2007PBWC NLH FinalTable - Bot mistake? or misfortune?

[DevinLake]

Ok, because this dumbass thread is still going I'm going to bring some pretty simple poker related discussion to this "programed in a vacuum" poker bot discussion.

I'm not using ROR because no one uses it.

I'm not giving the villain a range of ATC for min raising, because I doubt the program is that stupid.

I'm going to give the villain a pretty wide range for min raising IMO. Although I didn't look at the HHs that were posted because they are a pain to read. So, if the bot isn't min 3-betting consistently, I have the min 3bet range of TT+, AJs+, AQo+. To re-iterate, I think this is wide.

Now, we know A9o is crushed by this range. Which means this hand is all about fold equity. So, lets look at the best case where the villain only calls with AA.

So, villain's 3bet range is 5.0%. Villain's calling range is 0.5%.

Hero's shove will get called only 10% of the time. So, 90% of the time Hero wins the 17 in the pot, that a cEV of 15.3 chips.

Of the 10% of the time he gets called, he has 6.2% equity in the pot. So, 6.2% of the time he's winning the 958.50 that the villain had to call the shove plus the 7 chips already in the pot that don't belong to the villain. The other 93.8% of the time he loses the 958.50-6 (his previous raise). So, that's a cEV of -83.32.

So, in total this shove is 15.3-83.32 = -68.02 cEV.

[DevinLake]

[ QUOTE ]
Slim,

The numbers were based on the widest possible hand range from Guaran. My point was that if it's a bad play against the widest range possible then it certainly can't be any better for what we'd consider a normal range for Guaran (regardless).

RIIT

[/ QUOTE ]

I'm not sure exactly what you are saying here. But, if villain is min 3betting ATC and calling the shove with ATC, shoving with A9o is good. It'd be +213 cEV.

[RIIT]

[ QUOTE ]
[ QUOTE ]
Slim,

The numbers were based on the widest possible hand range from Guaran. My point was that if it's a bad play against the widest range possible then it certainly can't be any better for what we'd consider a normal range for Guaran (regardless).

RIIT

[/ QUOTE ]

I'm not sure exactly what you are saying here. But, if villain is min 3betting ATC and calling the shove with ATC, shoving with A9o is good. It'd be +213 cEV.

[/ QUOTE ]

Devin,

2 random hands dealt to Guaran and HTC with Guaran getting the best puts Guaran on the top 2/3 range (HTC on the bottom 2/3) in which case the bots losing chances are close to 60%. My apologies for not explaining this sooner if this is the answer you were looking for.

If it had been headsup then yes I agree that the push looks good against ATC.

RIIT

[pokerdude@ub]

[ QUOTE ]
Ok, because this dumbass thread is still going I'm going to bring some pretty simple poker related discussion to this "programed in a vacuum" poker bot discussion.

I'm not using ROR because no one uses it.

I'm not giving the villain a range of ATC for min raising, because I doubt the program is that stupid.

I'm going to give the villain a pretty wide range for min raising IMO. Although I didn't look at the HHs that were posted because they are a pain to read. So, if the bot isn't min 3-betting consistently, I have the min 3bet range of TT+, AJs+, AQo+. To re-iterate, I think this is wide.

Now, we know A9o is crushed by this range. Which means this hand is all about fold equity. So, lets look at the best case where the villain only calls with AA.

So, villain's 3bet range is 5.0%. Villain's calling range is 0.5%.

Hero's shove will get called only 10% of the time. So, 90% of the time Hero wins the 17 in the pot, that a cEV of 15.3 chips.

Of the 10% of the time he gets called, he has 6.2% equity in the pot. So, 6.2% of the time he's winning the 958.50 that the villain had to call the shove plus the 7 chips already in the pot that don't belong to the villain. The other 93.8% of the time he loses the 958.50-6 (his previous raise). So, that's a cEV of -83.32.

So, in total this shove is 15.3-83.32 = -68.02 cEV.

