Background: Im a live player, 90% 2/5 10% higher.

Id like to examine software in studying multiway spots specifically and am curious whether Snowie or PIO would be a more helpful approach. Budget is not a concern but Id like to focus on one for now.

Also, to be clear, this isnt necessarily to improve my game at a higher level. I actually find 2/5 to be more difficult than 5/10, or, put another way, you are put in much more bizarre spots at the lower stakes...and Id like to examine them from a new perspective.

(Example, 7 limps to you in the sb. Example2. You 3b 4x and get 5 callers. Example3, you flop tp as the pfr and get donked into with 2 callers)

These may sound trivial/lol but Im finding myself quite curious in 2017.

Iyo, which direction should i start with? What does each program offer wrt multiway theory?

Fwiw i have read up on both and I think pio makes more sense to me from a mechanical approach but snowie just seems so interesting.

tyty

I do not believe PIO does multiway pots.

Simple 3-way postflop does, snowie does too *i think?* but never liked snowie

In general in MW Pots you shouldn't be getting exploited if there are players left to act behind you because the more players left the more ranges, so generally there will be enough ranges to defend (theory)

Usually a bet & call with multiple villain's left to act is quite strong if they have reasonable stats/are tight players, use population tendencies/reads to exploit pretty much the same thing i find being the PFR the hardest part of MW Pots tho deciding when to bet or x...

I don't believe either one will be helpful studying live multi-way play. You'd want to look at MonkerSolver, which also may not be up your alley.

Snowie is really quick (instant) at providing "solutions" to multiway spots, but the solutions won't be very useful for live multiway pots, because live players suck. Snowie plays extremely cautiously in multiway pots (e.g. hardly ever cold-calling or overcalling pre-flop, folding overpairs to a single bet on the flop when not closing the action) and while I think its nut-mining style might be the "optimal" way to play multiway pots in theory, it could teach you to fold too often against the live droolers. Similarly, it hardly ever iso-raises limpers pre-flop. I should think that isolation raises and c-bets are a key source of your EV in live games, so playing like Snowie would cost you money.
As you know, I find it an intellectually stimulating tool for studying what optimal play might look like, but it's pretty well known that the way to crush live games is to massively exploit the fishes' imbalances, not play "GTO" against them.

Very true and having read some of his posts I think Op is pretty well versed with that. I think op is just seeking new technical (I hate the misuse of the word "theoretical") perspectives.

Op, I'm interested in this too... (although I'm only online) so I'd be interested in your conclusion.

Although you say budget is no object, you might want to check the hardware requirements for piosolver.

Thanks for the words of confidence and yea the hardware requirements were something I was (ignorantly?) not thinking about.


This looks very cool. You strike me as someone who is always on top of the latest developments, a quality I envy and one that is certainly a necessity in 2017+ poker.


As we've chatted about previously, I simply want to start thinking about live poker exploits at a deeper level than I believe has been done.

Here is the bread and butter synopsis of the 2017 approach:

At a high level, stacks are shorter and people don't fold, so play an unbalanced value heavy tag game and bet big.

However, I find myself awake at night wondering how a max exploitative adaptable ai would play in such a game. I realize snowie or pio won't help me much here, but I thought they'd be an interesting perspective

As far as I know, ai and solvers have totally revolutionized what people basically thought were standard approaches to the game in 2010.

I wonder if ai would do the same thing to what we consider "standard" approaches to live poker, where 50% of the table sees flops and people have a 0% fold to cbet.

I wonder if ai would do things we'd consider "weird". Such as:

The best player I have ever played with does not ISO raise multiway. He overlimps hands as strong as JJ/AQs after several limpers. At higher games he plays very differently / much more standard TAG, making this all the more peculiar.

I have been tracking him for some time now, and his wr is north of 20bb/hr. This is why I am interested in different approaches to the game.

Thanks for the post Arty, seems like what snowie is doing is sort of correct imo theoretically at least never liked snowie lol.

Make of this what you will, but Snowie's strat when facing multiple limps is super-weird.


It also folds AKo, and polarizes its iso-raising range with some random small suited hands.


I think it's very interesting to look at Snowie's suggestions, and studying it could help you conceptualize multiway pots in a way that would improve your play, but I don't think copying it would be a good idea. This can't be the best strategy for exploitable real life oppponents, can it? In addition, if you folded 94% of hands every time there were a couple of limpers (like Snowie does), you'd die of boredom.

