Challenge completed. Played 6039 1/2 cash hands vs Snowie:







But won't be included in Pokersnowie's "winner list", because for some unknown reason they exclude players who don't play 0.50/1 or who has an "error rate">25.

According to the variance calculator there is a 0.1255% chance that I am break even or losing vs Snowie with this playing style. Conclusive proof that Snowie is light years away from playing gto imo...

The hands was played across four sessions. The screenshot is from the first session.

edit: I lost $6,278 at showdown, and won $10,503 in non-sd pots. Total winrate 35 bb/100.

I just made the following comment on there blog.

http://www.pokersnowie.com/blog/2013...d#comment-6610

I wonder if it will pass their review.

"A user called fontaine has posted on the 2+2 Pokernowie game theory thread that they beat Pokersnowie for about 35bb/100 over about 6k hands. They are playing very lose and aggressive to take advantage for Pokersnowie folding to much. But you won't put them on the list because they have an error rate above 25, and was not playing 100nl."

So when you make too many plays that aren't +EV according to Snowie, you are disqualified, even if they might be winning plays?

If I understand this correctly, that would automatically disqualify many extremely exploitative strategies, which would make the whole challenge pointless. Because exploiting Snowie is, per definition, exploiting a weakness that Snowie doesn't know about. So Snowie would think you made an error. Right?

How have you determinate the BE chance ?
You were aware of the standard deviation?
I guess with that playing style it was extremely high so 0.12% chance of BE is too low of a number imo

Prob 12.55%

I made the calculation with this calculator:

http://pokerdope.com/poker-variance-calculator/

I assumed a standard deviation of 90, and entered a real winrate of 0 and observed winrate 35.

It's possible that I'm a bit off with regards to the st dev.

Yeah, i think 90 is too low for the style you described.

SD=100bb/100: 0.33%
SD=125bb/100: 1.47%
SD=150bb/100: 3.48%
SD=200bb/100: 8.69%

You guys are probably right, 90 might very well be too low. But 200 is way too high. There were very few preflop all ins or huge pots in general. And i used a lot of minbets(!) to control the pot size, as i found out that Snowie responded very poorly to minbets.

fontaine your results seem quite impressive and the pokersnowie team has confirmed your results on their blog too.

If you're strategy really works this well against snowie it would indeed clearly show that their strategy is not working and very exploitable.

Are you sure you did not just run very well? 5k hands is pretty small, however your winrate is sick.

btw fontaine can you explain how you exploit snowie? Or what you think its weaknesses are?

Thanks Ofcourse I can't completely rule out that it was a massive heater. But I lost quite severely in showdowns, I would maybe have expected to do better in showdowns if in fact I was on a heater. Anyway, I am sure the Snowie guys will look closely at my results. It's certainly interesting that it was possible to achieve those results with such an unconventional style.

In general, I think Snowie is folding too much. It is quite easy to bluff scare cards. And minbets get too much respect, Snowie should raise them much more often. I won a lot of hands with medium strength holdings after calling 3bets by simply donk minbetting 3 streets, instead of checking and getting bluffed off my hand.

The biggest adjustment I made was to 3bet A LOT. Especially as the stacks got deeper this was very effective, because then you can flat Snowie's 4bets in position. It seemed like Snowie assigned me a super strong range for calling 4bets. Just for fun I decided to 3bet 100% vs sb open, and Snowie let me get away with it.

It was quite fun to experiment with a style that is completely different from my regular playing style I usually play 26/21-ish at midstakes NLHE. But ofcourse, I would have got completely destroyed if I played my Snowie style in an online midstakes game

I'm not sure but that variance calculator looks like it may be using the old style PT BB per 100 whet 1BB=2bb. If your using at putting in small bb100 it will mess up the results if this is the case.

Really does not matter for the results unless you start mixing bb/BB values.

Same result doing it "manually", winrate=35bb/100, sd=90bb/100:

I played another session last night and lost for the first time, mostly due to shipping 7 bi after making a backdoor flush on a paired board in a 5bet pot lol Probably would have found a fold if it wasn't playmoney I still believe I am a favourite over Snowie, but probably not a 35 bb/100 favourite...

I haven't found anything useful in Snowie yet. The only good thing about it is it's the first player ever I can beat HU for 100bb.

It does thinks a certain % of the time. It might c/c TP 70%, c/r 15%, donk 15% as a crude example.

The point is that snowie team claims that they have found a near gto solution by using all kinds of exploitative strategies against it.

If you are really winning against snowie at a decent rate, whatever they are claiming to have done to find this gto strategy is not working.

I.e. the snowie team should have had a similar exploitative strategy applied vs snowie so it could learn this strategy.

If there are lots of possible ways to play vs snowie that snowie has no experience against than obv snowies strategy is flawed.

Snowie team claims that they have bots that play against snowie in extremely exploitative strategies and then snowie learns from these and finds unexploitable strategies. Ultimately this would lead to GTO.

If you are able to exploit Snowie then either:

a) their whole approach is not working. I.e. they are not finding a balanced strategy in the first place

b) they have not used ur strategy against snowie.

Either case makes the entire thing extremely useless since this means that snowie is very beatable and you can;t use it as ur gto benchmark.

if b) is true which i think is the better case, this would mean that they are failing to find strategies that exploit snowie, which is one of they things we need to be able to count on in order to trust the GTOness of snowie.

I am not yet convinced that you are actually beating snowie as 5k is a very small sample.

It would be nice to see if you could keep up a winrate vs snowie. It would also probably be good to keep your game closer to 100b. I think snowie only works up to 400bb anyway but it seems to closer to 100bb the more snowie is used to it.

I hope we will get some decent responses from the snowie team. i am sure they are looking into a lot.

I'd be interested to hear about any decent player who actually lost vs snowie.

The stats they claim to have are obviously heavily biased as people usually don't play seriously vs an AI bot. It might prove that snowie can beat play money, but I'm not even sure of that.

Yes it plays a mixed strategy, but who it's playing and how they have played any hand apart from the one being played have no effect on how it's randomised.

Oops, didn't realize you were quoting someone when I said that.

Here's an example of where pokersnowie is weak. There's a half pot open on the btn then a pot sized 3 bet from the CO which is called. You can see the range which Pokersnowie will bet vs a check, and what happens on the same flop vs. a min donk.

Pokersnowie should raise almost as much as it's betting here to avoid being exploited.

Why are people helping snowie when if you help them the result is them killing poker? This may have even been their plan...

We'll just move on to Omaha. And after that 5c Omaha. And after that 6c Omaha. Then maybe Courchevel? Who knows. Anyway, poker won't die. Hold'em will.

Correction to my last post it was CO vs. BTN