Recommend
19 
 Thumb up
 Hide
57 Posts
1 , 2 , 3  Next »   | 

Modern Art» Forums » General

Subject: Modern Art computer game w/ AI available for download rss

Your Tags: Add tags
Popular Tags: [View All]
Ylaine Gerardin
United States
Boston
Massachusetts
flag msg tools
Avatar
mb
http://www.dillgroup.ucsf.edu/~grocklin/modart.zip

4 Players, you against 3 AI opponents. Quite good AI. Let me know if you have any comments or suggestions!

GRocklin07@mckenna.edu

Readme underneath...

-------

MODERN ART*README*1-1-2005*************************************************************

Modern Art


High Stakes Bidding in the Galleries
Designed by Reiner Knizia
Programmed by Gabriel Rocklin

Version 1.0 - Release Version

This file contains:

-System Requirements
-Required Files
-How to play Modern Art
-How to use Modern Art
-Modern Art Strategy
-Programming Notes
-Credits
-Links
-Disclaimer

SYSTEM REQUIREMENTS:*******************

Modern Art should be able to run on any computer capable of running
Windows XP at 1024x768. In fact, it should run on earlier versions
of Windows, but I haven't tested it. Don't worry if it's slow loading;
it takes time to load the AI.

REQUIRED FILES:************************

ModArt.exe - Modern Art main executable.

msvbvm60.dll - Main Visual Basic 6.0 Runtime dll. If you somehow
lose it, do an FTP search for it. Included in
program.

HOW TO PLAY MODERN ART:****************

Modern art is a game of strategy that is based around putting paintings
from your hand up for auction, buying paintings from other players in
auctions, and selling paintings bought in auctions back to the bank
for a profit.

The paintings are from five different artists: Lite Metal, Yoko,
Cristin P, Karl Gitter, and Krypto. Each player begins with a hand of
9 paintings from the various available artists. On your turn, you
select a painting from your hand to put up for auction. The other
players bid on this painting and whoever bids the highest ultimately
wins the painting. They pay the auctioneer the price and recieve the
painting. The auctioneer may bid on his or her own painting, and if
the auctioneer wins, the price of the painting is paid to the bank.
Paintings that have been bought are displayed publicly in front of
the owner, and are a separate group of paintings from the ones in
each players hand. The paintings in a players hand are kept throughout
the game (unless auctioned), the paintings that have been bought are
sold back to the bank at the end of a round.

The game is played in four rounds, and a round ends as soon as the
fifth painting of a particular artist is put up for auction. This
painting is not auctioned off and is discarded. At the end of the
round, values are assigned to each of the artists based on how many
paintings of that artist were put up for auction. The artist with
the most paintings auctioned off (the artist who ended the round)
is worth the most, the artist with the second most paintings auctioned
is worth the second most, and the artist with the third most paintings
is worth the third most. The remaining two artists are worth nothing.
In the event of a tie, whichever artist is farthest to the left on
the board wins: Lite Metal wins all ties, Yoko wins all ties except
ties to Lite Metal, and Krypto loses all ties.

The values for first, second, and third are as follows:

First Place: 30
Second Place: 20
Third Place: 10

Each player begins the game with 100 cash to use on purchasing
paintings.

When the round ends, all players sell the paintings they bought
during the round back to the bank at the appropriate value. The next
round begins with the player after the player who ended the round.
The player who ended the round never got to hold an actual auction. At
the end of rounds one and two, each player receives 4 new paintings to
their hand. No new paintings are given after round three.

The values of the artists accumulate from round to round. If an artist
places first round one and second round two, their paintings sell back
at the end of round two for 50 (30+20). AN ARTIST MUST PLACE IN A
GIVEN ROUND TO BE WORTH ANYTHING THAT ROUND. Even if Krypto was worth
20 in round one, if Krypto does not place in round two, all Kryptos sold
are worthless and the players owning them will recieve nothing from the
bank.

