Beating the Heads up Limit Poker Machine - Page 4 - Heads Up Poker

Stac123 · 03-03-2012, 03:33 PM

Op is right. Machine has been programmed to only re-raise post flop with a strong nut range when opponent has limp/called pre. Themuppets is also right on.
I did a tour of locations looking for less smoke on a friday night. Most machines were empty. If the machine is long term easy money they would be filled.

Stac123 · 03-04-2012, 08:24 PM

The machine at the Rio is advertised as a dollar machine. It has a one dollar sign on it and has no other limits. It's actually a 2/4 limit machine. Blinds 1/2. The one at Green valley has 1$ $5 $10 stakes. The $1 is actually 2/4 limit with you having to post $2.

Stac123 · 03-04-2012, 09:01 PM

If I went to play a $1 slot machine and the machine said I have to pay $2 once the money is in I'm sure its a violation.

jlnoble2400 · 03-04-2012, 11:53 PM

Sorry I have not been on in a while. I made another 75 bb. But the machine does play for the most part now a different game than what I had been origionally beating it for. Here is my analysis. In its sb it raises 3/4 times preflop. If you check it will c-bet you can set it upo for the ck raise and this will determine its hand for that for the most part. But if it is in the bb it only raises 50 - 60% percent of the time from the bb, generally does not C-bet from the first position without something it is planned to call you down all the way to the river with. So the raise on this opportunity of betting should not be taken unless u have at least middle pair. Because when it bets here it almost always has showdown value. This is only for the lowest stakes. Also the sucker bet for the bonus bet is forced and u cant get out of it. It is a dollar bet that essentially is the house edge since there is not a rake.

jlnoble2400 · 03-05-2012, 12:03 AM

Sorry I was not on in a while. Been busy playing. My largest downswing from the start has been 28 bb. But I so far have shown an average gain at the lowest stakes, 9.5 bb per hour average. The machines personaility changes from time to time. Here is my analysis on the new change. The computer still raises at least 3/4 of time from the small blind . It will also c-bet if you check the flop going for the check raise here will get the computer to identify its hand of either strong or weak. But in first position a complete different story. It raises 50-60% of its big blind hands. It will only c-bet the flop if its prepared to call u down to the river or bet all the way down to the river. SO I reccommend calling with ace high or king queen high or better on nonconnected flops. I reccommend raising with middle pair or betteron non fitting flops. Bluffing after its cbet has no value because it will call u down if it bets first position

johnnyrocket · 03-08-2012, 06:01 PM

Quote:

Originally Posted by jlnoble2400

Sorry I was not on in a while. Been busy playing. My largest downswing from the start has been 28 bb. But I so far have shown an average gain at the lowest stakes, 9.5 bb per hour average. The machines personaility changes from time to time. Here is my analysis on the new change. The computer still raises at least 3/4 of time from the small blind . It will also c-bet if you check the flop going for the check raise here will get the computer to identify its hand of either strong or weak. But in first position a complete different story. It raises 50-60% of its big blind hands. It will only c-bet the flop if its prepared to call u down to the river or bet all the way down to the river. SO I reccommend calling with ace high or king queen high or better on nonconnected flops. I reccommend raising with middle pair or betteron non fitting flops. Bluffing after its cbet has no value because it will call u down if it bets first position

sounds like you've ran off the charts, with the edge vs this thing you should be having a lot larger swings, continue the run hot

johnnyrocket · 03-08-2012, 06:03 PM

Quote:

Originally Posted by jlnoble2400

I am going to first tell u where the limit heads up machines are in my neck of the woods. Then I am going to tell u point blank how to beat them.
They are just outside of the poker room on Kansas Speedway.

Here is the basic Strategy for beating these machines on limit play. I am making a consistent income beating it so u probably would want to listen to me as I give u a machine basic strategy.

First Off preflop
Always limp on the button never raise then what the computer is going to do is raise u most of the time about 75% of the time u will call his raise with any two cards.
If u are first to act not on the button always check the flop with the plan of check raising the computer if he bets his button this will force the computer to identify his hand strength a lot of the times he calls your bet some times he folds and most of the time this gives u free card on the next street and a steal opportunity on the river if he reraises u and u have top pair u call him down, even with middle pair, if u have a lock hand call with the purpose of betting the turn and river for value on the latter streets. If u dont have a hand u should fold, if u have a flush or straight draw call his reraise.

Now Turn Play
This is very important as the computer will sometimes change personaility and try to steal on the turn but most of the time he checks the turn I have found a check raise bluff on the turn to be effective if the flop has been checked by both sides and the computer bets the turn on the button. If he checks the turn and u dont have a made hand and u are on the button go ahead and bet it and try to take the pot right there. If the computer calls your turn bet it is highly likely he is calling the river bet. But if u check in first position and he checks back he most likely folds to a river bet.

River Play
If u have a made hand and the computer called your turn bet bet the river he will call it and u probably will get paid off. If the computer called your turn bet and u are last to act and the computer checks to u go ahead and bet it and try to win the pot. If u are first to act and he has called your turn bet and u have nothing it might be smart to bet anyways if u are first to act and see if he is calling. Proportionately the river bet wins u money long term in folded pots equity.