[/ QUOTE ]

100% Dead-on

[DevinLake]

[ QUOTE ]
Devin,

2 random hands dealt to Guaran and HTC with Guaran getting the best puts Guaran on the top 2/3 range (HTC on the bottom 2/3) in which case the bots losing chances are close to 60%. My apologies for not explaining this sooner if this is the answer you were looking for.

If it had been headsup then yes I agree that the push looks good against ATC.

RIIT

[/ QUOTE ]

I'm usually pretty good at understanding things, but I don't have clue what you are talking about. One Random hand vs another random hand is 50-50.

If one guy has top 1/3 and other has bottom 2/3, they no longer have random hands.

[RIIT]

[ QUOTE ]
[ QUOTE ]
Devin,

2 random hands dealt to Guaran and HTC with Guaran getting the best puts Guaran on the top 2/3 range (HTC on the bottom 2/3) in which case the bots losing chances are close to 60%. My apologies for not explaining this sooner if this is the answer you were looking for.

If it had been headsup then yes I agree that the push looks good against ATC.

RIIT

[/ QUOTE ]

I'm usually pretty good at understanding things, but I don't have clue what you are talking about. One Random hand vs another random hand is 50-50.

If one guy has top 1/3 and other has bottom 2/3, they no longer have random hands.

[/ QUOTE ]

Devin,

We have 3 players:

AsbakAlpha (bot)
Guaran (man)
HTC (man)

There were a total of 3 hands dealt (not 2). All I'm attempting to do is to tell you how I arrived at my 57% numbers (I'm not trying to argue with you or challenge your analysis).

If we deal 2 random hands to 2 opponents and we consider the folded hand (HTC) to always be the weaker and the non-folded hand (Guaran) to always be the stronger then HTC's hand would be centered at 2/3 (effectively a range of bottom 2/3) and Guarans hand would be centered at 1/3 (effectively a range of top 2/3).

I never said I considered Guaran to have ATC. I said "widest possible range" and it was my fault for not describing my methods more clearly (again I apologize) and explaining what that phrase means to me in the context of 2 opponents. Any confusion caused was my fault and not yours.

And again, I agree with you that the push looks fine against ATC. And this might actually be the reason the bot made the play if it did indeed act as if it were a headsup situation.

Also, there was 19 in the pot (not 17) ... your bad [img]/images/graemlins/smile.gif[/img]

RIIT

[Slim Pickens]

Ummmm.... .no?

When HTC folds, all you learn about Guaran's hand is that contains 2 of the 48 remaining cards. Those 48 cards are slightly weighted toward the higher ranks because there are more foldable hands containing the lower ranks, but after a raise and a reraise, HTC should be folding all but the best hands anyway. You would also have to take out your ace and nine. All together, card removal is going to be a very meh correction to Gauran's hand range. This 1/3, 2/3 whatever business isn't based on anything.

The most useful information comes from the action and this is going to dwarf any other effects. Guaran's hand range goes from (almost) any two cards to something very specific after his min-reraise. My guess is that it's something like 80% {QQ+} and 20% {random crap}. This accounts for him bluffing some of the time.

OK, so if he's calling with only AA, that makes 0.8*3/15 = 0.16 he calls with AA. The 0.8 is my guess at 80% QQ+. 15 comes from the total combinations of QQ+ that can be made with the 4 Q's, 4 K's, and 3 A's left in the deck. 3 is the AA combinations possible. Anyways, so now you can go through what Devin demonstrated with 84% winning the pot uncontested and 16% being crushed for your stack and figure out if it's a good idea.

[RIIT]

[ QUOTE ]
Ummmm.... .no?

[/ QUOTE ]

no what?
- you disagree with my analysis methods?
- you disagree with my conclusion?

[ QUOTE ]
When HTC folds, all you learn about Guaran's hand is that contains 2 of the 48 remaining cards.