The auctions are conducted in five different ways, and each painting
indicates how that paintings auction should be conducted. The five
types of auctions and their symbols are:

-Hidden ( * ): Each player secretly selects an amount of money they
are willing to pay for the painting, and then all amounts are revealed.
The highest offer wins the painting. In the event two players tie,
the player closest to the auctioneer in turn order wins the painting.
The auctioneer wins all ties.

-Fixed Price ( $ ): The auctioneer chooses a price not exceeding his
or her current amount of money and offers the painting for sale. The
player to his or her left has the first chance to buy the painting.
If he or she refuses, the painting continues around until either
bought or all players refuse to buy the painting. If no one buys
the painting, the auctioneer is forced to buy the painting at the
price set.

-Once Around ( G ): The player to the auctioneers left announces the
price he or she is willing to pay for the painting. The next player
in turn order either passes or raises this amount. The third
player does the same. Finally, the auctioneer has the chance to
purchase his or her own painting.

-Open ( + ): Any player can bid at any time. When once player makes
an offer that no other player is willing to raise, that player
buys the painting.

-Double ( = ): When a dual auction is played, the auctioneer has the
option of playing a second card OF THE SAME ARTIST and auctioning off
both together as a unit. The auction is conducted in the manner
indicated by the second painting. A painting with a double auction
cannot be played as the second card to another double auction. If
the auctioneer does not play a second painting, the next player in turn
order has the option of playing a second painting, in which case he or
she conducts the entire auction and receives the entirety of the payment.
The play then continues clockwise from this player, as if the double
auction was that players turn. This means that if someone refuses to
play a card on top of a double when they have the chance, they
effectively forfeit their turn. If no one plays a painting on top of
the double and the play returns to the person who played the double,
then that player recieves that painting for free, and play continues
to his or her left.

The game ends at the end of round four, and whoever has the most
money is the winner.

HOW TO USE MODERN ART:*****************

The four players are represented on the screen in a circle, with you
at the bottom. Your cash is in parentheses next to your name. The
values of the artists throughout the game are indicated by the white
boxes in the center underneath the apporpriate artist. The number of
cards each computer player has in their hand is indicated by a white
box. Your hand is represented by icons at the bottom of the screen.
Each icon has a color representing its artist and a symbol representing
its type of auction. Hover the mouse over a painting to see which
artist and type of auction it is.

When you purchase paintings, the number of paintings purchased of each
artist will appear below your name as a colored box with a number
on it representing the number of paintings. Each of the computer
players purchases will appear in the same way underneath their names.

The player whose turn it is has their name indicated in red. When it
is your turn, double click on a painting in your hand in order to put
that painting up for auction. Information about the painting will
appear in the small light green box in the lower right. To bid, enter
a value into the text box and click bid. If you do not wish to bid,
click pass. When you confirm your bid, the game will tell you who
ultimately purchased the painting in the status bar, and you can
acknowledge this by pressing the confirm button again. In the event of
an open auction, you will be given the option to continue bidding
higher and higher until either you or the computer give up.

MODERN ART STRATEGY:*******************

There are two ways to make money in Modern Art: buying paintings from
other players and selling them back to the bank for a profit, and
selling paintings to other players. Selling paintings to other players
makes more total money; however, in a game where all players are selling
paintings equally effectively, buying paintings to sell to the bank will
make the difference. Both avenues of profit must be used successfully
in order to win.

There are two keys to making profit selling paintings to other players:
Setting up artists to be valuable in late rounds, and using double
auctions effectively. If you have a large amount of a particular artist
in your hand at the beginning, playing a few in the early rounds to
put that artist on the board will make your other paintings more valuble
to sell in later rounds. Remember: buying a painting for 20 early
and selling it for 30 in round one is no different to the purchaser
than buying a painting for 60 and selling it for 70. The profit is the
same. HOWEVER, for the person AUCTIONING that painting, it makes a huge
difference whether it sells for 20 or 60. Thus position the artists
you have early in the game so that they can be extraordinarily profitable
late in the game.

