Inside PokerSnowie's brain reveals the work of the Snowie AI Team. It explores first hand how the brain of PokerSnowie evolves and learns advanced strategic concepts, on its own.

PokerSnowie's ultimate aim is to produce the perfectly balanced game, find the ultimate un-exploitable equilibrium for all No Limit Hold'em configurations. Join us on this fascinating journey, which is just starting, into the future of poker.

The PokerSnowie Video Quiz series aims to answer the question: "what is the right play?" It is a set of poker coaching videos presented by French professional poker player Sharp. In each video, Sharp sets up an interesting hand in the "Scenarios" tool of PokerSnowie and explains how to analyse the situation and learn from PokerSnowie's advice, based on the Game Theory Optimal model.

Challenge PokerSnowie: 1 month, 2900 challengers, €16 million profit

Jan 17, 2014  |  Posted by PokerSnowie  |  34

PokerSnowie wins 21 bb/100 against 3000 challengers

Since the 12th December, nearly 3000 challengers have taken on PokerSnowie, playing almost 2 million hands against it.
 
Had this been a real money challenge, PokerSnowie would today sit on a very comfortable booty worth €16 million. Not bad for a month’s work.
 
The precise figures stand at 1,852,809 hands played, with an average error rate for all players of 16.89 and a global winrate for PokerSnowie of 20.83 bb/100.
 
When concentrating on the 1500 challengers who have played at least 100 hands, and with an error rate below 25, we find 1,577,227 hands played and a profit of 14.11 bb/100 for PokerSnowie. That would still have translated into a generous €8.5 million payday.
 
Let’s dig into some details behind those figures
 
Unsurprisingly, PokerSnowie continues to win more against its weaker opponents. In the last update, we had a low number of extra-terrestrial level players, but none have managed to maintain this level. As a result, we do not have any relevant data anymore for this level. For all other play levels, PokerSnowie has consolidated its winning position. 
 
PokerSnowie's win rate by Error Rate level
  Hands played Error rate PokerSnowie's win rate
(in bb/100)
Extra Terrestrial
(0-4)		 less than 200 NA NA
World Class
(4-8)		 170,064 6.56 +1.94
Expert
(8-12)		 456,876 9.89 +5.47
Advanced
(12-16)		 389,258 13.71 +10.63
Intermediate
(16-20)		 336,386

17.76

 +25.17
Beginner
(20-25)		 216,063 21.77 +32.04
 
We’re now also able to detail the results by game format and level of play. As a result, each line has a lower number of hands and hence some figures will be strongly affected by variance. We have added a variance indicator to PokerSnowie’s win rate, based on a 95% confidence interval:
 

 

Challenge PokerSnowie results by game format
Game format Play level Hands Played PokerSnowie's win rate
(in bb/100)
Head's Up
 		 World class 54,368 -5.19(+/- 9.8)
 
 		 Expert 144,096 +5.66(+/- 5.7)
 
 		 Advanced 149,910 +18.63(+/- 5.7)
 
 		 Intermediate 136,916 +32.86(+/- 5.7)
 
 		 Beginner 97,367 +54.67(+/- 7.0)
 
 		 Aggregated result 671,984 +27.30(+/- 2.7)
6-max
 		 World class 111,414 +2.53(+/- 4.7)
 
 		 Expert 236,826 +4.73(+/- 3.1)
 
 		 Advanced 180,205 +5.24(+/- 3.5)
 
 		 Intermediate 110,712 +14.56(+/- 4.5)
 
 		 Beginner 71,307 +18.65(+/- 5.6)
 
 		 Aggregated result 801,465 +12.83(+/- 1.7)
Full ring
 		 World class 7,909 +8.14(+/- 14.5)
 
 		 Expert 79,349 +4.43(+/- 4.6)
 
 		 Advanced 68,529 +6.45(+/- 4.9)
 
 		 Intermediate 31,723 +9.39(+/- 7.4)
 
 		 Beginner 51,314 +30.94(+/- 5.8)
 
 		 Aggregated result 299,778 +20.11(+/- 2.4)

 

PokerSnowie exhibits even stronger results head’s up, although it did lose against world-class players. (Note however that variance can affect this result either way)
 
Are PokerSnowie’s winnings always consistent? No. In fact PokerSnowie had its first losing day on the 16th January and we’re expecting that; as the quality of opponents increases, the win rate will fall slightly.
 
Pokersnowie's winrate per day
 
Finally, there are 56 players who individually played in excess of 5000 hands against PokerSnowie. They played at an advanced level of 14.16, with over a total of 540,648 hands. PokerSnowie’s result? Winning 9.85bb/100.
 
