100BB, pocket 4 is better than pocket 8, micro stack?

XXPXXP

XXPXXP

Legend
Silver Level
Joined
Sep 7, 2012
Total posts
5,511
Awards
2
Chips
0
Just using some data
these are NL5, player s
results return from 44, and 88

have 150BB difference
1
2

but NL100

barely any difference
NL100 2
NL100

can anyone explain why 44 runs better results than 88.
thanks.

BTW: I made a java program to test the hand strength in EV

runs like 100big blinds

44 on average have +0.16Big blinds
88 on average have +0.27Big blinds

want to know why the results and assumption is not the same or might be anything wrong here?
 
vinnie

vinnie

Legend
Silver Level
Joined
Apr 12, 2013
Total posts
1,208
Awards
1
US
Chips
50
This is just variance over a really small sample size.

If you normalize these win-rates and use the standard deviation times to calculate what two time the standard deviation is in each direction, I am willing to bet both are well within range of each other.
 
Last edited:
vinnie

vinnie

Legend
Silver Level
Joined
Apr 12, 2013
Total posts
1,208
Awards
1
US
Chips
50
I mean,

88 hands you have 10,481.2 bb +/- 2,619.6 bb to cover two std deviations.
44 hands you have 7,508.8 bb +/- 2030.2 bb to cover two std deviations

Normalize the 44s to have the same number of hands
44 hands now have 11,752.5 bb +/- 2,619.6 bb
88 hands now have 10,481.2 bb +/- 2,619.6 bb

Now the 44 hands are ahead, for sure. But, when you look at the expected amount of swing that is typical for a 95% certainty, they are basically the same.

The end result of this whole discussion is the same conclusion that most of the real data geeks come to, our sample sizes are almost always far too small to say anything about results. We need tens of thousands of hands, if not hundreds of thousands or millions.
 
XXPXXP

XXPXXP

Legend
Silver Level
Joined
Sep 7, 2012
Total posts
5,511
Awards
2
Chips
0
I mean,

88 hands you have 10,481.2 bb +/- 2,619.6 bb to cover two std deviations.
44 hands you have 7,508.8 bb +/- 2030.2 bb to cover two std deviations

Normalize the 44s to have the same number of hands
44 hands now have 11,752.5 bb +/- 2,619.6 bb
88 hands now have 10,481.2 bb +/- 2,619.6 bb

Now the 44 hands are ahead, for sure. But, when you look at the expected amount of swing that is typical for a 95% certainty, they are basically the same.

The end result of this whole discussion is the same conclusion that most of the real data geeks come to, our sample sizes are almost always far too small to say anything about results. We need tens of thousands of hands, if not hundreds of thousands or millions.

300 run for a single hands is enough - because it is a single hands... it could be calling the long run.
besides these are from reg players. - the sample pick [I use the way for something like presidents election pick up pool methods <-- not sure if use correctly] is enough to cover the variance.
I use the allias for reg player, don't think the variance need to see this hug. - 150 BB/100 hands near 150/800=19% edge difference.
note:

44 wining rate 971BB/100 hands
88 wining rate 814BB/100 hands

besides I think I got an idea from one of my poker friends, want to know if any inspiration here .

thanks for the comment.:D
 
Last edited:
vinnie

vinnie

Legend
Silver Level
Joined
Apr 12, 2013
Total posts
1,208
Awards
1
US
Chips
50
300 run for a single hands is enough - because it is a single hands... it could be calling the long run.
besides these are from reg players. - the sample pick [I use the way for something like presidents election pick up pool methods <-- not sure if use correctly] is enough to cover the variance.
I use the allias for reg player, don't think the variance need to see this hug. - 150 BB/100 hands near 150/800=19% edge difference.
note:

44 wining rate 971BB/100 hands
88 wining rate 814BB/100 hands

besides I think I got an idea from one of my poker friends, want to know if any inspiration here .

thanks for the comment.:D

The sample size is nowhere near big enough. I know we're talking like a quarter of a million total hands to get this many samples of specific hands, but the numbers here are still not big enough and they fall well within the expected range of variance. I don't understand the rest of your comment about allias for reg players. And, what sample picking? You should have used every single hand of 44 and 88 available. It should have been a 100% sample size.

I highly recommend the book "The Mathematics of Poker" by Bill Chen and Jerrod Ankenman for a better understanding of how to calculate whether or not something is normal or within the expected range, when dealing with random outcomes like this.
 
vinnie

vinnie

Legend
Silver Level
Joined
Apr 12, 2013
Total posts
1,208
Awards
1
US
Chips
50
I've looked more into how HEM generates these numbers and it seems like the values are about 4 standard deviations apart. Which would be two standard deviations in each direction from the average. That means the results are completely typical and expected. Still, that book is important for understanding sample sizes required in poker. The numbers as astonishingly huge.

But, I also have a theory as to why 44 outperforms 88 at the lower levels, assuming that it does and this isn't just variance. At the higher levels, good players know that an unimproved 88 is about as good as an unimproved 44 when facing serious action. So, the better players lose less with 88 when they're beat. The weaker players will hold on to 88 longer when they're beat (or they call and give up to later action when more high cards come out). They won't hold on to 44 for as long on a bad flop. So, they lose very little when they don't hit a set with 44. Where, they might call a bet with 88 on a flop of 10-7-5, they are folding 44 on that same flop. The problem with calling with 88 on that flop (without a plan) is that you're going to have a lot of bad turn cards. If a Q comes on the turn, you might give up to a bet, or you might continue and find yourself paying off a better hand.
 
C

chronical

Legend
Silver Level
Joined
Mar 15, 2014
Total posts
1,128
Awards
1
Chips
1
if you've got 2 full table all-in callers you've got your answer(obviously this is not the case but a little bit more info would be nice). you've given a difference between 2 hands of ~80 blinds it can be covered by 2-3 fishy calls,they can happen on nl5
a1: how do your opponents play 44's and 88's. Do they shove 88 and fold 44?
a2: how do YOU play 44's and 88's. if you hold 88 stronger on flop/turn and flot river against folding any agression on 44...
a3: does that sample include som monsta wins were you clean up the table with quads or something like that?
 
Last edited:
Top