Double auctions supercharge this. Not only do they make it more likely
that the artist will place (because two paintings are already out), but
the auctioneer makes twice the income. In late rounds, doubles of an
artist that has placed previously can sell for over 100. A huge sale
of doubles in round four can often decide the game. This has two
consequences:

(1) The main purpose of the early rounds should be setting up the artists
you have doubles in to be valuable late in the game
(2) You should virtually NEVER play a double before round three, because
you don't know what cards you're going to get later and how the values
will shape up.

The strategy behind buying cards is more subtle, particularly because you
are increasing the sellers profit by bidding higher and higher. Buying
paintings early in the round will be more lucrative because the paintings
generally sell for less, but the reason they sell for less is because all
purchases are inherently riskier when the round is further from completion.
To bid effectively, a player needs to develop a sense of how likely a
painting is to place based on how many have been sold so far, which other
artists are out, and how many he or she has in his or her hand that can be
played to ensure that artist places.

Modern Art games are rarely won by defensive players who avoid the risks
of purchasing and attempt to win only by selling. Aggressive purchasers
are more successful, but a single expensive purchase of a painting that
ends up not placing can kill a player. Understanding which risks are
likely to pay off and which are not is the key to effective Modern Art
bidding.

There are many miscellaneous tactics that one picks up after playing,
including:

-Letting other players purchase paintings of artists you have already
purchased in order to broaden the number of people who have an interest
in ensuring that artist places
-Effectively using the lone double auction in order to either force the
round closed or pick up a free painting
-Bidding up open auctions on paintings you are selling
-Predicting other players likely plays based on the paintings they have
already bought and the bids they are willing to make

Modern Art is a deep game that requires both strategy - a one-shot game
plan that you attempt at the beginning and hope succeeds - and tactics
- turn by turn decisions that you improve throughout the game. The best
way to improve is through practice.

PROGRAMMING NOTES:*********************

Modern Art was programmed in Visual Basic 6.0. The bidding AI was
designed as a neural network and trained using backpropagation. The card
playing AI is based on a fuzzy logic system with various coefficients
determining how useful it is for the computer to achieve certain goals
such as selling a painting for a lot of money or playing a painting
that is likely to increase the value of paintings already purchased by
that player. These coefficients were evolved using a genetic algorithm.

A neural network consists of a set of neurons connected to each other via
a set of weights. The neural network used for Modern Art, called a multi-
level perceptron with one hidden layer, consists of three sets of neurons:
95 input neurons, 40 hidden neurons, and 1 output neuron. All of the
input neurons are connected to each of the hidden neurons, and each of the
hidden neurons is connected to the output neuron.

The input neurons are assigned values based on various information
available to a player attempting to bid on a painting: which artist
the painting is, how many of each artist have already been sold, how each
artist has placed in previous rounds, how many of each artist each player
is has already bought, the contents of that player's own hand, etc. The
hidden neurons are mathematical functions of the input neurons. Each
hidden neuron takes the value of each input neuron multiplied by the
weight for that input neuron and sums them all up. For example, imagine
the following values:

Input Neuron 1: -1
Input Neuron 2: 0.5
Input Neuron 3: 1
Input Neuron 4: 2

Weight from Input Neuron 1 to Hidden Neuron 1: 2
Weight from Input Neuron 2 to Hidden Neuron 1: -1
Weight from Input Neuron 3 to Hidden Neuron 1: 0
Weight from Input Neuron 4 to Hidden Neuron 1: -0.333

Hidden Neuron 1 multiplies all of the inputs by their weights and adds these
together, making (-1)(2) + (0.5)(-1) + (1)(0) + (2)(-0.333) = -3.166 .
This value is then "scaled" by using it as the parameter in the sigmoid
function f(x)=1/(1+(e^-x)), where x is the value we just calculated, -3.666 .
By scaling in this manner, if the sum of the inputs times their weights produces
a very negative value, the hidden neuron will assume a value near 0, and a very
positive sum of inputs and weights will produce a value near 1. Though the
example only showed this process for four input neurons and one hidden
neuron, in Modern Art all 95 input neurons are connected to all 40 hidden
neurons. The hidden neurons then have weights that connect them to the output
neuron, which again uses the sigmoid function and again produces a value
between 0 and 1. A value of 0 causes the AI player to bid 0; a value of 1
causes the AI player to bid the absolute maximum the painting could possibly
be worth. If you are unfamiliar with the sigmoid function, a web search
can show you a graph of it.

