Can't be good. 0 chance.

as far as the pre flop changes PS 2.0 also seem super sensitive to the stake. If you try looking at pre flop ranges at different stakes you will see different opening sizes.

I have also seem some odd things with the 1/4 bet size as well.

Funny I had different take. I wanted them to add 1/3 pot sizing.

I wonder if there is some reason they decided to use the 1/4? Did they run some analysis that pointed to it being the best choice for some reason?

Here's a hand analysis from the new AI:
It's a hand from PokerStars Zoom

- Button opens 3BB

- I accidently called in the SB with J8 - misclick on my part !
- BB Folds

- Flop comes T73

- I check - Snowie says I should lead out which makes sense

- Button checks

- Turn comes A

- I bet 3/4 pot - Snowie agrees but says I should have gone with a pot sized bet

- Button calls

- River comes 9

- I bet pot with the nuts - Snowie says the optimal bet size for me is 1/4 pot and that I really should check instead. It's expecting the Button to bet so I can raise....that's what happens when I move the hand to the scenario analyzer. Not sure that makes sense as the button won't always bet if I check.

Well, first off even analyzing the hand from Snowie's POV here kinda sucks if you're not supposed to have the hand in your range to begin with.

Donking flop isn't great.

Betting turn seems quite bad.

Checking river is meh, but betting 1/4p is aids. It's either bet large (polarized) or check-raise (polarized). But as played I'd assume a large bet is best, because you're guarranteed value from all AX and < 10 pairs will just fold and I'd imagine snowie doesn't have many bluffs OTR as played + honestly is unlikely to call the x/r.

Look at Snowie's game tree and see how he reacts vs. 1/4p cb, and vs. a check-raise and do an EV analysis quick. I dislike Snowie quite a lot for river plays so I'd imagine you could look at it briefly and see which is best under those assumptions.

The quarter pot thing is very odd. I've only looked at a few spots where it uses that sizing, and it's obviously extremely value-heavy when it uses it (because it doesn't expect to elicit a fold very often). In the challenges I had to keep calling with 4th or 5th pair due to the lolpotodds, and found myself paying off top pair that was going for thin value every time.

With that J8 misclick, it's hard to say how that hand fits into your range, because you probably shouldn't be flatting a 3x open with anything in the SB (up to 100NL, at least). Snowie can't give you range advice, because it uses a 3-bet or fold strat vs that sizing in SBvBTN in the micros.
If you bet 2x pot on the river (instead of quarter-potting), you'll gain more EV than with a PSB, but Snowie rates a check-raise even higher, although you'll only get the opportunity about 35% of the time (It checks back A6 or worse one pairs). Personally, I would jam 5x pot. Betting 1/4 pot with the nuts would make me too sad.

Yeah, well I can state I have not played with Snowie even close to as much as Arty, so maybe he can chime in, but man some of his river plays are just absolutely garbage. The addition of things like min-raises and 1/4psb really makes me question the integrity of the opponent.

FWIW I played multiple matches vs. Snowie (maybe like 5k hands, LOL) and was generally up. However, I did spend a decent amount of time analyzing Snowie's plays and found that they were quite awful on Rivers, multi-way spots, and some turns (specifically coordinated turns).

That said, Snowie was pretty dece HU overall (lolsample), as my above statements are regards to 6m and I felt most of his flop strategies were within reason.

On average when I used Snowie I say I spent 1 hour playing and 2 hours critiquing the value of said plays.

@Brokenstars

Are you talking about the current version or the previous one?

previous

Haven't solvers made Snowie obsolete?

Not really (yet). Solvers can give you advise for heads-up situations, but not multi-way postflop, nor for multi-way preflop ranges. If you do believe your preflop range estimates in HU pots are pretty good, then ofc. solvers > Snowie by far.

Arty, BrokenStars, ETBrooD

Thanks for your clear and thoughtful posts. I use Snowie fairly extensively, along with other tools like Flopzilla, Holdem Resources, and especially PIOsolver. PIOsolver is the most valuable by far.

