It is not about math logic what Snowie is doing...and as their creators said,when ranges come extremely short some recommended action is not always correct.

Snowie is not calling the river against snowie's range . It's going through all it's experience of calling this river against all the "exploiting agents" or w/e it has trained against . And from that experience it tells you that calling here has lost 45bb on average .

OK, sure, but isn't the issue here that Snowie's overbet shove range is crazy? If I'd been in Snowie's position, it'd be telling me that I should be bluffing with this mad losing range. This isn't some wild multiway hand involving weird bet sizes or a one-in-a-million board, it's all pretty standard and unremarkable right up until the point that Snowie decides to blast off into the stratosphere with ace high. If Snowie goes crazy in fairly straightforward spots like this, how can you rely on its advice in other spots? This is far from the first instance where I've seen Snowie make an unexpected big bet with a hand that I'd not really want to play that way - I initially thought "OK, there's something for me to learn here", but now I'm wondering whether that's actually true, or whether Snowie just plays badly on big rivers and turns.

It's 200%pot size bet that is optimal for his range because it has full house and a straight as value hands or A high air hand with some blockers to hero value hands that bet twice and checked the river.Hero's checking range probably doesn't has boats in it.And there is no point to bet smaller to induce a call from 2 pair type of hands or other bluff catching combos so Snowie is betting 200% but since SPR is betting all in.

Sure, I get the reason for having a nuts-or-air overbet shove range, and I'm not complaining about that per se. What's concerning me is the combos it's putting into that range, and the very unbalanced bluff to value ratio it ends up with. If it was only using, say, its four combos of AJs as bluffs when it does this, I'd have no problem. The problem is that it wants to do this with over twice as many bluff combos as value hands.

Probably it learned that is getting more fold equity and is making more money over betting in that spot since it has unbalanced range bluff to value combos.

It's a move of a maniacs players that u can find at micros.Move sometimes used by LLove at tournaments in multiway spots(4bet rejamming 100bb deep with QJs)that even high stakes regular tournament players r complaining

This kind of spot has been posted as an example of Snowie's weirdness and generally unbalanced river play several times in the thread. I think it's mostly just a sample size problem, due to the way Snowie trained, as Frogman mentioned. During its "reinforcement learning" phase, it just can't have experienced every possible action sequence with every possible combo vs every other possible combo in order that the river EV is correct.
It's not like with a solver, where you have the exact ranges - and frequencies with which each combo appears at that decision point - and let the software calculate a near-GTO solution. Snowie had to "play" billions of hands against "random" opponents. On that particular runout, it must have done quite well with its bluff shoves, so it kept doing them, but it did badly with its calls of shoves vs the agents it trained against. The game of poker is just too big for a neural network to have found solutions for every river card on every board vs every possible "random" strategy.
As has been said before, you can use Snowie to draw some general conclusions about optimal ranges, particularly on the earlier streets where Snowie's EV estimation is a bit more trustworthy too, but its river play is sometimes quite weird/wrong, and the numbers don't "add up".
Don't lose sleep over its percentages being unbalanced in one specific spot imo. You're unlikely to ever have that exact runout (and those exact combos) in real life. (If it came up again, you could 'exploit' Snowie by calling wide, but in other spots that look quite similar, you might find Snowie is considerably under-bluffing. Snowie's "mistakes" probably kind of even out in the long run.)

Fair enough. So what would you say is the best way to study using Snowie, then? Just disregard all its plays on rivers and big turns, and focus on early street strategy? I got the cheapo sub a week ago and have just been playing pretend 6max 500NL against it, which has been interesting but I'm not sure it's the optimal way to take advantage of it, or what to make of the quality of its play in general: over 8400 hands, I'm beating it by more than 32bb/100, which seems like an implausibly large amount for a recreational microstakes player against a genuinely strong opponent, even after taking variance into account.

I´m using snowie, especially the training tool as you seem to be doing, and I´m aware of its weirdness on later streets

Just my 2c, winrate over 8400 hands means nothing, real opponents or vs snowie. Disregard these results. If you can keep a big winrate after, let´s say, 50k hands or, better yet, 100k, then you may have find a way to exploit the bot. Until you get a reasonable sample size, it´s just variance.

Cheers

Is there an issue with using Snowie to extract optimal preflop ranges vs unique sizings that Snowie isn't regularly using? For example, RFI v OOP 3B, I presume Snowie has played this spot mostly with a pot sized 3B option. If I force Snowie to 4x 3B, are the ranges used (both 3B and the defense) of the same integrity considering that it hasn't practiced these sizings as much as others?