Whilst PokerSnowie is demonstrating its strength at the table, we would also love to hear how it’s also helping you progress. Please tell us your experiences using PokerCoach and Challenging PokerSnowie to test your game.
 
New features will be added to Challenge PokerSnowie in the next few days, in particular a shortstack and deepstack option. What features would you like to see? How can we make Challenge PokerSnowie and PokerCoach even more useful for you? Please let us know in the comments or send us your thoughts to support@pokersnowie.com

Comments

Carlos (not verified)

Fri, 01/17/2014 - 20:54

Yes, please include bb/100 on the profit report.

PokerSnowie

Sat, 01/18/2014 - 18:59

Thanks Carlos, well noted.

Jay Cordes (not verified)

Sun, 01/19/2014 - 11:25

Love these posts...so much interesting info. For example, there are so few players at "world class" for full ring games; maybe this is the format in which humans have the most to learn from PokerSnowie? I also love to see the validation of PokerSnowie's rating of the human level of play, with its clear correlation with win rate. I'll be very interested to see if humanity continues to improve and reduce PokerSnowie's daily winrate as well as whether or not the world class players can continue to win at heads-up. The only other thing that would be fun to see is a running top 10 players list (over 5k hands) for each game format, listing the highest current win rates against PokerSnowie as well as the least error-prone players. Each game format could be considered a distinct 5k hand challenge and be tracked independently. If anyone plays at least 5k games in each game format and wins in all three, now THAT would be truly impressive.

PokerSnowie

Mon, 01/20/2014 - 20:35

Thank you for the feedback Jay. The list of top players would be interesting indeed, and we are considering what we can do to enable this in the future. Many thanks again for your feedback.

black (not verified)

Tue, 01/21/2014 - 18:03

Jay make some realy nice suggestions. but how long it tooks to implement some of them? You need so much time to bring the pokertrainer online that i see no chance that you implement Jay`s ideas this year...

Fred Aces (not verified)

Sun, 01/19/2014 - 13:23

In the last "Challenge PokerSnowie" update, there were 18 players that were winning again PokerSnowie. How many players are now ahead of PS? Also, I'm one of the players that played more than 5K hands (at Full Ring) and winning again PS for about 2.0bb/100. How do I rank among the current winning players? Thanks

PokerSnowie

Mon, 01/20/2014 - 20:25

Hi Fred, there are 17 players ahead of PokerSnowie in that group. Amongst full ring players having played in excess of 5000 hands, you actually have the best result in terms of win rate (about 6 bb/100) and are one of 3 winners only.

Carlos (not verified)

Thu, 01/23/2014 - 08:48

Nice job Fred. I too get excited to see my name mentioned amongst those who are winning against Snowie. I have played 10,000 hands of 6 max and had an error rate of 8.86 and a winrate of 6.31 bb/100 at the 1.5 million hands update. I was going to wait the next update to see where I stand now, but since you asked I guess it is ok. Pokersnowie, can you tell me my latest numbers? I tried to calculate my winrate, but it is way off from what you reported last time. I am doing something wrong. Also, where do I rank amongst players who have played 10K hands of 6 max vs Snowie. Thanks. PS: The idea of a leaderboard would be a great marketing tool considering how excited we are about hearing our names mentioned.

PokerSnowie

Thu, 01/23/2014 - 10:24

hi Carlos - As of today, we have recorded 10029 hands played, for a win rate of 6.15 and error rate of 9.05. Considering only the 6 max tables, you have a win rate of 5.62 for just above 7700 hands played. if we had a leaderboard for 5000 + hands players at 6 max-tables, you would currently stand in 7th position. Do you feel PokerSnowie is helping your game overall? Would love to have your feedback.

Carlos (not verified)

Thu, 01/23/2014 - 22:51

PokerSnowie is having a dramatic impact on my game. By strengthening my opening ranges, it has helped me to get into less trouble post flop. Also by encouraging me to sometimes 3 and 4 bet with the proper ranges, it is preventing me from being a nit. I have a decent understanding of preflop play at this point. Now, I am seeking to understand flop play. Most of us who are winning players have learned to c-bet too often because few opponents will raise these c-bets in practice. PokerSnowie is helping me to learn when to not c-bet on some flops to avoid this. I am still having trouble with turn and river play, but soon I will improve here as well because I practice with PokerSnowie every day.

