Arty I think you should definitely start thinking about GTO in terms of Heros dynamic strat versus exploitative strat instead of trying to maximize the EV against static GTO strat. While they both end up to the same situation at the one datapoint, it's way more useful to think like the first way.


And yeah we are balancing just for the sake of balancing, trying to be unexploitative. If we are not balanced/gto at some point of our gametree, then v can increase his EV by adjusting and therefor we are not playing GTO. And the point of GTO is to minimize v's winnings not maximize ours, this is the very basics of game theory. If we are just trying to maximize our EV strat vs V , then we are just playing exploitative, obviously when we try to exploit GTO strat, we might end up playing GTO ourself.

A lot of this has to do with having perfect frequencies and so making V indifferent for adjusting as he can't increase the EV by doing so.

Like lets take the example of checking topset at some situation, we are choosing to check top set because it's the most profitable decision against V's strat because V is playing aggressively against our checking range. But in this situation our checking range is so small that when we check 100% of our top set, our checking range becomes so strong that v isn't playing as aggressively anymore which decreases the EV of checking. So now we wan't to start betting it with the frequency which makes v play so aggressively against our checking range that the EV of bet is same as check.

If you think we play against a static GTO strat in the same situation you might think like "oh the ev is the same, doesn't matter what i do".

With this example you see that the frequencies are very important in GTO as if one players frequencies are off, the another one's counter-GTO strat van be very different from the static GTO strat. Obviously it's impossible to play with GTO frequencies so all of this may sound a bit fancy, but for me it helps to think how to play exploitative.

3 hours of sleep so parts of my post doesn't really make sense. But what I'm trying to say here is that we aren't trying to maximize our EV against a certain strat (maximizing our EV against static GTO strat won't make our strat GTO, because we would be indifferent with so many holdings in so many spots that we would just **** up the frequencies, and thus be exploitable), but we want to maxmize our EV or in other words minimize villain's maximum EV, against all the possible strats V can have.

doctor877,
https://en.wikipedia.org/wiki/Minimax

I just played with the snowie challenge and had settings at 100NL 6m cash. Why does PS suggest to open for 1/2 pot and then min raise in the challenge?

I seriously don't get you guys getting so caught up about the hype with Pokersnowie and GTO.

The Pokersnowie team comes to brag about how great the software is and give numbers to sell as much as they can of this thing. Buy into the hype!

http://www.pokersnowie.com/blog/2014...s-hands-played

Note the following:
- The player pool was a random pool of all levels. Posting all the results of anyone who ever played Pokersnowie is just pointless.
- This was without money, so you can't expect everyone to play seriously and competitively. Probably a majority of hands were stupid trials on Pokersnowie.
- The assessments from Pokersnowie were, well, under its own standards! Like it claims it is the best in poker before you even play against it. It tells you how bad you are, like it has worked poker out!
- Most likely, what people would be trying to do would be to try out Pokersnowie or to work out strategies to trick Pokersnowie for free
- Most who would be using HUDs online wouldn't be able to and would be trying to work out some other strategy against Pokersnowie, which puts them at a disadvantage because they would not be playing their regular game.

It is already tricky to play against any computer program. Let me put it this way: Do these players beat even a silly game like "Poker Superstars III" which is far from a sophisticated AI engine?

Do these players beat say something like this?
http://neopokerbot.com/

Neopokerbot doesn't post its amazing success results claiming its huge winrates, but it probably has beat more people than Pokersnowie since it was out.

Furthermore, if the fear is that in the future GTO and automated playing may appear, the best thing anyone could be doing is to find ways to beat it.

For what matters for the greatness of Pokersnowie, here is how hard it is to beat 10-handed. I beat it for breakfast. An 8-year old can beat it.

https://vid.me/qSw3

How do you beat Pokersnowie, really? Let me think.... If you were playing online, what would you do? PROFILE YOUR OPPONENTS. Let's imagine Pokersnowie sat at your table. How would you mark it? As a TAG? A Nit? A Fish? etc. We are playing against a computer which does perfect calculations and will always have top EV. Well, you would mark it as a GTO bot, and then hit its weak spots, just like for any other opponent.

And for anyone struggling to beat Neopokerbot, I did it right after last post. A neural network robot that learns from players.... I bet it would still have a very high bb/100 among a lot of players.

There is an extremely simple strategy to beat it, which I didn't apply as I was in a hurry, but I stacked it every time before. This time I was just lucky catching a hand and stacked it quickly.

https://vid.me/r8zL

And here is another thought.... Pokersnowie is not ONLY one of the first GTO-thinking bots. It is also the first program that actually imports your hands and analyzes your play and recommends what it thinks you should do.

