spoken like someone that is about to reverse engineer the thing... let us know how it goes.

the machine is unbeatable over extended timelines.

How do u know that? Players claim they make steady income

LOL. I find my comment kind of amusing given that I'm almost for sure the most prolific poster in HULHE over the past few months.

Maybe some are, maybe not. I have noticed that all the guys who post here saying that they're beating the machine sort of disappear from these threads after awhile. Nobody ever checks back in and says, "Just so you guys know, I'm still crushing it." Perhaps they're too busy rolling around in piles of money.

I'd be absolutely shocked if this machine were beatable for any rate worthy of playing it for.

People popping into this thread giving dollars won/lost is largely irrelevant anyway since you'd need a very large sample size in order to ascertain if you actually have an edge on the machine big enough to play it.

fwiw, if u actually were crushing it, saying so on a forum(causing others to try it and potentially profit thus maybe having the guys refigure the machine, or recall them)

would be very dumb. that doesn't mean ppl wouldn't do it tho.

any idea if these machines will be offered in other games?

Awesome

^ this.

It can still be a highly profitable machine, since:

"Critical Fumbles" * equity < avg.(Vegas casino goer's HULHE neg. edge) * equity

This isn't even taking into account performance variables (user fatigue, "free drinks", desire for big-wins, etc.) or the psychology of random reward.

Oh, welcome to what happened to every poker game ever since 2004.

Pretty scary that is exactly how the ring of 10 bots played on pokerstars in 2010 that racked in 850k+. They would snap call then fold to a 1 bet. You can look it up it's all over the place. Sounds like they are running a very similar logic/program.

^^ when faced with identical situations, the machines play will vary. it is not programmed for absolute decisions, instead it applies frequencies. to use your hand as an example, if faced with this totally identical decision 100 times it could be programmed something like (call/raise/fold) 15/84/1. it just so happened this was the one in a hundred fold.

i dont know **** about limit holdem but on the Q/Q/Q/4/4 board pretty sure bot figured you would not bet river in a spot where (you assume) hes never going to fold and why would you bet to risk losing another bet or folding if he raises makes more sense to c/c when you playing board in this spot bot assumes you know this (incorrectly)


but idk stakes in hand guess he might have to call depending wut he getting on river hoping 4 55 66 77

fyi i did not read your post just saw board and 1st sentence below picture

^^ perfect handle

Bot routinely calls that board with 3 high, it "figured" nothing of the sort. I do believe it's programmed to take certain actions a certain % of the time for every situation it faces, like clay said. But I will admit sometimes it's very suspicious how it seems to make big laydowns at the right time and never at the wrong time.

Has anybody seen the machines in California?

^Not me...

I could be way off-base with this idea, but I've often wondered if there is a phenomenon that occurs when playing against humans where if you throw random small mistakes into an otherwise great strategy, that it results in a human opponent playing much worse against you.

For example, if I'm playing someone who does something completely outlandish every once in a while (like some of these screenshots of the bot seem to show), the result is that I both underestimate an opponent and I find it harder to think of a logical strategy against that opponent. If the bot was doing this, it would be some sort of non-GTO exploitative thing where their trials have shown that random weird actions cause humans to play worse against it.

Does anyone think this makes any sense or is it stupid?

perfect sense. open limp every once and a while and call an overbet with 8 high like a boss. then you are guaranteed max value from the whole table because they all have you color-noted as turbofishhh

Seriously, though, I think the bots perceived mistakes fall into at least 3 categories: (1) really bad ones that are basically just relics of the development process where it does something x% of the time that it would eventually learn to do 0% of the time if they had the computing power and development time, (2) balancing/change-up type plays that might appear weird but have the inherent value of making it hard to read/difficult to exploit and/or inducing further mistakes by its opponent (3) plays that are actually pretty standard in HULHE but aren't understood as such by the people posting the examples.

I def think there's a lot of value in various types of set-up/change-up plays versus our human opponents. I wouldn't necessarily characterize these plays as mistakes, however, and they likely don't require making choices that fall into the first category above.

I do think there's some merit to the idea that after the AIs have been evolved, they are likely tested against various human players to guage their performance in that arena before being pushed out to production as one of the bot's personality choices. It may be the case that some personalities that include a lot of the type 1 mistakes actually outperform others for the reason you suggest: they succeed in accidentally exploiting (human) population tendencies, perhaps in part by inducing weird adjustments that actually lower the human player's EV.

I think the "worth playing for" thing is huge. I do think its beatable, just that the edge is going to be so small that its simply not worth the time. I will play the one at Aria while waiting for a seat to open up in the room. I haven't meticulously tracked so I'm sure I'm going to be flamed, but playing at it over the last few months I am 99% sure I'm up on it, but nowhere near enough to be anything other than a pleasant way to kill time. Anyone at Aria is welcome to come watch me play it, I feel confident enough that I'm willing to risk public shame and ridicule...

And maybe it's just the one at aria, but it has a few patterns that are fairly reliable to the point where if it does assign distributions to each possible decision, it must be 98/1/1 in favor of the predicable act. The one that I feel most confident in is that I have never once seen it call my bet as opposed to raising me pre-flop with an ace...period. If you open and he has an ace, he 3bets. If you 3bet and has an ace, he 4bets. So if he flats me pre-flop, closing the action as opposed to raising another bet, I can 100% eliminate Ax from his rage. I'm sure someone will flame me and tell me I'm a moron, fine...I'm just reporting my experience...regardless of whether it's my open or my 3bet, if he calls as opposed to raises, Ax is not a possibility.

And if this has been discussed earlier in the thread, my bad.

I've seen $20/40. The Venetian machines are no comp no points despite not being labeled as such.

This line was awesome!

First time ever reading this thread and I am enjoying these photos of the hands and some of the lay downs it has made while posters have been playing.

Interesting.

Its not super difficult to believe the weak full house lay down could be part of either an exploitative or GTO strategy since the reasons that "no one in their right mind" would fold there is primarily because they are playing fish, bluffing that texture is basically a reverse freeroll. Its really easy to think of a heavily exploitative strategy there where always calling would indeed be very bad. Thus the machine needs to revert to mixing it up.

The second nut hand is a lot more difficult to believe though. That line is so extreme that it just isn't plausibly defending against anything. If your opponent only bet the nuts there, which is the only line folding the second nuts defends against, they would be losing such tremendous value that the particular spot or ones just like it would likely affect their overall results even if they only did it 1% of the time. The line of only betting the unchoppable absolute nuts on the river is so completely dominated that it just seems like it could never be right under any circumstance, GTO or not. So there would be no point in defending against it any fraction of the time. I could be 100% wrong on that though.

It may only fold second nuts 1% of the time.

I understand but my point is that the line that folding the second nuts defends against would seem to be not only strictly dominated but so far into that category that there's no conceivable way it could be anything but.