Against who?

Forgetting the creators have incentive to lie, against who?

Certainly it would lose straight up against Linus as it is not GTO (close in spots) and has known weaknesses that can be exploited. Snowie is never entered in computer competitions, either.

That statement from the creators means nothing and I am sure Snowie could win far more bb/100 against very poor strategies.

I am thinking Snowie is probably strongest if you only play the exact sizes it trained against.

I am not knocking Snowie and know some strong players have worked with it; simply looking for opinions-- especially from someone like ArtyMcFly.

Thank you for responding and maybe we will get more replies!

Won against Jungleman in a trial and they claimed that noone is accepting a challenge in a heads up match.
You can check Andrew Brokos thinking poker podcast episode where gest is one of creators. Snowie played against itself,so a lot of Snowie bots that played randomly...

Episode 79 I think it is...

Sorry,episode 54 with Jonathan Levermann...

Hey, is there a way to watch this part without being a paying member of upswing?

Don't think so.Maybe you can find it on the black market for off price.

is there not a support thread for snowie in the commercial software forum?
all I could find was this thread. thx

What is your question been using it for 4 years

There is not. This is it. Best to ask any questions you have in this thread.

Hey guys, thanks for responding.
Whats up MCA.

I bought the intermediate version a couple months ago, been playing a bunch, and wanted to know more about how the AI was built. Seems like there's pretty limited information out there. I watched a webinar with the 'head of research' from way back in 2013/14 when it was released but it didn't shed much light. They mentioned GTO a lot and also that Snowie had practiced by playing trillions of hands against real opponents and itself.

I guess I can look through this thread and see if there are any conversations about it.

Hey RWE,

Haven't seen you in a while. Nice to see you around the forum again!

I think you'll be able to find a good deal of the information about Snowie by reading through this thread. You could also check out the podcast disident mentioned above, although I have not checked it out myself.

I doubt that their method of training Snowie resulted in a true "GTO bot." I think it's unlikely due to the limited amount of bet sizes that they used, the actual time it would have taken for many of the less common spots to converge to true unexploitability with this method and with the computer processing resources that were available to them at the time that they put it out for public consumption.

I think it can still be a decent training tool, but I certainly wouldn't advise trying to precisely replicate its advice at the actual tables.

Just take Snowie's out of position preflop 3betting for instance. Very few, if any regs, actually use a pot sized 3bet from the blinds versus an in position player's open raise. The most common size is probably closer to 150% pot. Because of that, the ranges will change pretty significantly, both for what the out of position player should be 3betting at a different size and the ranges that the in position player should be continuing with when facing that size. It would probably even change the opener's range of hands that it chooses to open with a little bit if it knew that it was likely to face a larger 3bet sizing.

There's several other spots that I'm sure you will either come across by playing around with it or by reading about it in this thread. That's just one aspect to consider and it's preflop. The further you get into the game tree after such differences at the start of the hand the further the exact advice is from being beneficial to the games you actually play in. You know, the ones where half of the table limps and then all of them call your large Iso sizing.

Even in tougher games many of the solid regs are far more aggressive than what Snowie advises and many of them are probably capable of making much more money at the actual tables with their strategy than somebody who could precisely imitate Snowie's strategy at the tables.

Anybody who can beat snowie at World Class level at at least 50k hands - will do just fine at 50 nl

I think that is probably a true statement for the majority of players who could average at least World Class level with the caveat that they're not using the live advice to achieve that level.

I know RWE did, and probably still does, play higher stakes than 50NL.

Like I mentioned, I think it's a decent training tool, but I also think that trying to precisely imitate its play at the actual tables would be suboptimal. I think trying to understand why Snowie plays the way it does and understanding its limitations for application at the actual tables is equally important to improving at poker.

I used snowie for the first time a couple weeks ago and have played over 20,000 hands since. I learned to play poker by playing Hoyle Casino back in 2001, and have always wanted a competent CPU flop opponent, so this I absolutely love.

Two things, recently the training mode will randomly crash and all the hands played in that session are lost. A couple heartbreakers in there when stringing together really strong 500+ hand sessions. It didn't do this at all the first week, also it buffers far more often in between hands now. Is there anything I can do about that?

And finally, how's this for a blunder? Btn opened, I pot from sb, btn calls. I quarter pot, he quarters, I call. I spike ****ing gin on the turn, and its a monumental BLUNDER for me not to donk out quarter pot? I don't get it.

In typical snowie fashion, if we're playing for stacks, they have IT.

I guess it thinks you are missing easy expected value on the turn with a proportionally small size bet in a big pot.

Snowie Opens

What are the opening sizings Snowie generally recommends, like on it's Advisor widget?

It looks like 3.5 x is a button standard under most conditions but I am not sure..

It mins or 2.3 from every position almost always.

100bb deep usually 0.5pot from every position,some combos 0.4pot from LowJack,1pot from Button and SmallBlind.

Thank you.

I wonder why Snowie goes so large on the button when Pluribus on average went smaller and Monker is smaller.

Even rounding to half pot or pot, my incorrect guess would have been that Snowie would settle on raising to half pot on the button.... after having played itself over and over and over.


Thoughts?

Well, I've played well over 20,000 hands the last couple weeks and recall snowie opening for pot once and it wasn't from the button.

The opening sizes and the ranges Snowie uses are rake and position dependent. You can see this for yourself by choosing to play around with Snowie at the lowest stake available, a mid-stake and then the highest stake available. At lower stakes Snowie will often open raise bigger while at the highest stakes it often open raises smaller. I believe the reason for this is that we're more incentivized to raise larger to disincentivize callers and end the hand preflop when the effective rake is higher so that we pay the postflop rake less often. The higher the stakes the lower effective rake is and the less of a negative effect it has on our winrate when we see more Flops.

It depends only on rake.Because rake is bigger at low stakes,Snowie goes bigger to end action preflop(no flop no drop).Pluribus is AI so is similar to Snowie but was competing in high stakes and 200bb deep i think.Monker is not AI,it depends on input and will solve whatever task u ask to solve.

Thank you for your answer.

Here's an odd occurrence. Cut opens, I pot button, they call. I check back flop. They pot turn, I call. They check river.

It categorizes our hand strength as strong, which it isn't. Best hand at showdown 85%? How can that be, what hands do they bluff pot with on the turn? I doubt many 66-99.

I turned my hand into bluff and shipped, unsuccessfully obv. Snowie hated it, and it said all bet sizes are really bad. Hmmm, ok, maybe. But not because we have some powerhouse showdown hand.

When CO checks river on a total blank after potting turn they are very polarized between a few traps and complete air. When you turn third pair into a bluff you are essentially trying to make a weak top pair or a middle pair fold, but that hand category wouldn't pot turn to begin with, so your bet doesn't accomplish anything. Your hand is categorized as having 85% showdown because it beats all the bluffs that gave up, which is the vast majority of the CO checking range. As a general rule do not bet showdown hands into a polarized range.