Open Side Menu Go to the Top
Register
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc"

11-25-2012 , 10:06 AM
meh, i think we should just have a duel: durrrr vs. the ike/kanu/sauce trio. surely those three working in unison could qualify as a GTO bot, no? they would have to attempt to play as non exploitative as possible, and make their decisions strictly based off of game theory
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 10:22 AM
Quote:
Originally Posted by bellatrix
6tabling HULHE is not really doable, imo
Durrrr was 3tabling HU Triple Draw a couple days ago. If his brain can handle that then 4tabling HU LHE shouldn't be a problem.
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 10:24 AM
ike/kanu/sauce will crush durrrr
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 10:25 AM
Quote:
Originally Posted by Gunner171
Durrrr was 3tabling HU Triple Draw a couple days ago. If his brain can handle that then 4tabling HU LHE shouldn't be a problem.
it's pretty tough as the bot won't be making any mistakes as it is in fact, a bot while durrrr playing many tables may increase the chances of him making a mistake.
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 10:53 AM
Quote:
Originally Posted by Chinamaniac
Hoss could prob beat it, I know of a few people that more than likely could not only beat it but could prob crush it because I know they have experience crushing some of the best bots (similar to this one ) over very large samples


Durrr may be able to beat it but it may take time to adjust and the time he spends adjusting may be the diff in winning and losing.
I think people fail to realize how much stronger slumbot is than Polaris 2008 (which played the match against the pro humans). Hoss was probably beating Polaris back then, but he almost certainly would be at a disadvantage to slumbot.

Anyone who thinks durrr has the slightest chance is deluded.
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 11:17 AM
I doubt you would get more hands in per hour adding in more than two tables, assuming the system instantly deals cards/board and the bot doesn't require thinking time.

HU FL isn't like NL where there's a bunch of folding preflop and there's a delay in cards being dealt. in FL 95% hands will see a flop, 70%+ reach the turn and near 50% go to showdown.

Honestly the fastest might just be 1 table very fast. You can probably get in 500-600h/hr. So a 200k hand challenge would take 400 or so hours.
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 11:28 AM
i completely agree w\ Hood. HU LHE can be played extremely quickly 1table
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 11:34 AM
Quote:
Originally Posted by Cangurino
Wow, this thread got out of hand. There is so much nonsense which is posted and goes uncontested... things like "The goal of GTO is to make your opponent indifferent" (wrong), "GTO in poker would win against any other strategy" (probably wrong), "GTO in poker would break even against any other strategy (wrong)... can't go through them all.

Earlier somebody asked about opening sizes in GTO HUNL. I would be very surprised if a solution would use the same raise size for its whole range, and if the sizes didn't depend on stack sizes.
Thanks for clearing this up. The statements you pointed out were causing me some confusion.

That said, is it possible that GTO play could be -EV in some situations?
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 11:40 AM
there shoud be a differfent thread just for durrrr's challenge, and you know so it gets a little more serious too.
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 11:49 AM
Quote:
Originally Posted by look at me now
Thanks for clearing this up. The statements you pointed out were causing me some confusion.

That said, is it possible that GTO play could be -EV in some situations?
"-EV" is another one of those terms that are used loosely and aren't properly defined.

Of course, playing GTO has a lower expectation than playing maximally exploitively against a known strategy. Moreover it may be losing (have a negative expectation) in games that are inherently unfair - like HUNL where villain has the button in each hand.
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 11:54 AM
Quote:
Originally Posted by AlanBostick
BZZZT!

Thank you for playing.

The "O" in "GTO" stands for "optimal." A GTO strategy cannot be playing sub-optimally in some spots; otherwise it would not be a GTO strategy. A GTO strategy will be playing optimally in all spots. That's what "optimal" means: an optimal strategy is a strategy both is unexploitable and maximizes EV.

Playing GTO against an opponent who is playing exploitably will at least sometimes result in giving up EV. Optimal play won't win as much as exploitative play. But you cannot exploit the mistakes of an exploitable player without exposing yourself to exploitation.
bit late to reply to this, going back to the example of your opposition only betting the river with the nuts, GTO will call with a wider range of hands (some non nutted hands) to remain unexploitable, is that not calling a sub-optimal range? (given that villain only ever bets the river with the nuts and won't ever adjust or widen his river betting range)
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 12:01 PM
A GTO strategy actually has less than 50% chance of winning against the Nemesis in a HUNL game.
Just putting that out there for everyone to think about.
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 12:04 PM
Quote:
Originally Posted by partywme
bit late to reply to this, going back to the example of your opposition only betting the river with the nuts, GTO will call with a wider range of hands (some non nutted hands) to remain unexploitable, is that not calling a sub-optimal range?
There's some conflation of two meanings of the word optimal, here: the commonly used form that's synonymous with best; and the form in GTO, which has different connotations. With the former, what's not optimal is sub-optimal. With the latter, what's not optimal is just, like, not optimal.
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 12:23 PM
so am i getting this right: If there was a supposedly GTO bot playing hunl vs whoever and that human player finds a way to beat it over a very large sample size, its actually proof that the bot doesnt play GTO since GTO means that getting exploited isnt possible right?

So the only question is whether there is GTO for nlhu, not whether it was beatable if there was a GTO.


also another question. GTO for hunl is obv beatable by a super user right? For example if you take the nash equilibrium shoving charts, they arent beatable even if one could see the cards of the shover. I just cant imagine this would be the case for a deepstacked hunl game too for bot playing GTO against a super user.

edit: and final question then if its beatable by a superuser can it be called GTO then regardless?
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 12:24 PM
Durrr saying he'll only do the bet if he can play up to 2k/4k with no buyout certainly sounds like proposing an offer he thinks (hopes?) won't get accepted. Even Ike, Kanu and Sauce combined can't pool together enough money for that.
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 12:27 PM
Quote:
Originally Posted by Cashcid Linc
so am i getting this right: If there was a supposedly GTO bot playing hunl vs whoever and that human player finds a way to beat it over a very large sample size, its actually proof that the bot doesnt play GTO since GTO means that getting exploited isnt possible right?