To put this another way... There are already so many ways to play poker, and so many strategies to follow, but no one ever made a computer program which either applies them, or imports hands and analyzes people's plays based on them.

When was a program out that did the same and plays Sklansky-like? Or maybe like Daniel Negreanu. Or according to the strategy suggested by any poker book? How about a program applying the strategies in, say "Easy game" by Andrew Seidman (as an example)? And also evaluate your play afterwards... There aren't any.

In a computer-run world, no one made programs like that. They just made PokerSnowie and some neural network bots...

It would also be interesting to see such bots playing against Pokersnowie. Then perhaps you'd know better if Pokersnowie is as good as it claims. And you would have more training software.

And to put more clearly what Pokersnowie does: It uses a very new GTO theory which isn't even established to coach your plays. And you pay for it.

So you pay for someone to teach you the new hype before it establishes its success. It is "the best" before it has even been tried out....Ehm.... Would you rather leave great strategies to learn and pay for this new hype? Isn't it a bit risky at best?

I bet you a lot of people are using Pokersnowie because they think they will save money from hiring a poker coach, because coaches charge $50/hour upwards, whereas this software charges a monthly subscription which is way smaller. Are you trying to save money on poker training and so buy any software out there which is simply cheap?

This is crushing PokerSnowie heads up.
Apparently 4,550 challengers lose to it!

I am Phil Hellmuth! I am beating this super smart piece of software! Wow.

How many bb/100 is that? 360bb/100? Ah yes... "variance"

https://vid.me/1hlN

Because the previous videos had few hands, here is a new game with 300 hands with Pokersnowie.



I wasn't following my original strategy all the time, and took my chances, but here are the results. 100bb/100 good enough?

Continued on video...
https://vid.me/NmdA

Um, 300 hands is still an absurdly small sample size. Play a few tens of thousands, then come back and brag.

I thought you left 2+2? 300 hands, 3000 hands, 30k hands.... all not enough.

This is like the joke which says "How do you make an idiot wait 24 hours?"
"I'll tell you tomorrow".

Apparently my posts are read. Not a good place to discuss strategy for what I want.

How is 1,000,000 hands to prove the software is nonsense? I'd rather play with money then spend time testing overhyped, overmarketed software for free.

Anyway, I am not having a go but personally
1) Some new piece of automated software doesn't impress me.

2) All you hear about GTO is that it is so hard and complex that you won't get it cos you are daft. This must be like a new Nobel prize level discovery, too hard for us to perceive. Is that an attempt to intimidate us? I'm tired of reading the Rock Paper Scissors analogies as well. This is the best "explanation" people who speak about GTO give about why it works. Not good enough. Different game!

3) I don't buy the "this is some fancy new theory you'll never understand".

I don't buy the marketing hypes for any new piece of software.

I don't share the programmers' joy for building whatever they wanted.

Here is my bottom line: Does it work, does it apply to my game, is it useful in anything at all?

4) It is not even clear if Pokersnowie is using it or not because the whole heads-up game alone hasn't been fully decoded. So you are telling me Pokersnowie plays GTO? Impossible.

5) The people who sell software which does GTO analysis sell it for much more as a matter of fact. Pokersnowie isn't GTO as I understand.

6) Practice shows that a few minutes on this software show its flaws. I beat it. Look at my posts. And I am not even the only one. Google for "Crushing Pokersnowie".

You can keep denying it and give me your wild stories and hypes. I beat it badly. Very badly. Very very very badly. And it took so little thinking. Yeah, Pokersnowie will rank me the worst player in the world probably with a negative score. LOL yeah, I didn't play as it wanted me to. Oh my.

7) The GTO theory isn't even applicable for most stakes. You really want to analyse your games with it? Why? At your stakes level? I could go on why there is no reason to import your hands because you won't need it for your game. What would you find out? And as before, Pokersnowie doesn't seem to use GTO anyway. So what good is it?

There's just too many questions raised. The simple solution to all of this is: I downloaded the software, played with it, beat it. That's it. I don't need all the analysis. Big hype confirmed.

Don't buy it then ??

You can buy it though.

Firstly, I don't think anyone will deny that Snowie has some slick marketing "hype", designed to induce subscriptions. It's commercial software, after all. The (overwhelmingly negative) comments towards the start of this thread actually led to the manufacturers changing some of the wording on their site.
Secondly, it's also accepted that Snowie has some exploitable leaks. It's not as "perfect" as the original marketing campaign made out.

