Unlike solvers that can be programmed to choose from multiple sizes and multiple sub-ranges ("range-splitting" I think it's called) in order to marginally increase total EV, Snowie chooses one size that maximises the average EV of the range/strategy. It chooses a single size that works best for its range, not just one hand.
While it could increase the EV of JJ by betting bigger, some of the other hands in the range would lose EV if they also bet big, so Snowie picks the size that gets the best average.
I think a closer approximation of a GTO strategy would balance several sizing strategies (with a slightly different range for each size) and use them at different frequencies. e.g. Sometimes bet small, sometimes big and sometimes huge, but always balanced.

2 noob questions.

1.Does Snowie show EV in terms of multiples of big blind? ie. does EV of 0.5 mean 50% of bb? If not what does it mean?

2. I'm a bit confused as to what Snowie considers a pot size (or any other default size) raise. Eg. at 1/2 BTN opens to $6 and if SB wants to make a pot sized raise it 3bets to 20 (when pot is 9 before the 3bet). For the BB (assuming SB folds) it's a 3bet to $19. Why is this?

1. Yes. EV is measured in big blinds.
2. Google "pot-limit calculation". When you make a pot-sized raise, you include your (imaginary) call in the size of the pot before adding to the current betsize. The SB has to call 5, so a pot-sized raise is to 9 + 5 on top of the 6 that villain used. The BB only needs to call 4 to match the initial raise, so a PSR for him is to 9 + 4 + 6 = 19.
In both cases, the villain gets 2:1 odds on a call (just like when you make a pot-sized bet post-flop). When SB makes it 20, the pot lays 28:14 on a call for the initial raiser. When BB makes it 19, the pot lays 26:13.

Thanks guys.

Does anyone have any idea why at the challenge it suggests to open x3.5 from the btn?

In simulations of games with high rake (sub-100NL), it prefers a large size with a tightish range OTB, because it apparently calculated that more EV is generated by stealing the blinds with about 38% of hands than by seeing flops (and getting raked) with a wider/weaker range of 40%+. In short, when rake is high, more of your EV comes from pre-flop fold equity, and larger sizes help increase it. At high stakes, where rake is minimal, you can play looser, and minraising (and defending super-wide) is de rigeur.
Curiously, there is no rake in the challenges (or at least there wasn't when I was a subscriber), so Snowie actually plays too tight in them.

Thanks for your answer, that makes sense! How come though it isn't suggesting the same sizing for the other positions?

It's complicated and goes counter to what many players do in real life. (Many people open larger in EP and smaller in LP, but they also open tighter than Snowie UTG, and wider on the button).
When you open in earlier positions, you're more likely to get called (or 3-bet) and have to play OOP. Since getting 3-bet or playing OOP is disadvantageous, it's apparently better to risk less of your stack with your initial bet when you can't guarantee you'll be in position. The same thing applies post-flop. (OOP players should often c-bet small or check, whereas IP players can make larger c-bets and at a higher frequency).

makes some sense at least in my games running reports with the stat "face 3b" usually are higher in ep, so i think we also lose less when fold to 3b by raising smaller (?)

Hey,

Apologies if this has already been posted somewhere but i'm currently trialing Snowie for HU and maybe some 6max situations.

So far i'm beating it quite convincingly heads-up but i'm yet to play it at 6max.

If Snowie was put into Stars, which stakes (for 6max/HUSNG) do you think it would be profitable at? I know this is a kind of situationally dependent question but just looking for a little more about Snowie's ability.

Cheers!

^^

While I can't answer this question (and nobody might be able to), it does seem that Matthew Janda from CardRunners seems to respect Snowie's results quite a bit, as he frequently uses it in his videos to illustrate / validate concepts.

I figured this may be the case... perhaps a better question to ask then is, would the average NL10, NL25 etc player be able to beat snowie? I'm surprised there isn't any research into that... unless they're scared of the outcome? lol

How would you conduct an objective "research" though? Letting the bot run on a poker site is out of the question obv ... and what "average NL25" player would play the bot for long hours on end unless he's being paid?

There was something like "Snowie Challenge", and you can read about it here:
https://www.pokersnowie.com/blog/201...s-hands-played

Which sample size was this? Jungleman had good things to say about its HU game.

Snowie is stronger HU than at 6max; the latter's game tree is just so much more complicated.

It's a great tool, but you have to know the assumptions when looking at its advice, i.e. that all players are playing like Snowie (tight but fairly balanced).

Is the only reason Snowie doesn't change its 3bet size based on stack depth because it's bound to fixed sizes? ie. it can only raise a 3bb open to 9bb because this is exactly a pot raise?
In reality you should 3bet bigger than this 200b deep, correct? What's a standard size assuming a 3bb open and a 2.5bb open?

Correct, though it will change its range for that fixed size as stack depth de/increases.

Not necessarily.
Pot.

Thanks guys

Not a huge sample size, granted. I just play it in 'challenge' heads up SNG with 100bb starting. I usually achieve mid-World Class ranking, occasionally high-intermediate.

I usually play 6-max/FR, so it's nice to see my HU skills aren't too shabby

I've messed around with snowie a little bit and it seems either too tight or unbalanced in certain situations 6m, so I haven't tried going deeper with it. It doesn't seem to value a preflop range with good coverage and I've seen some odd 3b/4b scenarios where snowie only has AA, or it's only flat from the sb is 66 which has to be flawed for obvious reasons

The proflop ranges are as good guesses as any top guesses. It basically 3-bets or folds like a limit player (including the calls in the BB). Postflop it is similarly constructed GTO (guesses) and there are major weaknesses. When I played it, it sucked more than fair share.

Is possible to keep in consideration to introduce another time the Monthly Subscription?

I don't remember - Can PS runs sims for NLH with antes?

Does anyone know this?

Does Poker Snowie take into account prior street actions when giving advice?

For example, I bet the flop and turn (Snowie said I should check in stead). Then on the river Snowie says I should jam pot sized bet 100% of the time. Is this taking account of my betting in prior streets? or is the recommended action "in a vacuum" and based purely on the my hand and run-out.