So the only question is whether there is GTO for nlhu, not whether it was beatable if there was a GTO.


also another question. GTO for hunl is obv beatable by a super user right? For example if you take the nash equilibrium shoving charts, they arent beatable even if one could see the cards of the shover. I just cant imagine this would be the case for a deepstacked hunl game too for bot playing GTO against a super user.
If a human beats is over a large sample it wouldnt be proof that the bot isnt GTO but it would be very, very likely.

There is no question. A GTO strategy for hunl exists and it is not beatable.

If you see the hole cards you are not beating the strategy, you are cheating.
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 12:30 PM
Quote:
Originally Posted by Dehlyago
There is no question. A GTO strategy for hunl exists and it is not beatable.

so what about this then?

Quote:
Originally Posted by Cashcid Linc
also another question. GTO for hunl is obv beatable by a super user right? For example if you take the nash equilibrium shoving charts, they arent beatable even if one could see the cards of the shover. I just cant imagine this would be the case for a deepstacked hunl game too for bot playing GTO against a super user.

edit: and final question then if its beatable by a superuser can it be called GTO then regardless?
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 12:36 PM
Quote:
Originally Posted by Cashcid Linc
so am i getting this right: If there was a supposedly GTO bot playing hunl vs whoever and that human player finds a way to beat it over a very large sample size, its actually proof that the bot doesnt play GTO since GTO means that getting exploited isnt possible right?

So the only question is whether there is GTO for nlhu, not whether it was beatable if there was a GTO.


also another question. GTO for hunl is obv beatable by a super user right? For example if you take the nash equilibrium shoving charts, they arent beatable even if one could see the cards of the shover. I just cant imagine this would be the case for a deepstacked hunl game too for bot playing GTO against a super user.

edit: and final question then if its beatable by a superuser can it be called GTO then regardless?
I don't think anyone is questioning that. I think Durrrr is making the claim that no bot can be GTO. He probably is correct in the idea that the current technology is not very close to producing a GTO bot for HUNL, but the idea that no GTO solution exists is absurd.
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 12:37 PM
Quote:
Originally Posted by Cashcid Linc
also another question. GTO for hunl is obv beatable by a super user right?
Yeah. Ranges only exist because of incomplete information. You can't exactly construct unexploitable ranges/frequencies when your range is your exact hand, and when a superuser can exploit all of your "ranges" by playing perfectly against them.
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 12:44 PM
Quote:
Originally Posted by Cashcid Linc
For example if you take the nash equilibrium shoving charts, they arent beatable even if one could see the cards of the shover.
Im not sure if i am reading that right but if you think you can use the shoving charts and still play break even vs a decent opponent who can see your cards then you are wrong.
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 12:57 PM
Would be very interesting with no buyout
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 01:12 PM
Quote:
Originally Posted by Herro Prease
Durrr saying he'll only do the bet if he can play up to 2k/4k with no buyout certainly sounds like proposing an offer he thinks (hopes?) won't get accepted. Even Ike, Kanu and Sauce combined can't pool together enough money for that.
durrr is a joke
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 01:14 PM
whatever happened to the durrrr challenges against Patrik Antonius and Dan Cates a.k.a. Jungleman12? They never finished right? did either player buy-out or something or are they officially still going?
Kind of weird to start a 200k hand challenge when you can't even finish a 50k challenge.. however the bot will just be online all the time..
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 01:14 PM
Quote:
Originally Posted by Dehlyago
Im not sure if i am reading that right but if you think you can use the shoving charts and still play break even vs a decent opponent who can see your cards then you are wrong.
yes you are right actually. So it is still GTO even though a super user could beat it
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote
11-25-2012 , 01:15 PM
Quote:
Originally Posted by Kozaa
Who the **** will pay money for the bot in case it loses? LOL
From what we've seen so far, people are lining up to take the LHE bot's side of this. If you're on the "durrr is a poker God and will crush the bot" side of things, you're likely to get as much action as you want.
Quote:
Originally Posted by Chinamaniac
Hoss could prob beat it, I know of a few people that more than likely could not only beat it but could prob crush it because I know they have experience crushing some of the best bots (similar to this one ) over very large samples
I don't recall any modern example of this. Has any big-time LHE HU specialist done well against modern bots? If you have a link about a match from the last couple years, I'd love to see it. I think having a NL/PLO specialist playing is a help to the LHE-bot side, as you say it provides early profit even if he's a genius. Still, I'm not sure you wouldn't find money against even the best LHE HU specialists.

The thread is interesting looking at where the opinion lies:
Railbirds -- bot has no chance
Normal poker players -- may be even money. Some like poker genius some like bot
LHE players -- bot will crush, thus could be ugly. At the stakes proposed, human could lose 8 figures

Here's my prediction if the match runs, as we've seen it in many interactions with humans and bots.
  • Match starts, bot makes strange plays, and many say "fish computer, it is going to get killed"
  • Some strong players start saying that the strange plays aren't actually -EV
  • Apparent runhot by bot
  • Human starts losing for a big number and then says "I just need a little runhot to get even, I get even and I'll quit with dignity"
  • Losing continues, with cries of "rigged!!"

My prediction is that Tom loses a moderate to large amount if he plays. It seems clear why the talk of a buyout, because this could get very ugly.
Hoss_TBF: "All top players use game theory, distributions, bluff ratios etc" Quote

      
m