CalcioPoker

A successor to Scoby the Poker Bot -- Calico is under development.
Calico is under going testing now.
During the testing period people can play Calico and see information that is usually hidden. This includes his cards and the data he uses to make a decision.

Calico is written PHP 5, a scripted language.
This means there is no download.
Your browser or iPhone can run it.
Test Calico

Comments are welcome, also you can reach us through the contact link on Calico.

2008 Poker Bot Competition Summary

The 2008 poker competition consisted of 12 competitors from 6 countries and 19 programs. The results were announced at AAAI 2008 on July 15, 2008 in Chicago, IL.

Heads-Up Texas Hold'em Limit

Competitors

Program Name	Affiliation	Team Members
Dr. Sahbak	(No affiliation)	Roy Fox
GUS	Georgia State University	Nicholas Larsen
GS4-Beta	Carnegie Mellon University	Andrew Gilpin, Tuomas Sandholm, Troels Bjerre Sørensen
Hyperborean08-Online and Hyperborean08-Equilibrium	University of Alberta	Michael Bowling, Duane Szafron, Nolan Bard, Neil Burch, John Hawkin, Michael Johanson, Morgan Kan, Nick Abou Risk, Dave Schnizlein, Kevin Waugh, Martin Zinkevich, Bryce Paradis
Ian Fellows	University of California, San Diego	Ian Fellows
GGValuta	University of Bucharest	Mihai Ciucu, Stefan Popescu, Mihaita Alexandru Leoveanu, Diana Dulgheru
PokeMinnLimit1 and PokeMinnLimit2	University of Minnesota	Mohamed Elidrisi, Brett Borghetti

Results

The Heads-Up Limit competition features two winner determination rules. The bankroll rule orders the competitors according to the total number of chips won. The wins rule orders the competitors according to their wins and losses against the other competitors, and uses the instant runoff technique (see the rules page for further information).

60,000 hands were played for each pair of competitors, for a total of 2,160,000 hands. The following table summarizes the pairwise and overall performance for each entry.

	GS4-Beta	Hyperborean08-Online	Hyperborean08-Equilibrium	Ian Fellows	GGValuta	PokeMinnLimit2	PokeMinnLimit1	GUS	Dr. Sahbak	Average
GS4-Beta		-0.03 ± 0.012	-0.029 ± 0.009	-0.02 ± 0.015	-0.036 ± 0.009	0.2 ± 0.018	0.113 ± 0.016	0.695 ± 0.018	0.713 ± 0.007	0.201 ± 0.007
Hyperborean08-Online	0.03 ± 0.012		-0.002 ± 0.014	0.004 ± 0.005	-0.007 ± 0.012	0.202 ± 0.017	0.184 ± 0.01	0.52 ± 0.009	0.616 ± 0.01	0.194 ± 0.006
Hyperborean08-Equilibrium	0.029 ± 0.009	0.002 ± 0.014		0.024 ± 0.003	0.005 ± 0.004	0.182 ± 0.008	0.161 ± 0.02	0.476 ± 0.019	0.566 ± 0.013	0.181 ± 0.004
Ian Fellows	0.02 ± 0.015	-0.004 ± 0.005	-0.024 ± 0.003		0.003 ± 0.004	0.153 ± 0.017	0.154 ± 0.02	0.467 ± 0.012	0.532 ± 0.019	0.163 ± 0.005
GGValuta	0.036 ± 0.009	0.007 ± 0.012	-0.005 ± 0.004	-0.003 ± 0.004		0.144 ± 0.013	0.143 ± 0.008	0.452 ± 0.017	0.494 ± 0.094	0.159 ± 0.011
PokeMinnLimit2	-0.2 ± 0.018	-0.202 ± 0.017	-0.182 ± 0.008	-0.153 ± 0.017	-0.144 ± 0.013		0.03 ± 0.021	0.743 ± 0.02	1.265 ± 0.017	0.145 ± 0.008
PokeMinnLimit1	-0.113 ± 0.016	-0.184 ± 0.01	-0.161 ± 0.02	-0.154 ± 0.02	-0.143 ± 0.008	-0.03 ± 0.021		0.299 ± 0.014	0.921 ± 0.02	0.054 ± 0.011
GUS	-0.695 ± 0.018	-0.52 ± 0.009	-0.476 ± 0.019	-0.467 ± 0.012	-0.452 ± 0.017	-0.743 ± 0.02	-0.299 ± 0.014		0.512 ± 0.014	-0.393 ± 0.007
Dr. Sahbak	-0.713 ± 0.007	-0.616 ± 0.01	-0.566 ± 0.013	-0.532 ± 0.019	-0.494 ± 0.094	-1.265 ± 0.017	-0.921 ± 0.02	-0.512 ± 0.014		-0.702 ± 0.01

Heads-Up Limit Bankroll

The number in parentheses is the average number of small bets won per hand.

