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.
My Head's Up experience with PokerSnowie
I was approached by the team behind snowie a couple of months ago to have a look at what they had developed. I was only interested because of the snowie name - they had solved backgammon, a game I used to love. Now they had taken poker on but poker is a completely different game to backgammon in the fact it is a game of incomplete information- they couldn’t have solved poker, it’s impossible. I told them how sceptical I was and stated if it’s system based then it’s exploitable. I was involved in the development of the early short stacking systems years ago which led to the influx of players “ruining” the regs fun. I also used to do data analysis on millions of hands a week so I thought I knew what I was talking about- their mouths were making claims their arses couldn’t cash.
I was given access to the prelaunch preflop advisor and then the Pokercoach/Pokertrainer app. I played some hands of poker for the first time in three years with my own money and won a little, put the hands into PokerCoach to be told I had made loads of mistakes. I won so how can I be making mistakes. I disagreed with a lot of the advice it was giving me- snowie was wrong. I decided to prove it was wrong by playing the snowie way over a lot of sessions…. I won again damn!!! I got to playing at an extra-terrestrial level according to PokerCoach and was now winning 90% of the sessions I was playing. I still disagreed with the system though (poker brain saying this is a mistake according to snowie but I have to continue in this hand). After logging plenty of hands and analysing the hands it said were mistakes I found I had won some big pots I wouldn’t have won as snowie said I should have folded, unfortunately I had also lost some….. net result I lost more than I should have and if I had played the snowie way I would have won more overall. I hate being wrong more than anything and tried to explain it with variance, running bad, my opponents were fish, etc.. ( you can take the poker player out of England blah blah blah).
I decided to then focus my efforts on playing snowie direct on the poker trainer. This software lets me play from heads up to full ring games against snowie. This is obviously in a virtual environment/play money if you prefer as snowie is not allowed to play on any sites and is a training tool. I played thousands of hands trying to find weaknesses/leaks. I found what I thought were a few, but overall it was tough- the thing just doesn’t make mistakes and plays exceptionally well post flop. A challenge was then set by the team after I claimed I could exploit it.
Challenge 100 bb heads up. When one player has the lot start again, best of five format.
I started with a 50 minute game and pounded away at snowie. I had him on the ropes and set him up in the hand for the allin on the river when holding second nuts (he could not have the nuts as it wasn’t in his range). I was up against a system so I knew his range and I knew he would call as the board had hit his range meaning the minimum he could have was top 2. He had 54 left on the river and I had him covered so I bet 47 into a pot of 42….. wtf was I thinking???? He snap called and was left with $7.
</p> <p>This is human error/ showing off/ whatever you want to call it- I can’t mock and rub down a computer program so why the hell did I bet $47 instead of allin…. 10 minutes later I was tilted as anything after he took the first game. He won a few allins with hand v hand but the first game loss was purely down to my stupid mistake. I took game two easily as I had found an area it was playing incorrectly. He took game three in a cooler and I took games four and five with ease. I was the first person in the world to beat snowie (it has played far better players than me and crushed them). I was confident I was superior and had proven it could be exploited. Unfortunately the stats proved I had run well and got lucky in some key situations. I was so confident I agreed to a rematch over a bigger sample size… sigh.</p> <p><strong>Challenge 2 100bb heads up. Best of 11</strong></p> <p>Hand 3 of the first match snowie limped the button. In thousands of heads up hands it had never limped the button before. This negated one of the main ways I was exploiting it. The AI updates itself every few days having run millions of hands. If it spots anywhere it feels it could be exploited it corrects itself and changes the system….shit, this isn’t gonna be easy.</p> <p>I don’t want to go into too much details but I had my arse handed to me on a platter served with a side of embarrassment. The stats showed I ran as unlucky as I had ran good in the first challenge but a 9-2 whipping was delivered. Even if I had ran like god I honestly believe I would have still lost.</p> <p>The real worrying thing about the Pokertrainer is the ease with which it can teach you how to play just like snowie. Imagine having the world’s best player sat on your shoulder telling you when you make a mistake as you make it and why it’s a mistake. This perfectly balanced player says you need a minimum of x to play in this situation. You learn very fast and try to not make mistakes. When you clear your mind of everything you know about poker and play the snowie way you win- it’s as simple as that. Regs and good players will get a lot out of snowie but the novices are going to absolutely love it. When I started playing before the internet was even invented I invested a lot of money and learned from my mistakes- nobody was sat guiding me telling me where I was going wrong. New players are now going to have access to this which will turn them into winning players very quickly.</p> <p>I am still trying to work on a counter system to exploit snowie but where I was once 100% confident I could exploit it I am now more doing it as a challenge. I once spent many months trying to solve the Beal Conjecture and I am now beginning to think I may have a better chance of doing that and may go back to it.</p> <p>Ariston<br /> The only player to have beaten snowie ;) (kinda)&nbsp;</p> <p>&nbsp;</p> <p>About Ariston<br /> <em><img alt="" data-cke-saved-src="/blog/sites/default/files/Ariston.jpg" src="/blog/sites/default/files/Ariston.jpg" style="width: 100px; height: 150px; border-width: 4px; border-style: solid; float: left; margin-left: 10px; margin-right: 10px;" /><span style="font-size:14px;">I was ranked in pocket 5s top 5 online pros when they first launched, due mainly to number of satellite wins. won tournament leaderboard on stars when they first launched it and finished second the following week. I have played at tournaments all over the world with a betfair patch on- many won in via&nbsp;sats and some sponsored. Wrote the first formula on how to win seats in sats and won 6 world series packages in one year. won/chopped irish masters 3 years in a row 2005-2008. Moved into staking and theory and helped set up the first short stacking systems and rake farms. Worked personally with players in world's top 10, helping them identify leaks in their games. Best known on the circuit for the "ariston" moniker given to me by someone in a live game after getting a headache ( allowing me to get in his head to make him lose all his money) - he said " you are like that commercial, you go on and on....ariston and on and on"... it stuck.</span></em></p>
Jim welch (not verified)
Wed, 01/15/2014 - 08:57
Wed, 01/15/2014 - 11:14