The role for Snowie is to insure my execution is as perfect as possible. With Snowie I can play 800+ hands an hour in a Challenge and have them assessed for multi-player, pre, and post flop actions. No other tool can do that. I only use 100BB cash play allowing stacks to get no lower than 50BB. I don't think the lack of antes is an issue for improvement of tournament play. I open wider than Snowie suggests and ignore those small EV errors.

I can look at my "blunders" -- I set that in the options at about 0.3BB EV -- and assess whether they are blunders at all. Typically about half the blunders I would call less than optimum play, and in some cases far less than optimum. Many of them are preflop errors given my opponent tendencies and I don't know of another tool that could help me discover those. Others are caused by not paying enough attention to my or the opponent stack size. Some are just plain caused by not having a consistent thought process applied to each decision.

Also Snowie allows me to understand the ranges -- for good players -- in a specific spot preflop and on the flop.

As you have pointed out, there are many places where Snowie plays poorly, espeically as the SPR gets small, and especially river play. I also find it check raises on many turns when leading would be a better play. I imagine it is very difficult to train Snowie for those spots as each one does not come up very often in the training sets. Also Snowie CC's very little preflop.

Snowie flags some pretty interesting spots as blunders, and then I set those up in PIOSolver and look at them in more detail. I always learn something from those.

Some of the spots where Snowie uses the 0.25 pot bet, particularly on the flop, actually make a lot of sense and PIO Solver selects the same bet size.

As far as the 65ss, 54ss recommendations in Snowie, I see PIO choosing post flop bluff combos that have little range overlap with the opponent -- always suited and sometimes strange hands like J4ss. Because Snowie doesn't get to a 3bet flop,or Snowie opens UTG, with hardly any 6s or less -- with exception of A5s, A4s etc , then in certain spots these 65ss hands are reasonable bluff hands.

Here is an example of a poorly played hand by Snowie, and by me, that was flagged as a blunder. By analyzing in PIO I learned quite a bit.And there is liberal use of the 0.25 pot bet in the example below.

Snowie Raises to $2 With AdQh in hijack, I 3bet to $5 on button with AhJd, call. So far so good.

$11 in pot, $90 behind.

Flop AsTd8d

Snowie checks -- I check -- and Snowie says Blunder -- recommends I make 0.25 pot bet -- PIO agrees with snowie

Turn Js

Snowie checks -- PIO recommends Snowie should usually bet pot after check check on flop

I bet $11 pot -- Snowie says 50/50 check or bet -- PIO says I should definitely check

Snowie check raises to $28 -- PIO thinks only a check raise with KQ or QdJd makes sense.

I call -- Snowie says blunder -- and PIO agrees -- I should fold to the check raise.

River 7s

Snowie bets 0.25 pot --same as PIO recommends, I call -- Snowie and PIO OK with my call.

Pio tells you optimal bet size? I thought you had to put the bets in yourself? I guess you could design a tree with different sizes...


If I start playing around with snowie do you guys recommend the one built into holdem manager or the real one? I guess I can use the free trial on both and see what I like better... My post was kinda pointless.

Yes, you select the bet sizes, and then see which ones Pio prefers for which parts of the two player's ranges.

I also use PIO and happened to look at some BB/SB defense ranges vs BTN 3x with 45% of hands. PS way over folds based according to PIO. I think PS was folding BB ~67% and folding SB 84%. Their 3B ranges and % were pretty close.

Over folding blinds v BTN is pretty big issue since it happens so often

PS developers - thought on this and why the 1/4 sizing?

Can you set the rake level in Pio, or is it fixed? I think the rake level can have a dramatic effect on pre-flop ranges. e.g. Snowie's button range and opening size at 100NL are different to its range and sizing at 1000NL.

Yes, you can set the rake amount, and you are correct, it has a dramatic effect on which hands you can profitably call with in the big blind. Also, if you are opening in LP you should open tighter, and make a larger opening raise that what you would open with in earlier positions.

You essentially select in PIO an opening range that is wider that you expect to use for the LP opener and the same for the OOP SB or BB caller. You can run that against 60 weighted flops and then see the aggregate EV for each hand from OOP and IP. From there you can remove those combos that you dont think have enough positive EV to be worth opening or calling -- as you will be removing $6 from the live table so its not worth it unless you have some positive expectation. Then run another simulation to iterate.