Because the sigmoid function is non-linear, the value of the output neuron
is NOT a linear function of the input neurons. This means that each input
neuron can cause a higher bid in some situations (for example, by bumping
one particular hidden neuron's sigmoid function over the steep part of the
curve), but cause a lower bid in another situation (because the hidden
neuron affected in the previously described situation is in the flat
part of the curve and thus the input neuron's "positive" contribution
to the final value becomes irrelevant). This is essential in a game
like Modern Art.

The true value of neural nets is not their ability to determine an
appropriate bid when given input data; it is their ability to learn what an
appropriate bid is by being trained. A neural network produces an output
based on the values of its weights, and by adjusting those weights it can
"learn" to produce an ideal output. A training program played Modern Art
against itself over and over and over again (thousands of times), and each
time adjusted the weights to try to make the output neuron closest to an
ideal value. For each auction that made a profit for the buyer, the ideal
value was set to be one greater than the the highest bid; for each auction
that made a loss, the ideal value was set to be one greater than the
highest bid that could have potentially made a profit, and 0 in the case
of auctions of artists that never placed in that particular round.

Training the neural network was a difficult process and much trial and
error was used in order to determine its appropriate construction. Many
of the input neurons are just combinations of other input neurons and contain
no real new information but were found to improve the quality of the final
results.

The card-playing system is based on a fuzzy logic system using coefficients
that evolved via a genetic algorithm. Each painting in a player's hand is
given five ratings based on how well it achieves five particular goals:

(1) Making the art owned by that player more valuable
(2) Making the art owned by that player closer to becoming more valuable
(3) Selling for a lot of money
(4) Increasing the value of the art inside the players hand to sell later
(5) Increasing the value of dual auctions inside the players hand for later

Imagine that in a given round, the following cards have been played:

3 Lite Metals
2 Yokos
2 Cristin Ps

Axel owns both Yokos and 1 Cristin P. If Axel plays a Yoko, he accomplishes
goal (2) by pushing Yoko up into a tie with Lite Metal and making it possible
to surpass it. If Axel plays a Cristin P, he actually hurts himself with
regard to goal (1) because even though his sole Cristin P becomes more
valuable, his two Yokos each decline in value by the same amount, reducing
the total value of the paintings he has already purchased. If Axel has lots
of Cristin P in his hand, however, or if he expects Cristin P to sell for
more than Yoko, then it is possible playing a Cristin P would be so helpful
toward goals (3), (4), and (5) that it would be even more useful than playing
a Yoko.

The ratings a painting recieves with respect to goals 1-5 are then each
multiplied by a coefficient specific to that goal and that round.
Obviously, playing paintings that set up doubles and improve the value of art
to sell later is much more important in rounds one and two than in rounds
three and four. Also, the coefficients change throughout the round: goals
(4) and (5) are usually more important at the beginning of a round than
goals (1) and (2), which are more important at the end. When there are
at max 3 cards of a particular artist out, goal (2) may be worth attempting,
but when four are out, it can be more useful to cut losses.

The coefficients for each goal for each round were set with little precision
by me, and then hundreds of variants of these were created, slightly different
from the originals. These variants played against each other thousands of
times, with the best variants being selected to try again and the least
successful being eliminated. New variants were produced based on the
previously successful variants and so forth. Over time, a good set of
coefficients evolved for each round. The "decay" - the change in the
coefficients throughout a round - was also evolved in a similar fashion.