Max

I would assume that its advice will be "less good" against uncommon sizings that it didn't face much in its training period. It used to say on the website (and may still do so) that Snowie was/is vulnerable to very large overbets post-flop too, because it didn't do much training against them. Until it trained with the quarter pot sizing, it was also exploitable vs very small bets. (An early challenger crushed it by playing 100% of hands and mindonking a lot, since Snowie basically treated a minbet as if it was a half pot bet.)
Unless Snowie had significant training vs 4x 3-bets, I would guess that it treats a 4x 3-bet either the same as a pot-raise, such that it doesn't fold as much as it should, or as if it's 2x pot, in which case it would fold more often than is optimal.

thanks Art

Problem is vs most opponents, regs and fish included, when someone calls twice on T55Kx board or whatever, and then overbets 2.3x pot the river, calling with KQ vs most opponents is gonna be lighting money on fire. Not enough people are turning Tx or 88 into a bluff to justify calling with such a weak hand. It's not a standard bluff spot for them.

Great posts by Arty.

Is there any reason there is so many cold 4bets in 6 max from the "AI" that folds 95% of the time to 5bet all ins? That can't be right.

In a real game a cold 4bet would fold to a 5bet all in maybe 5% of the time in almost every game that exists today.

I understand the AI is far from perfect, but what is the point of practicing against it if these situations don't come up nearly as much as the AI throws at you. If anything it's encouraging you to 5bet shove almost all cold 4bets and light your bank account on fire.

Snowie is a decent tool for specifically getting some reps in heads-up. He's a pretty good opponent and it's nice to have an always-available opponent since online HU games are basically non-existent, at least on my sites, and doubly so for the micro stakes.

In my perfect world for my current situation, having a steady supply of 5nl-10nl HU tables against real people (or even bots for that matter) would be ideal so as to provide a reliable method of quick, on-demand practice to reinforce the work I do "in the lab". But alas, the world ain't perfect. So Snowie is a good alternative.

Hence my one complaint about Snowie: it would be an even better tool if they implemented an optional shot clock to further simulate real competition.

Hi,

I guess PokerSnowie support team is not involved in this thread right?

I have a question regarding https://www.pokersnowie.com/pftapp/

It seems that the charts we can find there are very different from the actions suggested by the PokerSnowie software.

For example here (https://www.pokersnowie.com/pftapp/#bet3_aftermy_raise) when we are BTN vs SB, it is said to call with K7s (!).

The PokerSnowie software says it's a 100% Fold. It's also a 100% Fold with K8s and K9s sometimes (depending on the stakes and 3bet sizing).

So how can information be different in their online advisor VS their software ???

The online advisor (and presumably the mobile app) uses the old AI and was based on either 100NL or 200NL before they even trained the bots with the newer betsizes. The subscription-based version has a newer AI and will give different results depending on which format you're simulating. The "paid for" version is likely to be more accurate.

Is Snowie still playing itself?

Thanks, but I am asking if Snowie is still playing itself back at Snowie headquarters, trying to improve its play.

Like trillions of hands r not enough?

well... no

Don't underestimate this software . and don't over estimate it either . It did help me correct some fundamental errors I was making but its limitations for analyzing downloaded hands are a little limited and will give you a very high hand score in the vent you play 100 hands and are basically card dead and find yourself playing peek and fold you can be rated extra terrestrial just because you hand to fold 90% of your hands . By the same token it can grade you as a beginner just because you stood up to the table captain and played back at him with QQ or JJ . At the very least it does give you a comfortable way to analyze hand histories that we should be doing anyway just don't expect this to turn you into a world beater its a tool to help you solidify your fundamentals not much more

Well Educa-poker cash course for Upswing is 50% based on Snowie so people should take it seriously.In my opinion Snowie is the fastest and cheapest solver out there because is based on data and practically u don't need a powerful machine to run it.The only thing is that everyone should learn to use it properly. Like,they say it folds too much when analysing hand history. That's because bot is giving u an analysis of his own play/ranges.It folds because there r no combos to call or raise in a particular spot.If hero or the villain makes a mistake in some spot, the analysis after that is useless. (For example u raised pre with combo that Snowie is folding 100%...analysis after that will not be so relevant ).