As the open raiser you also have to consider the percentage of times you expect to be 3bet, and how often you fold versus call. You can run another simulation with the 3 betting and OR calling ranges set, and with the net on the flop dollars in the pot, and see what your expected EV is for each combo IP and OOP.

A really important parameter is the expected 3 bet percentage from the SB, and from the BB. PIO doesn't help you with that -- you need to decide what you think is a reasonable set of values for that and the relevant range for the 3 better. Of course the SB 3 bet range is weaker than the BB. Now you could use Snowie for that -- see what ranges and therefore what percentage of time Snowie 3 bets in the SB, and in the BB, and add those assumptions into your speadsheet calc for overall ROI, and the Snowie ranges into PIO.

I've performed all of these calcs and created a chart using PIOs excellent color charting capability where each combo is shown as to the expected range of EV. It is amazing how much more valuable the suited cards are. For my casino you shouldn't be calling a 3x LP raise with 100+BB behind from the BB with AT or KJ for example, but you should call with lots of "crappy" suited cards.

Yes you can plug in whatever rake you want. In the example I was talking about I don't know what rake they did/didn't use as it was a training video and wasn't mentioned.

That said PIO was calling WAY more in BB. I think I ran PS at 1/2NL.

Well, bear in mind that you said the BTN was opening 45% of hands for 3x. Snowie doesn't expect the BTN to be opening so wide for that sizing even at 20kNL. I'm sure it would defend in the BB a lot more often if it "knew" that the BTN's range was so wide/weak. From its own simulations/training, it thinks that GTO pre-flop play OTB is a lot tighter than what is widely practiced by regs.

I don't understand why you care what PS knows/expects?

If it has a strong and robust set of strategies it shouldn't need to know exactly what it's opponents are doing. Did I miss this very high fold rate in the blinds or is it new to PS 2.0?

At 1/2NL facing a potsize open from the button Snowie folds a combined 60% from the blinds, 3-bets a combined 25.5% and sees a flop (only from the BB) 14.5%
Betting 3.5bb to win 1.5bb requires >70% FE for immediate profit. Snowie defends 10% more than required and only about 36.5% of his defends are calls. Considering how rarely Snowie just calls and the fact that postflop the winner of the pot has to pay the rake I find it unlikely that Snowie is exploitable using this strategy. In fact it could be argued that by calling more often he might become more exploitable.

The point I was trying to make is that the Pio pre-flop opening ranges you were using/looking at might not be optimal in the first place. (I have no way of proving that, nor do I have any strong reason to believe Snowie's are better).
As I understand it, Pio's post-flop calculations are based on what's optimal for the ranges you enter yourself. If you specify a 45% BTN opening range, it will naturally lead to an 'optimal' BB defence strategy that's wider than if you specify a 40% BTN range. Snowie thinks playing GTO means opening tighter than 45% on the button, and therefore it defends correspondingly tighter. To put it another way, if Snowie thought opening 45% was optimal, it would do that, and then it would defend wider in the BB too.
Presumably, Snowie's "tight" BB defence would still at least break even vs a 45% stealer, but against such a player, it/you could do even better by 'exploitatively' defending even wider (by using the defending ranges Pio suggests, for example).

This observation is especially interesting because at 2/4NL up to 3/6NL Snowie steals ~39.5% with a 1/2 sizing. At 5/10NL and higher he steals ~44% and uses the 1/4 sizing. And as we know at 1/2NL or lower he steals potsize and even less often.
So there are multiple thresholds caused by higher rake where Snowie sees more profit with his range by raising less often and betting larger.

But this does not neccessarily mean that we can conclude the same for ourselves. The network Snowie trains in is improving and removes -EV plays over time. This affects not only Snowie's EV but also the way he achieves that EV. This means that essentially Snowie has arrived at his stake-specific strategies through combined efforts. If his opposition hadn't effectively increased their own EV in a certain way then Snowie would've seen different results and he might've arrived at different stealing strategies at each of the stakes than he has now. I have no idea.

Who won in the match between Snowie and Jungleman?

Good question. Afaik, results were never released.