Thanks for the limping range. Arty, does Snowie show utg's limping range? I assume utg+2 and hj are similar to co, maybe a tad less conservative?

Ava, if you had to guess limping ranges for The Best (utg and from mp after three limps) what would you say?

Interesting topic.

His main utg open limp range is small to mid pairs (fairly standard even amongst strong winning players at 2/5) and suited broadways as strong as KQs (not as standard).

His overlimp range in lp is very wide (56o wide) but also strong hands (he'd limp TT/AQs in that example)

Thanks arty, yes i knew i wouldnt be mimicking, but im a curious cathy and snowie seems really neat. I am curious if snowie assigns any premiums to the limps.

That's helpful. So TB isn't mixing premiums into his utg limp range?

I'll let others confirm, but I believe that Snowie does assign premiums to ep/mp limps, which is why it plays so conservatively from, say, mp vs three limping Snowies. I can understand why getting limp/reraised with TT would suck. But I don't get why you'd open fold it

Not sure how software would handle that, but in practice in most 2/5 games a player doesn't have to protect his limping range by mixing in premiums because he has a couple other players do that job for him.

Snowie does not recommend open-limping any hand, because limping is less profitable than raising. It also doesn't assign ranges as such.
It does, however, have some "experience" of playing against limps, from when the neural net played against bots that acted randomly or with fixed non-GTO strategies, like limping 100% or raising 100%. It basically just looks at its huge database of past results and asks "Did I make money if I called or raised with XX when there were three limps in front of me?"
Evidently, it wasn't profitable to raise the limpers too often, in case they showed up with aces, or started limp-reraising aggressively. Even if the iso-raise just gets a string of callers, that's pretty bad for you if you have TT, because you're gonna hate the majority of flops.
The curious thing is that the iso-raising ranges have a lot of similarity to 3-betting and squeezing ranges, so it often appears to play vs the limp as if the first player had actually raised.
--
FWIW, my understanding of multiway pots (or at least the way Snowie plays them) is that you aim to make the nuts, not the second nuts. I think that's why it's not keen on hands like TT/99 or 98s. When there are multiple players in the pot, making a second best hand is incredibly costly. Those mid-pairs and middling SCs apparently have significant reverse implied odds if there are 3 or 4 players in the pot. e.g. If the flop comes KT9 4-handed, then TT is unlikely to get all in vs anything but QJs.

this all gives me a headache, but it's encouraging to note that I'm not the only one who overlimps in many spots (not all) where others ISO.

Can't claim a 20bb wr tho...maybe 2bb

I've found a monkeyspanker online; is that the same thing??? I'm not sure I want it up my alley tho

That's one of the bigger differences between live and online. If you play 6max on Stars, nobody will help you defend limps by overlimping and calling isoraises wide. That makes limping way less profitable than in a live game where 3 guys behind you overlimp and everybody who wants to isoraise knows that multiple calls might be coming.

I think actually many people iso-raise far too wide both online and live in an attempt to exploit very weak limping ranges that fish have. I've always thought limp/calling might be a reasonable strategy with lots of hands at tables that iso wider than they cold-call. For example KTo might be a button iso raise against a couple limpers, but never an overcall against a raise and a cold-call.

About Snowie, it's not a GTO bot. It can be a nice training tool in some circumstances, but I don't think it's reasonable to use it for studying theoretically optimal ranges, especially in less common circumstances where it hasn't had a lot of training against itself. For example, I doubt the TT example given in the thread is remotely close to GTO. It is definitely not GTO if we give the opponents common limping ranges and use that as a starting point.

I'm not sure if there's a good tool for multi-way pre/post-flop analysis. Analyzing the decision trees by hand might be the best way?

https://forumserver.twoplustwo.com/1...olver-1651569/

I'm pretty cautious about iso raising live low stakes because it's so difficult to get heads up and playing hands like KJo or 88 3 or 4 ways just sucks. Also that limper chain may well have people holding AQ or TT in it, so just blazing away like playing 6 max online is not always optimal.

Yeah earlier position they limp likely stronger limping range then mp/lp.

have you considered crev? dont use it personally but i use snowie and i think u would get a lot more out of crev. (i do really recommend snowie btw, apart from its bet sizing the bot is really good and almost anyone can improve by playing against it)
+1 to monkersolver btw, has a good PLO preflop tool as well

Crev would not be useful for studying multi-way pots.