One day back on APT, and it makes my head explode.
Playing BCL, $1-$2, $200 stacks. Folded around to be me on the button with 9d-7d. Raise to $7 and get called by the BB.
10-8-4 rainbow flop, BB throws out a weak lead bet of $10 and I shove. BB calls with K-J offsuit and wins with K-high.
How in the world does BB call a $200 shove vs. preflop raiser with just two overs, and a backdoor straight draw????
I'll be kind and say there was suboptimal play all around here. As I understand it, these bots are supposed to mimic low stakes live cash players. For that reason, they will do all sorts of things that make no sense from a sound theoretical standpoint. Would you ever see a real player making that call? Maybe, even though it is insanely bad play. On the other hand, what in the world are you doing shoving ~96BB's here? Before now, I wouldn't have thought anyone would shove $193 at a $25 pot. So, you had really spewy bad play on both sides of the equation and 1 of you had to win the pot. Not sure there is much else to say about this one.
I don't disagree - shoving was just the easy way of coming over the top with the open ended straight draw when the bot demonstrates weakness. In a live situation, I would have been more precise and come over the top to $50 or so. Then again, even if I only come to $50 and the bot shoves, I'm at least close to priced in for another $140 with, presumably 8 outs, twice over. If I'm going to call the bots' all-in, it makes less difference if I'm going to $50 or just shoving myself (with the presumed extra fold equity - $50 might not fold K-10, but the shove might - the shove might be better).
But the principle is still the same - and even more true with the extra weight of the all-in shove. The read and the play, I would argue, was right (although, of course, the Skalansky Theorem assessment would say that it's wrong, since it places no value on induced fold equity) as long as I price it out to draw to, really at most, 6 outs, because it has to figure that it has to hit a K or a J to be ahead after I raise preflop and come over the top.
My point simply being, if that call is open to the bots, then what the heck are we supposed to be getting out of that?
From a learning perspective, probably nothing. I agree that some of these bots play so poorly that it hurts my head. For good or ill there are many live 1/2 players who are just as bad. I understand that this site wants to help us train for real play vs theoretical optimal but I agree that some things are just plain silly. I try to ignore the absurd plays and move on. I do wish they could program some better bots to mimic at least low-stakes online play but right now we don't have that here.
Yeah, there are still lots of terrible 1-2, and some terrible 2-5, players. Over the long run I, you, or anybody with a modicum of skill and insight will crush them.
My objective here is to try to get myself back to thinking correctly about how to beat the best 1-2 and 2-5 players. Any suggestions as to what settings to try to simulate that?
IMO - I don't think KGB is right. I had settled at one point in BCL being maybe close to the right behavior, but now I'm not so sure.
Play KGBs Dungeon
My own experience is that i have been mediocre at 6max not even beating 2/5 as bad as the bots are there.
A completely unrelated idea came to me. Play 9max KGB. I have been spanked by those nits over many thousands of hands.
Funny thing happened while adjusting ranges, reading their ranges, improving bet sizings, etc. I improved quite alot.
How do I know?
I dropped back to 2/5 6max. Lifetime 6max over 38.6k hands i'm 83.83 bb/100.
Suddenly last 30 days at 2/5 6max I am 187+ bb/100 and 234 over last 7 days.
Conclusion : playing against the high levels makes one better even if one can't beat the higher levels. ( I am a -9 bb/100 loser at KGB 9max.)
Understood - that makes you better against the bots at lower levels, but does it actually make you a better player against live, human opponents?
I'll add - the bots continue to do weird things. Have a "passive, intermediate" bot in the cutoff who, with no action to her, limp calls with 8-6 offsuit preflop, but the "aggressive, intermediate" bot behind her folds K-10 on the button?
"Passive, intermediate" bot find other hands where she limps preflop from middle position with J-7 offsuit, but other hands where she won't limp preflop from middle position with 7-5 suited???
As for the examples given, it sounds like common low stakes play to me. With those ranges, it would be 3bet mania for me in position. @MAM4
Maybe, but the weird part seems to be the same bot being willing to limp-play lower equity hands and fold higher equity hands in the same spots.
It is one of those things where I expect even bad players - human or bot - to have some pattern that makes sense to them. I don't see how these patterns are anything but random.
I worry about getting too tuned to playing against completely random bots vs. people who are not completely random.