For more information about the specific details and challenges of programming
Modern Art, including:

-Why the input neurons were chosen the way they were chosen
-The method of backpropagation used to train the neural network
-Details of neural network training including biases and the use of momentum
-Why those five goals were chosen for the card playing system
-Final optimization of the neural net and the card playing coefficients

feel free to e-mail me at GRocklin07@mckenna.edu

CREDITS:*******************************

Gabriel Rocklin (Snapper):
Design, Programming, Graphics

Beta Testers:

Corey Reese

Thanks to:

Reiner Knizia, Modern Art Designer
Mat Buckland, "AI Techniques for Game Programmers"


This program has been created on a budget of 60$ (40 for the Buckland
book and 20 for the Mayfair edition of Knizia's game), several hundred
deceased artificial neurons, an ungodly amount of time that almost
certainly should have been spent on other things, several design text
files, three (precisely) brilliant ideas, and over 200 hours of leaving
the AI evolving program running both in various tests to improve it
and to evolve the final AI.

The AI evolving program, incidentally, is the most incomprehensible
code I have ever written or seen. It is likely that Visual Basic
simply became so frustrated at me that it spat out a reasonably good
Modern Art player on its own in order to get me to leave it alone.

LINKS:*********************************

http://www.boardgamegeek.com
-The best site for boardgaming on the web, including a link to this
program.

http://www.boardgamegeek.com/game/118
-BoardgameGeek's page on Modern Art

http://www.mayfairgames.com
-The US publisher of Modern Art

http://www.mayfairgames.com/mfg-shop/3300-4499/qps/4402.html
-Mayfair's page on Modern Art

http://www.fairplaygames.com
-The cheapest place I've found to buy board games on the web

http://en.wikipedia.org/wiki/Reiner_Knizia
-Wikipedia's page on Reiner Knizia, the designer of Modern Art

http://en.wikipedia.org/wiki/Neural_networks
-Wikipedia's page on Artificial Neural Networks

http://www.cwal.net/forums/rr
-The CWAL Rubber Room, where you can usually find me hanging out
as "Snapper". A community of StarCraft players who now mostly
discuss other things.

GRocklin07@mckenna.edu

DISCLAIMER:****************************

I am not responsible for whatever happens to you due to usage of
this program. If you ceased attending classes to play Modern Art
and dropped out of school, that's hilarious, but it's not my problem.
If you reverse-engineered this program to turn in as your senior
thesis in CS AI development and your readers eventually found this
program on the internet, that's also hilarious, but you really should
have given it a try on your own. And try to challenge yourself - we'd
all like to see Tigris and Euphrates or El Grande done well with
computer AI, right?

Any similarities to real-life entities in this document or this
program are purely coincidental.

1,024,000 bytes were harmed in the creation of this program.

Fortunately, none of them were animals.

Except Barry, Harry and Larry's Brother.

*sniff*

But don't tell anyone.

"There is often a large gap between theory and practice.
Furthermore, the gap between theory and practice in practice
is much larger than the gap between theory and practice in
theory."

"Kryptos? We don't need no steenkin' Kryptos!"

"I prefer to kill one bird with two stones: one to kill it,
and the other to make sure it's dead."

And in the immortal words of the great Snapper,
6 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Anson Li
Canada
Vancouver
British Columbia
flag msg tools
Avatar
mbmbmbmbmb
Re:Modern Art computer game w/ AI available for download
Snapper (#74358),

I have a question about the double auction card. In my rules, the profit is split between the person who plays the original (and decides to skip the 2nd card) and the person who plays the matching 2nd card.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Dan Blum
United States
Wilmington
Massachusetts
flag msg tools
badge
Re:Modern Art computer game w/ AI available for download
anson (#74637),

That's what the Mayfair rules say, but the original German rules have it the way it is implemented in the program - the person playing the matching card (if any) gets all the money.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Ylaine Gerardin
United States
Boston
Massachusetts
flag msg tools
Avatar
mb
Re:Modern Art computer game w/ AI available for download
anson (#74637),

You're right, I chose to go with the original German rules. I like it better and it would be difficult to have the AI adapt to different rule systems like that, though I may at some point.

