Weird HPFAP blind defense strategy inconsistencies

[Nigel]

[ QUOTE ]
Andrew Prock already knows all this stuff exactly, I guarantee it.

Edit - no, he isn't going to help you. E.g. here and here and here.

[/ QUOTE ]

Trying to solve this is less complex than trying to figure out what Andrew is ever saying. [img]/images/graemlins/wink.gif[/img]

When's his book coming out?

Hi Nigel,

Great post.

First off some data:

In 130k hands of 5-10 (6 max), the BB folded to my open-raise in the SB 25.6% of the time.

In 40k hands of 10-20 (6 max), it was 22%.

Work that people are doing on developing optimal open-raising ranges in the SB in, say, a 1-2 blind structure game will have to take BB looseness as a parameter (as has already been stated) to be useful. Against a BB that never folds (which hopefully happens a lot if your game selection is good), raising any two would clearly be bad.

My ideal way to tackle this problem would be to look through a 20 million hand composite DB of very successful players and actually see how the various hands did, then form rough guidelines of what hands to open raise with. However, this will assume an average looseness for the BB as you can't filter for tight or loosely defending big blinds. It will at least be a good start that we can modify incrementally to suit the particular oppoenent.

Using a mathematical or simulation approach to derive optimal raising standards may have the benefit of incorporating assumptions of the big blind in the model, but it suffers from not being a 'real world model', i.e. making assumptions that add inaccuracies that dwarf the numbers you end up getting for your result.

Ditto for defending in the big blind against a SB raise.

My own story is that I came to the same "realiziation" months ago that proper play in the blinds requires aggressive stealing and aggressive defending. Since implementing changes in my play, my blinds stats have taken a nosedive. The problem is, I could be running bad.

Someone may come out with a "theoretical guideline" for what raising and defending standards should be, but until it's been tested over millions of hands, we won't be sure that the assumptions they made were or weren't valid.

Sorry for not making a very constructive post. My contribution will be to suggest that maybe we can form a composite database of heads up blind hands, and maybe we can analyze this data and come to some meaningful conclusion.

-v

[The Truth]

Hey buddy, you mentioned "Abdul's work" could you point me toward that?


-blake

[Nigel]

[ QUOTE ]
Hey buddy, you mentioned "Abdul's work" could you point me toward that?


-blake

[/ QUOTE ]

Blake for most of his writings you have to sort through the hell that is rgp. You'll want to search groups by author:abdul author:jalib for his stuff, then add additional parameters.

For archives of his posev site, go to:

here

Nigel

[baronzeus]

this post made me realize i [censored] suck at blind defense. thank you.

[TStoneMBD]

Nigel youve really done an excellent job of inspiring me to work on my blind defense in HU situations. Unfortunately, I don't know where to start.

[Nigel]

vkh,

Thanks for the reply. A couple of quick things:

Interesting stats, you are getting called a little more than I am, but I imagine that has to do with changes in the 5/10 and 10/20 over time, or perhaps just sample size. Oddly enough, in all my databases, it seems that the BB is MOST likely to fold with position to the SB raise, then the CO, and lastly the button as I am seeing some crazy numbers when looking at my button raises - I am being called near 85%. Seems odd.

Going back to this approximately 25% number we are seeing for the BB in response to an SB open-raise, we are flirting quite closely with the 30% number that Sklanksy postulates (correctly or incorrectly?) allows us to raise any 2 cards for a profit. If we can raise any 2 from the SB against this opponent, how about from the CO or button with position?

Anyway, it's interesting that since you have made these adjustments, your stats have headed in the wrong direction. Let's hope it's just bad luck and not bad theory. If not, I may soon start to suffer the same fate.

I need to set-up Postgres and build one huge PT database as I imagine I alone have several million hands total stored. As you suggest, there is a lot of work to be done.

Nigel

[StellarWind]

[ QUOTE ]
Maybe the BB being out of position is the reason for the discrepancy, but I don't see how position changes the mathematics behind the SB showing an automatic profit from the BB folding too much pre-flop.

[/ QUOTE ]
It is unquestionably true. If you fold the BB very often it becomes profitable for the SB to raise any two cards. He should never fold.

What is far from theoretically obvious is whether there is anything wrong with that. Even if you know that his raise represents any two cards it may be true that defending with a trash hand is -EV due to implied odds. In that case you need to fold regardless of how much money SB may be making.

Indeed, worrying about whether SB is making an automatic profit by raising any two cards is irrational. Whether he is making an automatic profit depends heavily on the size of his blind payment because it determines his pot odds. But once I tell you that he is definitely raising any two cards the size of his blind payment cannot matter to you. It does not affect your EV. The EV of his raise for him is not your problem. Only your EV should matter to you.

If the BB has position postflop it is reasonable to believe almost any hand will play well enough against two random cards that it should be +EV to call. Position is a big advantage and even the worst hands have a lot of pot equity preflop versus a random hand.

Heads up the SB has the button and that changes everything. If you play a bad hand against a good player you are going to lose a lot of money postflop. Simply cannot be avoided when your opponent has both position and better average cards. The only question is whether the money in the preflop pot will compensate you for this.

This is not an easy question to answer which is why people disagree about correct blind defense strategy. But the "automatic profit" argument is not a valid shortcut for solving the problem. It assumes that SB is not entitled to make a profit by raising any two cards and there is no obvious reason to assume that when he has the button.

[bobbyi]

This sounds reasonble to me, although I have very little experience with 2/3 blinds and it is all in lives games where I usually chop.

About this part:
[ QUOTE ]
against a SB openraise from a guy with 8+PFR i call with 67o/68o and almost any 2 suited.

[/ QUOTE ]
What are your thoughts on calling with offsuit hands with one high card like K3o and Q6o? I tend to like the HPFAP advice to avoid hands with 2's and 3's, but I would call with maybe K5o and Q7o. Is that reasonable?

Nigel,

Thank you for replying.

When I get home I will look at HPFAP and see what information is given on the derivation for the '30%' number.

For now, some thoughts...

When the SB open-raises in a 1-2 blind structure, he is investing .75 BB to win .75 BB. Thus, if the big blind folded over 50% of his hands on average, the SB would show an automatic profit by raising any two cards; let's take 32o as a representative hand for 'any two cards.'

Since we will win some of the times we get called, even with 32o, we don't need the BB to fold 50% of the time. We can accept a lower number. Sklansky says 30%. It seems like the real world fold % we are seeing in our DBs is close enough to 30% that it's not obvious if raising any two is correct. What if Sklansky's 30% is really 25%? Maybe it's 35%. It also depends on the skill level difference between you and the BB. If the big blind plays much worse than you post-flop, you may only need him to fold 25% of his hands to show a profit raising 32o, whereas against a great post-flop player, you may need him to fold 35% of his hands for 32o to show a profit. So we can talk about an 'average fold %' for the 'typical' opponent.

I think with so much variability and so many different factors involved, the 'huge DB approach' is the most reliable way to answer the question of what hands to steal with.

Setting up one huge PT database would be great. I would be willing to send you my DB, and I can probably get at least one other DB of a friend who wins at 10-20 6 max. You can PM me if you need any help with this endeavor.

-v