open % from CO and button

[danzasmack]

so therefore it is correct to spew like a madman

[rzk]

[ QUOTE ]
[ QUOTE ]
but i thought Q6o, J4s etc. are far from being very very marginal. am i wrong about this?


[/ QUOTE ]

Based on what?


[/ QUOTE ]

based on reviewing data in stox's book and my own data for higher hands, simulations, equity analysis. not saying i know exactly how -EV they are, that's why i was asking whether i was wrong.

btw 5% equity between Q9o and Q6o accounts for probably more than 5% of an effective pot size difference in winrate. the "high card/high kicker equity" is different from the "straight equity" in the sense that when a straight wins it wins substantially more on average.

fwiw, from my analysis it seems possible (though by no means certain) that the difference in winrate between Q9o and Q6o could be as high as 0.3bb/hand. i think it would be cool if you grouped hands like Q6o, Q7o, J4s and other hands in the bottom part of your range and have pt calculate your winrate with them when open-raising otb.

grouping them attempts to battle the sample size issue. the stdev when the hand sees the flop is around 3bb/hand. so if you have 400 hands in your sample, the stdev of the winrate will be 3/20 bb/hand=0.15bb/hand.

[Oink]

Ok so I checked

I took all the hands I usually open but that Stox doesnt open according to the chart on page 12 in his recent book. I also added the bottom hand of Stox range.


Hands
A3o-A2o, K9o-K5o, Q9o-Q7o, Q5s-Q2s, J9o-J7o, J6s-J3s, T8s-T5s, T8o, 98o-97o, 97s-95s, 87o, 86s-85s, 76o, 75s-74s, 65o, 65s-64s, 54s, 43s


Filter: OTB, chance to steal and raised, not a blind


EDIT: I missed some limits. These are the real numbers

db 1:
2199 trials with winrate of 0.19. 28 winners and 15 loosers

db 2:
1196 trials with winrate of 0.08. 25 winners and 18 loosers


Now go tell Stox he is a nit! [img]/images/graemlins/grin.gif[/img]

[Oink]

Ok so I did the same for CO.

Hands
A5o-A2o, K9o-K8o, K5s-K3s, QTo-Q9o, Q8s-Q5s, JTo-J9o, J8s-J7s, T9o, T8s-T7s, 97s, 87s-86s, 76s-75s, 65s

db 1:
1655 trials with winrate of 0.23.

db 2:
782 trials with winrate of 0.10.


This thread rules. I am gonna start raising even more!

[rzk]

wow, interesting. so your winrate with that range otb is 0.15bb/h +/- 0.05bb/h.

but i wonder what would happen if you just put in the bottom range, the trashier of the bunch, something like:

K7o-K5o, Q7o, J8o-J7o, J6s-J3s, T6s-T5s, T8o, 97o, 96s-95s, 87o, 86s-85s, 76o, 75s-74s, 65o, 64s, 54s, 43s

these are the hands that i always assumed were solidly in the -EV territory. but it looks like they might be much more playable than i imagined, at least when you play them [img]/images/graemlins/smile.gif[/img]

[Oink]

I actually did that as well. Dont remember the results but it was +EV for both CO and OTB.

I included the bottom hand of Stox range because that is supposed to be neutral EV.

BTW it seems as if my results are a lot better in db1 than db2. Funny since db1 is mostly 3/6 and 5/10 with a high rake. I dunno if its significant tho.

[rzk]

all right then, i'm convinced that i need to lag it up.

the differences between the db's might not be significant. the 1stdev intervals overlap so could just be a stat fluke.

[vmacosta]

[ QUOTE ]
all right then, i'm convinced that i need to lag it up.

the differences between the db's might not be significant. the 1stdev intervals overlap so could just be a stat fluke.

[/ QUOTE ]

rzk,
I am a significant loser with the second range you quoted. However, I also open some even weaker stuff like K3o, Q6o, etc. and when I add it all in I am basically breakeven. I think the conclusion is that my sample size is pitifully small and poker is just too much of a crapshoot to come to conclusions before millions and millions of hands.

[rzk]

[ QUOTE ]
[ QUOTE ]
all right then, i'm convinced that i need to lag it up.

the differences between the db's might not be significant. the 1stdev intervals overlap so could just be a stat fluke.

[/ QUOTE ]

rzk,
I am a significant loser with the second range you quoted. However, I also open some even weaker stuff like K3o, Q6o, etc. and when I add it all in I am basically breakeven. I think the conclusion is that my sample size is pitifully small and poker is just too much of a crapshoot to come to conclusions before millions and millions of hands.

[/ QUOTE ]

i think you can make some reasonable conclusions if you keep track of stdev of whatever variable you are trying to estimate. in oink's case, he has a sufficient number of hands for his first range for his positive EV to be statistically significant.

what's your sample size for the second range i quoted+K3o,Q6o etc.?

[Oink]

rzk

Any Idea how to find the st deviation from the filter?

I would guess the standard deviation over a sample fo hands where our vpip/pfr is 100/100 is a lot higher than our overall std.

The first results I quoted looks like I am winning about 15BB/100 over a little more than 4k hands. (From the top of my head)

For that to be significant different than 0BB/100 at a 98% level the st deviation needs to be less then about 80BB/100. (Again from the top of my head)


I suspect I am a significant winner with those hands over this sample altho its close.