Aggression is the key to beating this machine. If u played a basic bet every street strategy after the flop in first to act position and raising his first to act betting every hand u are in this will not let him call u down with weak cards this could also be winning strategy as the computer will have a hard time calling with weaker hands. Since u are showing extreme strength every hand. But I dont quite play that way my trick to limiting is the first post flop opportunity he has to bet I generally check raise him to see where he stands and this gives me the control of the hand. If this is done on the flop it is highly likely when u check the turn u can get him to check back most of the time almost certainly letting u know if u bet on the river u win the pot.

not sure if level? always limp button so you can play smaller pots in position vs a similar range that you'd be able to identify with raising? Why not limp/3bet with this strategy, just too many flaws and I only read a few sentences in to the strategy.

jlnoble2400 · 03-09-2012, 04:03 PM

short buy strategy and pushing on button might work and cahsing out with profits.

jlnoble2400 · 03-09-2012, 04:05 PM

makes sense will try it

themuppets · 03-09-2012, 04:17 PM

JDalla had the same idea about LRR'ing and tried to get some discussion going toward the end of this thread. He too estimated that the machine was raising open limps about 75% of the time. The discussion didn't go very far but imo you could just start with hands that have a clear and significant equity edge and then grow it slowly depending on how the bot is responding. Alternatively, you could use PokerStove or Slice (the latter is really better for coming up with HU ranges on the fly) to find the bottom of a profitable 3-betting range. I like the former idea better because you'll have a stronger range to work with while you're sussing out the bot's postflop play in the LRR'd pots.

johnnyrocket · 03-09-2012, 07:02 PM

yeah, interesting post muppet, i'm definitely curious to play against it with some friends if i make it to the wsop, just for fun though, not for a living which apparently 50 2+2er's have tried and claim it's rigged since they don't understand the HU LHE theory well

themuppets · 03-10-2012, 05:13 PM

I guess since I bothered to work it out I'll post it here. The range of hands that have > 50% hot-and-cold equity versus the top 75% (as ranked by Slice using the HU configuration) is:

A2o+, K6o+, Q9o+, JT, A2s+, K3s+, Q7s+, J9s+, 33+

This is just less than the top 36% of hands.

Obv there's more to the story than aipf equity but it's probably a reasonable baseline if someone wants to mess around with it. It's interesting to note that JDalla's off-the-top-of-his-head range was pretty close. It goes to show how good people can get at guesstimating this stuff.

To be clear, I'm not convinced of the limping strategy at all. If it is true that the bot almost never folds his bb, and also that it responds to the limping strategy with postflop play that is straightforward and exploitable (I remain skeptical, which is not to say I necessarily doubt the reports itt), then it maybe has some merit.

I actually started trying to work out the scenarios under which it would make sense to apply such a strategy, but it gets pretty complicated even if you're just looking at it from a pf perspective. Roughly speaking, lrr'ing versus a bb that raises 75% of the time is equivalent to open raising if the bot never folds the bb and 3-bets 25% of the time, but in that case the 3rd small bet is only going in when the bot's range is (presumably) much stronger. If we limp everything, we also save money with the weaker part of our range, but again this only remains true if the bot would hardly ever fold his bb and as JR notes there's something to be said for positional advantage.

That's about as far as I want to go with it for now as I almost never limp versus my human opponents and i don't foresee playing the bot myself anytime soon.

philnewall · 03-11-2012, 06:43 AM

Any of you guys thought that maybe its raising of limps isn't grossly exploitable?

Although I doubt the 75% figure I don't think you'd do well against a figure just somewhat below that.

By limp reraising you'll only gain a bet against the hands that raise a limp and don't 3-bet against an open raise (say 30% hands. 65% raise-limp minus 35% 3-bet range). You'll cost yourself one bet against the hands that would call a raise but check vs. a limp (28%) and you'll give a lot of free equity to the hand that'll fold to a raise (7%).

Preflop is the easiest street for bots to get right so I don't think you'll print money against it with anything that simple.

themuppets · 03-11-2012, 08:37 AM

It seems to me a good observer should be able to distinguish between 65% and 75%, but I agree it's hard to accept without first hand experience. The only reason I even entertained the idea (other than boredom) was that I remembered JDalla's posts in the thread I linked above. And as I said before in an overly wordy way, even if we accept that number we would still have to believe that the bot's 3-betting and folding behaviors were also unusual before the LRR strat would make sense. Never having played it myself, I don't really want to be overly dismissive of those possibilities but I can definitely understand how you would have your doubts, especially given your study of Polaris.

Anyway, I don't think anybody's really advocating the LRR strat at this point. Remember: This part of the discussion began with JR basically telling OP, "If you're going to limp, I don't see why you're not LRR'ing."

TBH, I think depending on particulars such as the bot's c-bet % and how its postflop play is affected by the pf action, getting an extra small bet in pf could turn out to be relatively cosmetic. IMO the most interesting thing about OP's strategy is that he seems to think he's inducing a predictable postflop pattern by way of taking a non-standard line pf.

philnewall · 03-11-2012, 10:40 AM

Seeing as how the bots are still in action however many months later you have to assume that they are winning a fair amount to compensate for all the running costs. My feeling is that this bot's biggest edges are probably in its preflop play -- there just isn't that much variation from one bot to another. Most players from a full-ring LHE or NL/PLO background are just going to be making so many preflop errors there that it doesn't really matter if the bot drops a few clangers postflop.

I also intrinsically disbelieve the opinions of someone who claims to be beating it for 35+BB/hour with a max drawdown of 28BB. Lawl.