1. GS4-Beta (0.201)
2. Hyperborean08-Online (0.194)
3. Hyperborean08-Equilibrium (0.181)
4. Ian Fellows (0.163)
5. GGValuta (0.159)
6. PokeMinnLimit2 (0.145)
7. PokeMinnLimit1 (0.054)
8. GUS (-0.393)
9. Dr. Sahbak (-0.702)

Heads-Up Limit Wins

The numbers in parentheses are the overall number of wins and losses. Note that there is a three-way tie for second.

1. Hyperborean08-Equilibrium (8-0)
2. Hyperborean08-Online (6-2)
2. Ian Fellows (6-2)
2. GGValuta (6-2)
5. GS4-Beta (4-4)
6. PokeMinnLimit2 (3-5)
7. PokeMinnLimit1 (2-6)
8. GUS (1-7)
9. Dr. Sahbak (0-8)

Heads-Up Texas Hold'em No Limit

100,000 hands were played for each pair of competitors for a total of 600,000 hands.

Competitors

Program Name	Affiliation	Team Members
University of Ballarat	University of Ballarat	Peter Vamplew, Chris Turville, Robert Layton
BluffBot 3.0	(No affiliation)	Teppo Salonen
Tartanian2-Beta	Carnegie Mellon University	Andrew Gilpin, Tuomas Sandholm, Troels Bjerre Sørensen
Hyperborean08	University of Alberta	Michael Bowling, Duane Szafron, Nolan Bard, Neil Burch, John Hawkin, Michael Johanson, Morgan Kan, Nick Abou Risk, Dave Schnizlein, Kevin Waugh, Martin Zinkevich, Bryce Paradis

Results

The following tables summarize the pairwise and overall performance of the bots. Since the winner determination rule was instant runoff bankroll, the programs did not finish in the same order as they would have if they were ordered by total bankroll. The tables illustrate the winner determination process.

	Tartanian2-Beta	BluffBot 3.0	Hyperborean08	Ballarat	Average
Tartanian2-Beta		-0.611 ± 0.091	-0.625 ± 0.091	5.371 ± 0.223	1.378 ± 0.091
BluffBot 3.0	0.611 ± 0.091		-0.109 ± 0.046	2.566 ± 0.109	1.023 ± 0.052
Hyperborean08	0.625 ± 0.091	0.109 ± 0.046		2.13 ± 0.123	0.954 ± 0.054
Ballarat	-5.371 ± 0.223	-2.566 ± 0.109	-2.13 ± 0.123		-3.355 ± 0.107

	Hyperborean08	BluffBot 3.0	Tartanian2-Beta	Average
Hyperborean08		0.109 ± 0.046	0.625 ± 0.091	0.367 ± 0.054
BluffBot 3.0	-0.109 ± 0.046		0.611 ± 0.091	0.251 ± 0.051
Tartanian2-Beta	-0.625 ± 0.091	-0.611 ± 0.091		-0.618 ± 0.061

	Hyperborean08	BluffBot 3.0	Average
Hyperborean08		0.109 ± 0.046	0.109 ± 0.046
BluffBot 3.0	-0.109 ± 0.046		-0.109 ± 0.046

Final Rankings

The number in parentheses is the average number of small bets won per hand.