[/ QUOTE ]

I disagree. This statement isn't true until after Guaran shows his hand (which was not the case at the moment the bot was deciding to push). Now I realize this may seem like an insane statement to you but from a quantum mechanics style of analysis (which is what I prefer) where information is "uncertain" until measured, it is a perfectly rational statement. If you've not been exposed to this style of analysis then please keep an open mind before reacting.

[ QUOTE ]
Those 48 cards are slightly weighted toward the higher ranks because there are more foldable hands containing the lower ranks,

[/ QUOTE ]

Yes from the bots perspective this is accurate simply because of the non-available ace in it's hand.

[ QUOTE ]
but after a raise and a reraise, HTC should be folding all but the best hands anyway.

[/ QUOTE ]

Agreed.

[ QUOTE ]
You would also have to take out your ace and nine.

[/ QUOTE ]

Yes indeed, the tool I use does this.

[ QUOTE ]
All together, card removal is going to be a very meh correction to Gauran's hand range. This 1/3, 2/3 whatever business isn't based on anything.

[/ QUOTE ]

Well this is where I am going to disagree with you within the context of an infinite number of trials and random cards. I've chosen not to "guess" about what Gauran has or doesn't have. But instead I've chosen to use what MUST be true for an infinite number of trials with 2 opponent hands dealt.

Most if not all human analysis begins with an educated "guess" about the villain hand range (folding, calling, raising, etc.); our poker books are filled with this sort of thing and certainly 2+2 is as well. I'm not criticizing this method; I do it when I play live.

A qm style analysis puts Guaran on both the high and low range simultaneously (very much as if he is playing 2 hands and gets to use the better of the 2). The reason this isn't discussed in books is because no human can do this sort of calculation without a computer and one of the main purposes of a book is to help humans play well in live conditions.

[ QUOTE ]
The most useful information comes from the action and this is going to dwarf any other effects.

[/ QUOTE ]

Yes from a human analysis POV we all perform some type of bet interpretation (for better or worse). From a qm POV the action is meaningless insofar as card range is concerned and all action has one and only one effect - it changes the size of the pot.

[ QUOTE ]
Guaran's hand range goes from (almost) any two cards to something very specific after his min-reraise.

[/ QUOTE ]

Are you saying the rules do not allow Guran to re-raise unless he has the range you think he should have to do that? Or are you banking your entire analysis on the idea that your "guess" about his range must be 100% true and accurate?

[ QUOTE ]
My guess is that it's something like 80% {QQ+} and 20% {random crap}. This accounts for him bluffing some of the time.

[/ QUOTE ]

And personally, I think your "guess" is solid. But there is a way to look at the hand without first having to "guess".

[ QUOTE ]
OK, so if he's calling with only AA, that makes 0.8*3/15 = 0.16 he calls with AA. The 0.8 is my guess at 80% QQ+. 15 comes from the total combinations of QQ+ that can be made with the 4 Q's, 4 K's, and 3 A's left in the deck. 3 is the AA combinations possible. Anyways, so now you can go through what Devin demonstrated with 84% winning the pot uncontested and 16% being crushed for your stack and figure out if it's a good idea.

[/ QUOTE ]

If your guess is accurate then you've made your case very well.

Here are my numbers for AsbakAlpha:

winning chances: 0.400
losing chances: 0.568
tied chances: 0.031

pot: 19
call: 5
push: 947.50
call: 947.50 (from G.)

total pot: 1919
total pay: 952.50

ev: 1919*(0.400+0.031/2) - 952.50 = -154.06

RIIT

[DevinLake]

Ok, I get what u were saying before. Using the names screwed me up, I though u were talking about the two involved in the hand.

[ QUOTE ]
Also, there was 19 in the pot (not 17) ... your bad

[/ QUOTE ]

I believe there is 18? I thought villain was in the BB, but he's not, he's in the SB. So there is the 6 from the button raise, the 10 from the SB min raise and 2 from the BB. So, 18...

[DevinLake]

I think if this bot was programed correctly using your quantim mechanics, this hand demonstrates nicely that it is not a good method for programming a poker bot.