PokerSnowie

Fri, 01/24/2014 - 10:51

Thanks for your feedback, we're really glad to hear how it's helping you. Best of luck at the tables.

chrono (not verified)

Sun, 01/19/2014 - 18:47

The single most important upgrade you can make is to allow for the user, starting from all available actions, to think as abstractly as possible in terms of decision trees through the Scenario Builder. For example, if the user wants to construct a Check Raising scenario, he or she would want a reasonable amount of different flops where Check Raising instead of any other available action would be ideal, or where Check Raising and other available actions are equally ideal. Then, he or she would want to know which specific flops are better for which specifically different raising amounts for a Check Raise. Then, assuming he or she selects a specific flop, he or she would want to know what possible turn cards would make Check Raising be definitively or equally ideal, and then which possible turn cards would be better for which specifically different raising amounts for a Check Raise. He or she would then also want to know what possible turn cards would make Betting ideal (and then if Betting is selected, which possible turn cards would be better for which betting amount), or what possible turn cards would make Checking ideal, or what possible turn cards would make Calling ideal (and for what betting amounts). The same thing can then be extended to the river.

chrono (not verified)

Sun, 01/19/2014 - 22:23

Here's the second suggestion, which I believe is even better than the first. Add a new function that is similar to Challenge, except that PokerCoach is telling you on the sideline which action to make, and by what amount if betting or raising. Just by making the user do the action (and by what amount if betting or raising) him or herself against PokerSnowie, it allows the user to memorize by sometimes approximate and sometimes precise repetition over the course of tens of thousands of hands what PokerCoach considers to be the optimal game. The exponential rate at which the user will improve will be truly incredible.

PokerSnowie

Mon, 01/20/2014 - 20:27

Chrono, many thanks for your many suggestions and ideas. We are actively working on several of the ideas you suggested.

gdino (not verified)

Mon, 01/20/2014 - 01:26

A suggestion for the developers. Playing the challenge is fun but it gets a bit boring over time with all the folding, especially full ring. Can we have a functionality where we set it up so that we get dealt a top x% of hands and at least one opponent gets dealt top y% so we are more likely to get action to our pfr and at least see a flop. Can start as simple as that to later evolve into a real focused coaching.

PokerSnowie

Mon, 01/20/2014 - 20:28

Thanks gdino, excellent suggestion, well noted.

alex (not verified)

Mon, 01/20/2014 - 06:25

hi i dont understand what do you mean that snowie won for emapla 4.73bb/100 in 6-max against expert level, which one of the five snowies in the table won that? all together, average? or snowie was playing in a table against 5 expert players? thank you

PokerSnowie

Mon, 01/20/2014 - 20:33

Hi Alex, this is a very good question and we are guilty here of a small confusion. The win rate we're showing is in fact the loss rate of the players against PokerSnowie. In aggregate, it's the same thing (i,e, world loses by 20bb/100, PokerSnowie overall wins by 200 bb/100). But in your example, the expert level players lost 4.73bb/100 in the 6-max format and each PokerSnowie (i,e, 5 in this case) has in effect won 1/5 of this. Hope this clarifies the figures.

MEGA (not verified)

Mon, 01/20/2014 - 16:06

I'd like to have a button that will hide/unhide Snowie's advice when reviewing a session. I prefer to analyse the hand for myself and then see how Snowie analyses it.

PokerSnowie

Mon, 01/20/2014 - 20:34

Thank you Mega, suggestion well noted.

Kyle (not verified)

Thu, 01/23/2014 - 23:29

Hello guys, do you have any plans to make the software available for Mac? Thanks in advance.

PokerSnowie

Fri, 01/24/2014 - 10:53

Hi Kyle, In the short term, a mac version is out of scope. PokerCoach does work fine on a windows emulator for mac though.

Oliver Heuler (not verified)

Fri, 01/24/2014 - 22:31

Which Windows emulation on the Mac is suggested (has been tested) for running Snowie?

PokerSnowie

Wed, 01/29/2014 - 10:29

We have received positive reports of users using Virtualbox and Parallels.‏

Terry Linhardt (not verified)

Sat, 01/25/2014 - 20:23

You requested some suggestions for the Challenge game. Here are some things I would enjoy seeing: 1) It would be nice to be able to configure *each* individual opponent in terms of their "reload" play. For instance, it would be nice to have some players always "top off" to 100BB, while another tops off, for instance, when their stack shrinks to 50 BB and still another who tops off when their stack is a paltry 20 BB. 2) A "side bar" that would reveal PS advice. I realize the problem with this is that it could really wreck your stat gathering. An alternative would be to have the advice given immediately after an error. In short, I make a decision, execute it, and PS would *then* advise me. 3) Truthfully, I'm not quite sure how I would implement this: But I sometimes find I have a dozen errors pre-flop, but upon closer examination the error rate for all of them are pretty trivial. Or I can have ONE error which is far greater than another 10 or 12 put together. SO, it would be a nice feature if I could quickly find the significant errors. Now, I realize an error is an error. But some are just a lot more expensive than others. Of course, the really expensive ones we call blunders, and it's easy to find those. but I would like, I guess, an additional option to filter through "non-blunder" error to find those that are more than trivial, but not at the blunder level. Hope this makes sense. It would also be nice if I could configure this additional filtering level. For instance, I might want to look only at errors that are > .5 or > .8 BB. You get the idea I hope. As a side note, I love the software. I play the Challenge and find I am improving. And I run all of my on-line action through Snowie to get feedback. Keep up the good work.