1. Hyperborean08-NoLimit (0.954)
2. BluffBot 3.0 (1.023)
3. Tartanian2-Beta (1.378)
4. Ballarat (-3.355)

6-Player Texas Hold'em Limit

Competitors

Program Name	Affiliation	Team Members
AKI-RealBot and mcBotUltra	Technical University Darmstadt	Johannes F�rnkranz, Ulf Lorenz, Frederik Janssen, Sang-Hyeun Park, Eneldo Loza Menc�a, Jan-Nikolas Sulzmann, Oliver Uwira, Hendrik Schaffer, Thomas G�rge, Lars Meyer, Timo Bozsolik, Marian Wieczorek, Michael W�chter, Bj�rn Heidenreich, Andreas Eismann, Michael Herrmann, Arno Mittelbach, Bastian Christoph, Benjamin Herbert, Claudio Weck, Stefan L�ck, Immanuel Schweizer, Alexander Marinc, Kamill Panitzek
GUS	Georgia State University	Nicholas Larsen
CMURing-Beta	Carnegie Mellon University	Sam Ganzfried, Andrew Gilpin, Tuomas Sandholm
Poki0	University of Alberta	Aaron Davidson, Darse Billings, Jonathan Schaeffer, Duane Szafron, Lourdes Pena, Nick Abou Risk, John Hawkin
DCU	Dublin City University	David Sinclair, Neill Sweeney

Results

The number in parentheses is the average number of small blinds won per hand.

1. Poki0 (0.646)
2. AKI-RealBot (0.573)
3. DCU (0.252)
4. CMURing-Prototype (0.163)
5. mcBotUltra (-0.135)
6. GUS (-1.50)

BluffBot

BluffBot

BluffBot is a World Champion "poker bot" created by Teppo Salonen for the purpose of competing in the first two annual AAAI Computer Poker Competitions in 2006 and 2007.

In the 2006 competition (held at the 21st National Conference on Artificial Intelligence) the first version of BluffBot finished in the 2nd place in limit Hold 'em competition beating such opponents as GS2 by Carnegie Mellon University and Monash BPP by Monash University, losing only to University of Alberta's Hyperborean.

In 2007, the new BluffBot 2.0 went undefeated and took a convincing 1st place victory in the no-limit Hold 'em competition beating all other opponents, including Hyperborean07 by University of Alberta, GS3 by Carnegie Mellon University, as well as bots from Gomel State University, Milano Polytechnic, University of Manitoba and University of Minnesota to name a few.

Does It Really Bluff?

Absolutely! Any effective poker strategy requires a balanced use of bluffing and value betting. Contrary to common beliefs, computer programs utilizing game theory are in fact often better bluffers than most humans.

The AI (artificial intelligence) used in BluffBot 2.0 is a non-adaptive expert system designed using the principles of game theory. This approach makes a good practical use of variety of known expert strategies exploiting weaker opponents while at the same time effectively defending itself against exploitation from adaptive opponents.

Bot-running

Millions of different computations.
It's very simple, it's legal and no one on the other side of the screen will ever know. I've run a cable from the PC showing the game - or, rather, games - into a laptop running some specialist poker software. This displays an information-only Etch A Sketch-like rendering of the poker tables; the cards, the betting, the players contending the pot. The laptop is making millions of different computations based upon the strength of my hand and how the others are betting. Then it places 'my' bet. I don't have to lift a finger or even be in the room.

Online poker is nothing more than a busted flush.
If you're a poker player, this is merely unethical. But if you're an executive or shareholder in one of the top poker websites, the advent of programs that play for you is very bad news indeed. Online poker is a £3bn-a-year industry - £3m is gambled on online poker every day in Britain alone (we're now the fifth biggest gambling country in the world). But this depends on the punters knowing they're getting a fair game. When they're up against expertly programmed computer players, then they are, quite emphatically, not. And if these programs evolve as fast as the experts predict, online poker is nothing more than a busted flush.

The game is completely corrupt.
One expert in this powerful new software, 'Chopper', tells me, 'It's amazing to think of how much we gamble on online poker sites - mainly because there is no such thing as a fair game of online poker. It just doesn't exist. The game is completely corrupt; it has zero integrity. Online players are secretly using every means at their disposal to fleece you --and at the forefront of their campaign is the use of poker robots. When all this becomes public knowledge, the amateurs will leave and the game will die.'

Phil Robinson You'll never beat poker robots

How a program ranks a hand.

Assignment of an 11 digit number
A program can reduce any possible hand to an 11 digit number. This includes an array of 2 or 5 or 6 or 7 cards. The number represents the value of the cards as a poker hand and is used by the bot to make decisions.

The first digit
The first digit denotes the rank of the hand, to wit:

1. nopair
2. pair
3. 2 pair
4. trips
5. straight
6. flush
7. full house
8. quads
9. straight flush


The trailing digits
The trailing 10 digits consist of a concatenation of up to five two digit numbers which denote the strength of the hand within the rank where
24=ace
23=king
22=queen
21=jack
20=10
19=9
18=8
17=7
16=6
15=5
14=4
13=3
12=2