@MAM4 - I hear you but sadly these types of plays do replicate low-stakes live poker more than we care to believe. Humans change behaviors all the time, sometimes for no apparent reason and sometimes for reasons we don't see. For example, if you notice 1 player limping any 2 cards, then you can limp behind with all sorts of garbage and try to outplay them postflop. If on the other hand you notice a player with a strong limping range, you may fold higher equity hands than you played vs the 1st player. Other times people decide they want to play a pot for no real reason and will use any 2 somewhat decent cards as an excuse. The point is that some randomness isn't really random at all and sometimes it is but humans exhibit this behavior all the time in weaker games.
I think the value of playing lots of hands, vs weak or strong bots or players, is that you get to see lots of situations in a short period of time and experiment how to handle them. We sacrifice some of the quality of training here playing vs meh bots but get a lot of value from the volume of hands we get to play. Its a tradeoff of sorts. All in all I wouldn't use this as your only training ground but I definitely think it has more than enough value to continue using it.
Well, I did commit to the lifetime membership, so I'll probably continue using it.
I fully ascribe to the position that even bad players aren't random, they are each bad in very predictable ways. Some overvalue playing small pairs to the flop hoping to flop a set, some overvalue playing suited connectors, some overvalue playing any connected cards, some overvalue playing offsuit A-x, some are combinations of the above. But, once you seen them play for a bit, you should get a good read on which they are.
Here's a simple playing against the bots question. Playing 1-2 with 200 stacks, bot limps, you have J-J and make it 10 to go, bot calls behind you, and bot limper calls. 33 in pot and flop comes 10-8-5, two to a suit. Bot who limped preflop checks, you make the 25 bet that you absolutely have to with the overpair, bot behind you folds, but limper bot now check-raises to 75. Do you call the extra 50 on the flop?
Whereas with most human players I'm limiting my thinking to whether they have A-10, maybe K-10, a set, or a straight or flush draw, the bot will sometimes make that check-raise with a pair below the 10 or even seemingly nothing (like Q-7 offsuit). Maybe the bot is playing theoretically better by mixing in a basically pure bluff out of position, but I might see that from 1 in 50 or 1 in 100 low stakes players.
Villain2 has middling pair or AQish type hand. Leaning to latter bc middling pair likely stays on this flop. Villain1 has smallish pair or suited connector. Although mini % has QQ. Limp reraising is used frequently by bots with AA, KK. Im eliminating those. Even weak bots raise TT mostly. So you are looking at 88,55 or a J9, 97, 67 matching suit on board. Maybe a suited A flush draw.
Credit to Doug Polk for imploring players to think about their entire range. And so what is your limper raising range? And where does JJ fit in the catagories of raise, call, or fold?
I call to keep v1 bluffs in. Then decide again on the turn. Hero is ahead of V1 range at this point.
You say random play? Im in live tourney here at APT. Button limps. I raise 4x from SB with suited A. BB calls. Button calls. Flop A5x rainbow. I cbet 1/2 pot. BB c/r big leaving 1.5 BB behind. Button folds. Won't go thru my decision process here but after some short tank, I ship. No snap tells me im ahead. Bb calls shows 4,2 off........Random you say? A human called my 4x raise with 4,2 off.)
Spoiler: yes he binked the 3 and took a chunk of my stack. )
I like how you are thinking thru hands. Keep training.
hi hi i just read over this...
warlock and highfive are so experienced players and brilliant comments
i have some programming knowledge and i find it amazing that the bots are programmed as well as they are, it blows me away to consider the how complex it is to make the various algorithms.. like wow, really wow, and i sure it always a work in progress to tweak and improve robot performance in standard situations
in my limited experience with the zillion possible situations in poker how can we expect the robots to mimic it all?
i am also amazed at the thought you exhibit in examining the hand and play
i mostly now play real people and real money in micro-low things, rather than the robots
i am learning from warlock, doug polk, alex fitzgerald, Sky M, dozens of books, and ATP things and my poker software
real people do wonky things.. and online you have just seconds to make a decision
like tonight i get eliminated from a $10 SNG... i have JJ and totally dry board and i try get money into pot and successful, and the guy who goes with me in the pot to end had a 4 7o but got 2 pair (second one on river) .. i think how the hell can he call all my big raises with that lol
when real money at play and seconds to react... who knows what a human decides do
i embarrassed some of the dumb things i have done and I sure on other forums .. players are using me as a donkey example of things hahaha
and at different stages of my learning no i wasn't predictable... most of it was intuition, 'gut feel' with little else
if it was all mathematics and analysis Skalansky et all would be top money winners and they arent, or AllenBay would have his bots beat Negreanu, Seidel, Holz, Ivey, Hellmuth, Juanda et all