But... it seems you really have no idea how much variance there is in poker, or how you would go about proving the bot is "bad".
When Doug Polk and friends played Claudico, they had an overall winrate of 9bb/100 (IIRC) over 80,000 hands, but there were days when the bot beat some of them badly. Even with that winrate and that sample size, there wasn't quite a 95% confidence level that the humans were stronger than the bot, due to the high standard deviation in results.
Because Snowie plays an approximation of GTO, with some super-thin bluffs and bluff-catches, it can take a very large sample size to get a decent idea of your winrate (or more likely lossrate) against it. I only played about 10,000 hands of 6-max against Snowie. For the first 5,000 I won at something like 13bb/100, which made me wonder if there was something very wrong with the program, because I'm not very good at poker.
For the next 5,000 I lost at something like 17bb/100 despite getting a higher rating (with several sessions rated at "extra terrestrial"). My 10,000 hands was a "lolbad sample size". I'm afraid to say your 300-500 hands are virtually meaningless.
It's like the weekend warrior who buys into 100NL zoom on a Friday night, runs up a 600bb stack and thinks he's solved poker, when really he just had a heater. :/
Snowie probably isn't of much use to you at the present time, so if you didn't get much out of the trial, that's fair enough. I'm not sure why you need to rant about it though. Plenty of other people have found it useful in one way or another.

If you try to play your normal game against Pokersnowie perhaps you'll lose. As I said before, if Pokersnowie was sitting on your table you would want to play a different game, like you would play different for different kinds of opponents.

How you would normally face other opponents is perhaps you'd work out their ranges and play with adjusted ranges, work out your odds etc. It seems that perhaps this approach won't work for Pokersnowie. Plus, it will never tilt or make incorrect calculations, so one could eventually lose from that as well.

I wasn't trying to win based on cards. And I didn't win with 13bb/100. I won huge amounts. This includes all the times where I titled and called Snowie's bets when clearly I had lost, or the number would be higher.

I think that in the way that Pokersnowie is built, it is exploitable, and in that case it is not a matter of variance to beat it because it will always play the same way and it will always be exploitable in that way, and so a small sample size is enough to prove it.

I can't prove that it is exploitable theoretically but I can't see any theoretical proof that it isn't either so all I can see is the practical result at the moment.

Here is one thing about Pokersnowie I am thinking of.

So it analyses every situation (in whatever way) and then estimates a % to call, one % to raise and one % to fold.

I think how these numbers work is that Pokersnowie will ALWAYS pick the highest one. It will ALWAYS do that. And it will never do anything else.

If bet sizes are involved in working out those numbers, all one has to do is bet in such a way so that these numbers will change so that folding will be the highest option for Pokersnowie.

So when you bluff it, if in some situations and for some amounts it may be considered stupid, perhaps above certain amounts it is a 100% winning move because Pokersnowie will always choose fold even if it is a marginal decision. In real life, opponents would sometimes call there.

So to beat it, you would not be playing the cards, which also involve variance. You are playing to make Pokersnowie choose fold.

My idea is that playing any two cards deep stacked you can crush Pokersnowie and perhaps GTO.

Also, since Pokersnowie makes perfect calculations, the more information is available on the board, the more at a disadvantage you are, because Pokersnowie will calculate better than you. So the later the street, the worse for you. Preflop is the best point to beat Pokersnowie. This way you make its biggest strengths redundant.

All this is without really knowing how Pokersnowie works. It is just a rough guess and maybe Pokersnowie calculates different.

^ GTO cracked, the future of poker is saved

yes

You sir are an idiot. I rest my case.

You say that you don't like 2+2 because people aren't talking strategy or whatever, but you come off as very dense and arrogant (quite common in the poker-scene).

Lighten up, be able to digest what people have to say. Not all of it is worthless and if you could open up your eyes a bit and gain some humility perhaps you would see that.

I put forward 2 pages of arguments. Someone coming to say "No" like a baby is just ridiculous. No argument, no case. I appear arrogant? I put forward my arguments. I spent a good few hours playing with Pokersnowie and putting together a complete answer. No one spoke with arguments on these last 2 posts. Just No no no no no. Here is some news: Yes, yes, yes, yes, yes, yes....! And guess what else: Yes. Give me arguments, not personal judgments.

PS: Yes

I am very skeptical about snowie as well, but he's right, you obviously have no clue what you are talking about and should read up on poker theory.

If you are interested in GTO, you should read, or better take a course about game theory. Look for Nash Equilibrium as it is what people are talking about when they speak about "GTO" in poker.

The theoretical knowledge required to understand what this is all about is way below any math degree.

No. Google "mixed strategies".

By definition you can't beat poker "GTO" (nor can you beat a NE in any symmetric game anyway).