2+2ers hitting up a play money site?

[PokerParadimeMc]

[ QUOTE ]
The real fish in that hand is clearly Sonia. I actually emailed the site owner and conversed a bit about this, but Sonia does not appear to play good poker to me, but yet the stats they compile on her have her destroying the games. I can't really justify that in my mind when I see the way she plays.

-DeathDonkey

[/ QUOTE ]

Hi DeathDonkey, Sonia varies her play significantly depending on her opponents. Some strange plays may seem otherwise if you had seen their stats, however for obvious reasons they have to remain confidential.

We are in the process of adding Monies won from or lost to Sonia in your stats so you can see just how fishy she is!

[ QUOTE ]
this is hilarious. someone created a site where pbob runs hot and crushes all.

but wtf, how come theres no virtual victor?

[/ QUOTE ]

There is a virtual PokerBob because someone has signed up to the site with that username; there is no virtual Victor as no-one has registered with that name.

Marc

[inferno]

does sonia know your cards?

[PokerParadimeMc]

[ QUOTE ]
does sonia know your cards?

[/ QUOTE ]

No

Marc

[Spy Dog]

I call BS on Sonia.

I had a hand where she openraised from the SB, then I 3-bet and she folded preflop. WTF?? Maybe she read my 3-bet as super-super strong. No. I don't think so. I had 3-bet T7o in the same situation a few hands earlier and showed it down.

[gaming_mouse]

PokerParadimeMc,

The idea for this site very cool, but almost too good to be true. I understand that a detailed answer to the following question would probably give away more than you want to, but I'll ask anyway and maybe you will be able to give a general answer.

How is it possible mathematically to infer a "strategy" based on a player's hand histories that would be anything more than the crudest approximation, even after 10K hands or more?

How do these virtual clones create that strategy?

Do you take into account, eg, that a good human player is doing fairly complex oppoenent modelling himself? And do you try to replicate his opponent modelling?

Or are the models based simply on a complex collection of stats: VPIP, PFR, 3bet, cbet%, turn cr %, etc, etc?

Thanks for any info
gm

[gaming_mouse]

Okay,

I found the answers to my own questions here:

http://pokerparadime.com/StaticPage?template=aboutus.vm

and HOLY SH*T! The results you guys list at the end of the page are incredible:

[ QUOTE ]

Due to the different nature of the heads up game Sonia was started off with no knowledge of poker apart from a short self-play heads up match to enable to her to get a very basic grasp of the game. Despite this handicap Sonia's class showed against Vexbot as she went on to win the one million-hand session at an outstanding rate of 3.5 big bets per 100 hands. It was decided to let Sonia keep her experience of playing Vexbot when starting the test against Sparbot. Again Sonia outclassed Sparbot to win by 3.1 big bets per 100 hands over a full million hands.

To complete the tests Sonia was played against two tables of Pokis, one full ring and one short handed. One seat is always reserved for a human player in Poker Academy so the game had to play with one short of full ring. Despite not having any experience of full ring, Sonia rapidly adapted to the game to achieve a rock steady win rate, despite only feeling the urge to enter the pot with 15% of her hands. Having demonstrated her superiority long before the million hands, the session was stopped at 250,000 hands where her win rate stood at 5.6 big bets per 100 hands. For the final test the best Pokis were kept from the previous session. Able to play many more hands in 6-Max, Sonia made her class really work for her by achieving a whopping 9.5 big bets per 100 hands over 100,000 hands.


[/ QUOTE ]

[Spy Dog]

I just want to state that I like the site and what you are trying to accomplish.

One thing that should be noted is the winrates that Sonia is claiming. Keep in mind that these are rake-free winrates. A 1 BB/100 winner at stakes like 5/10 is actually a 4BB/100 if rake was ignored.

[gaming_mouse]

[ QUOTE ]
I just want to state that I like the site and what you are trying to accomplish.

One thing that should be noted is the winrates that Sonia is claiming. Keep in mind that these are rake-free winrates. A 1 BB/100 winner at stakes like 5/10 is actually a 4BB/100 if rake was ignored.

[/ QUOTE ]

Ah good point. I completely forgot about that.

[gaming_mouse]

I am also interested in how much lhe poker-specific information is needed to "train" Sofia?

Reading the description on the website, it sounds as if almost none is required. That is, it sounds as if Sofia is a general engine that needs nothing more than the game rules in order to work its magic, but that seems pretty difficult to believe.

For example, did you have to define situational categories like board textures, betting sequences, etc, that Sofia knew it needed to collect data on and base modelling on? Or was Sofia somehow able to create those categories herself (that seems almost impossible to me)?

J

[Tryptamean]

[ QUOTE ]
I just want to state that I like the site and what you are trying to accomplish.


[/ QUOTE ]
sorry if I'm being dense, but what is the site trying to accomplish?

anyway, I signed up last night and played about 20 hands, then I got bored and started playing like a typical play-money maniac for another 20 hands or so and quit. Does that mean virtual tryptamean is still playing like a maniac? or a tag? or something in between?

Also, why does sonia not get the <> around her name if she is also a virtual player?