Terry Linhardt (not verified)

Mon, 01/27/2014 - 21:44

Regarding my suggestion #3, I realized that I could accomplish what I wanted by going into "options" and setting my own "blunder threshold" to some value other than the default. For instance, at the default of 2.0 BB, I frequently get 0 blunders, but a fair number of rather trivial errors (.01 or .05), and also a number of "middling" errors (.4 and .5). By setting my blunder threshold to .4, for instance, I obviously get more "blunders." But I can concentrate on my more expensive errors and skip past those that are of fairly minor significance.

PokerSnowie

Wed, 01/29/2014 - 10:33

Thanks for feedback Terry. Your second suggestion is something we're actively working upon, and hoping to release in the near future. On point 3, besides changing the blunder thresholds, we will also improve the error filtering for an upcoming release. Thanks again for your comments and we wish you continued success at the tables.

Ivan Solvason (not verified)

Sun, 01/26/2014 - 16:32

I have enjoying playing against Poker Snowie. I am amazed how much it can bluff seriously with AK - hitting nada. Leak in the program is definetely to hit second or third pair. Then raise a bit every street and raise a bit more on river. PokerSnowie refused to fold the hand. Loosing a lot of smaller pots. Also I nailed it numerous times when it was bluffing. I have played Ring games and have ZERO chance to win. Heads Up was the only thing where I could beat up PokerSnowie on especially 100/200 Freezeout. Most frequently ended as Expert and on one occasion world class. Did not calculate outcome after all the battles but I am 100% sure I am far ahead. All in all - I think it is a very good training tool for any class player, and the human brain can always outsmart a program, switching from loose to tight frequently, setup traps with trips - trying to take down a two pair for Snowie with massive pots. Have also tried freqently to bluff and it works very often with big raises, but have also been caught with my pants down here and there of course. But fun to play for sure. Have better results in tournament play with a 23% for January ITM (best ever in 5 years) - so perhaps something have added to the mediorcre skills I have. Think you should move away from payservices, since my guestimate would be less than 1% would actually be paying for this. Find other ways like affiliate deals, advertising etc.. just my 5 cents.

Terry Linhardt (not verified)

Mon, 01/27/2014 - 21:36

Ivan: If I read your comment correctly, you seem to be saying that PS can be exploited through traps, over-bets, etc. If PS is GTO, your attempts to exploit would fail over time. Here is a question for the developers: Would PS change it's style of play if all cards were face up? My understanding of GTO play (and my understanding may be flawed) is that PS could play with it's cards face up, and it could not be exploited. Yes? No? Obviously there is an immediate issue of whether or not call a huge bluff. If the cards are face-up, the opponent would know whether a call or fold is appropriate. Comments ?

PokerSnowie

Wed, 01/29/2014 - 11:20

The 'cards face up' principle needs to be applied correctly. If PokerSnowie shows its cards at the beginning of each hand, it would certainly be easy to beat. The correct principle is, that a perfect GTO system could show its range at any time during the hand and not be exploited. For PokerSnowie this principle is the same, with the important difference that PokerSnowie is not playing GTO perfectly. So if you saw PokerSnowie's hand range all the time, it could still be beaten.

PokerSnowie

Sun, 01/26/2014 - 17:59

Official winner mention: Natalie played 6500 hands at an expert level of 8.58 and overall win rate of 6.12 bb/100. Very well done!

Kevin (not verified)

Tue, 01/28/2014 - 21:08

Hi, Thanks for integrating a short/deepstack option, I think it's an awesome idea! When will it be released? Also, will there be a selection of a few different stacksizes(for example 500, 300, 100, 40, 20) or will you be free too choose whatever stacksize you like from 1-500BB or whatever the limits are of what snowie can play? I think the latter would be better......but either way I'm excited about it! Thanks again, keep up the good work! CHeers

PokerSnowie

Tue, 01/28/2014 - 21:19

Hi Kevin, we will release the new starting stack options on Thursday 30th this week - you heard it here first! :-) This first version will only have a 50BB, 100BB and 200BB options - but we will be adding more options in the next few weeks, possibly a manual input choice as you describe. Stay tuned!

Kevin (not verified)

Wed, 01/29/2014 - 01:17

Thanks for the quick reply! 50BB is still too deep to train for my games unfortunatly :/......adding a 25BB option would be cool! I can wait though, or try and let my stack run down :-) Cheers!