngs is 60%, which can very easily lead to making a buyin an hour single-tabling turbos, or $100 an hour playing the same $100 at sit-n-gos instead of cash.

If by "h" you meant "hand", then the comparison is a bit more difficult to draw, but I'm still not sure it's all that certain. On full tilt turbos, blinds reach 100/200 after 24 minutes. It's *very* rare for a game to last this long (I think it's happened to me maybe twice), but even if it does last this long, there have probably only been somewhere around 80 hands. Which, with a 60% winrate, puts you somewhere around $0.25/hand at a $100 buyin level (yes, I know there is no $100 buyin level for turbos, that they only come in the $110 variety, but lets not make things more difficult than they have to be), compared to the $0.10/hand of a 10BB/hundred cash winrate at $100NL (I really wish I had any idea of whether this was actually a decent winrate. Somehow I don't think 10bb/hundred indicates anywhere near the edge that a 60% winrate at sngs does, but I don't know what else to use). And I believe this $0.25/hand can be viewed as a minimum. It will be much higher for shorter games (which most are).

[ QUOTE ]
One thing is certain which is that your $/h will be greater playing in a cash game . This leads us to believe that your variance should be lower .

[/ QUOTE ]
Explain please? I'm not sure I understand how $/h has any bearing on the variance at all. Almost by definition, the only things that can possibly affect variance are pot sizes (both average and real pot sizes) and how often you win those pots. It is certainly possible for 2 play styles to make the same average $/h, but have vastly different variance figures.

[ QUOTE ]
A $100 sng is higher variance than a cash game with a $100 buy in and blinds at 4/3 ,2/3 .

[/ QUOTE ]
I don't understand this sentence, unless by 4/3 and 2/3 you mean blinds of $1.33 and $.66. If that's what you mean, then yes, the sng would likely be higher variance, because the blinds in the sng increase, but the blinds in the cash game stay the same.

And I believe it would be safe to say that $100 .5/1 blinds would be lower variance than $100 .66/1.33 blinds, even though you are technically "deeper" stacked, in the same way, and for the same reasons, that the math appeared to show that $100 .5/1 blinds actually has the potential to be lower variance than playing $10 at the same blind levels.

(I say "appeared", because I didn't actually check stacksizes at the tables, just the average pot size. For all I know, many of those tables could have had small effective stacks)

You're potentially risking the same amount ($100), but since average pots are generally built up in increments of big blinds, the average pot (in $), would likely be lower at .5/1 blinds.

[ QUOTE ]
I will repeat this again . A $100 sng is higher variance than a cash game with a $100 buy in and blinds at 4/3 ,2/3 .

[/ QUOTE ]
You do realize that this is the point I was trying to make all along, right? That unless there are other factors that only become apparent in cash games (and there are certainly possibilities), that SNGs should be higher variance than cash games? And that this is exactly why it was so important to clarify the 10BB vs 100BB issue that blew the thread out into a massive flamefest in the first place?

In fact, the whole point of the original post was to try to get a discussion going about whether conventional wisdom (which says that cash games are higher variance than sngs) is correct, and if so, why, because there don't really seem to be any factors in favor of sngs being lower variance.

Originally, I suspected that they should be pretty close to the same, with factors in one direction being offset by contrasting factors in the other. (for example, in sngs, you're never playing for a double stack, but in cash tables, the blinds don't increase)

After everything that's been said, I'm fairly convinced that cash play is actually *lower* variance than SNGs, which is exactly opposite from "common knowledge".

[ QUOTE ]
3 $33 tables definitely has less variance than 1 $100 buy in cash game.

[/ QUOTE ]
Yes...

Which would be important if we were trying to compare $100 cash sessions to $33 sng tables.

Since we're actually trying to compare $100 cash sessions to $100 sng tables, you're going to have to help me out here and explain what you meant. [img]/images/graemlins/smile.gif[/img]
Reply With Quote
  #10  
Old 09-03-2007, 04:38 AM
TNixon TNixon is offline
Senior Member
 
Join Date: Apr 2007
Posts: 616
Default Re: Finally, a useful conclusion drawn in the variance thread!

A REQUEST FOR STATISTICS

It would be exceedingly useful to get some real-world data, comparing $100NL cash sessions against $110 sngs, from a wide variety of people, especially from anybody who plays both forms fairly frequently, but even just one side or the other would be useful.

The necessary data would be:

For both cash and HUSNGs:
Some general indication of your playstyle. More agressive players are going to tend to show higher variance than conservative ones. Maybe a simple 1-5 rating on looseness, and one for agressiveness? Obviously these change based on opponent, but an indication of your "default" mode would probably be useful.

For cash play:
Number of hands, winrate, and standard deviation. (I believe pokertracker can very easily give you all this stuff, including the std. dev).

$100NL would be the most useful, but if you provide data for multiple limits, please separate the numbers by limit.

Another very useful stat would be average number of hands per hour, but I don't know how you'd get that, unless pokertracker will also tell you how much time you spent playing. [img]/images/graemlins/smile.gif[/img]

For HUSNGs:

This is going to be simpler for the most part, but a lot tougher on one specific point, because there's one very important factor that I don't know if there's any way to get at easily: average tourney length, or average number of hands per tourney.

Other than that, a simple win/loss record is all that's needed. Ideally just $110s.

If there's no way to get that out of pokertracker, then there is something I can do, but it would only work for people geek-savvy enough to be comfortable installing perl, and running a perl script on their hand histories.

Alternatively, I could throw together an .exe that would do the same thing, but you really shouldn't trust executables from people you don't know, especially where your poker bankroll is concerned. Who knows, I could have thrown in a keylogger for your password, or a trojan that sends me your holecards if we're ever playing each other. The perl script would be simple enough that it would only take a very rudimentary understanding of programming in order to verify that I'm not an evil bastard after all. [img]/images/graemlins/smile.gif[/img]

This data would be used to show real-world variance comparisons between cash play and sngs. Stats would be listed per individual, but no names would be used.

If you would be willing to contribute, please PM me.

Edit---

UGH!

I could have sworn I saw std. dev shown on the graphs that everybody posts in the low-content thread. But apparently I was mistaken, as none of the graphs I've seen mentioned it.

Is it possible to get any indication of variance out of pokertracker? This is another thing I could get out of hand histories with a perl script if I had to, but with all the same problems as before. (you'd have to be willing and able to install perl and run the script)
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 04:46 PM.


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