I still love playing lone doubles as the 4th card of a particular artist, in order to force someone else to end the round (so my turn comes up again sooner) or to pick up a free (and frequently valuable!) card.

Let me know what you think of the AI!
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Mike Mayer
United States
Arcadia
California
flag msg tools
designer
Avatar
mbmbmbmbmb
Re:Modern Art computer game w/ AI available for download
In Mayfair, the once around bidding starts to the left of the seller, with the seller getting to go last. Is it different in the German version?

 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Ylaine Gerardin
United States
Boston
Massachusetts
flag msg tools
Avatar
mb
Re:Modern Art computer game w/ AI available for download
Mike Mayer (#75526),

I don't believe so; it's that the way it's implemented in the program?
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Ylaine Gerardin
United States
Boston
Massachusetts
flag msg tools
Avatar
mb
Re:Modern Art computer game w/ AI available for download
Erm, my last post made no sense.

I don't believe the German rules are different from the Mayfair rules. Isn't the way you described the once around auction the same way it is implemented in the program?

What makes you think there's a rule discrepancy?
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Mike Mayer
United States
Arcadia
California
flag msg tools
designer
Avatar
mbmbmbmbmb
Re:Modern Art computer game w/ AI available for download
Snapper (#75552),

Sorry, my mistake. I misunderstood what was happening when I was playing the progam.

Very nifty program, by the way. I've now played three times. Got killed the first time, but clawed my way to second place the other two times by selling a lot more than I was buying. I'll have to work on finding a happy medium.

-Mike M.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Ylaine Gerardin
United States
Boston
Massachusetts
flag msg tools
Avatar
mb
Re:Modern Art computer game w/ AI available for download
Mike Mayer (#75591),

Thanks! Always nice to know that someone enjoys it.

Good luck!
Gabriel
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Sean McCarthy
United States
Seattle
Washington
flag msg tools
designer
badge
Avatar
mbmbmbmbmb
Re:Modern Art computer game w/ AI available for download
Snapper (#75552),

Awesome! Can't wait to try it out. Even just reading the readme has been quite enjoyable—I'm interested in game AI but don't "know" anything about it yet other than the usefulness of genetic algorithms.

Thanks for creating this fun and complex program for free!
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Ron Sostaric
United States
League City
Texas
flag msg tools
mbmbmbmbmb
Re:Modern Art computer game w/ AI available for download
Snapper (#74358),

This is really impressive. Writing a game AI is a lot of work.

So far I am 0-2 against your AI...but it's going down!

Good work!
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Mike Mayer
United States
Arcadia
California
flag msg tools
designer
Avatar
mbmbmbmbmb
Re:Modern Art computer game w/ AI available for download
Snapper (#75594),


I found a bug, I think. Clemsons put up a double auction card, sent it round, and Beatrix completed it by adding another double auction card (!), which is illegal, and caused the program to stop.

Also, an observation, the computer players really throw the money around! The real-life people I've played with are more conservative, especially in the later rounds. If "Frank" bought a painting for 100 bucks from "Jim," hoping to make 110 at the end of the round, "John" would certainly point out that while Frank made a net of 5, he gifted Jim an unholy present of 100 bucks free and clear.

As a result, the last two games I played the computer, I didnt buy anything during the last two rounds, just selling, and came out well ahead of the computer players.

 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Mike Mayer
United States
Arcadia
California
flag msg tools
designer
Avatar
mbmbmbmbmb
Re:Modern Art computer game w/ AI available for download
Mike Mayer (#75740),

I meant that Frank would have made a net of 10, not 5.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Mark McEvoy
Canada
Mountain
Ontario
flag msg tools
badge
Avatar
mbmbmbmb
Re:Modern Art computer game w/ AI available for download
Mike Mayer (#75741),

Read: http://boardgamegeek.com/article/39573

It doesn't surprise me that the AI plays 'loose'. AlexFrog outlines the logic in that article, as to why it's perpetually better to pay N (provided N < EV) for a painting (not your own auction) than allow an opponent to buy it for N-1. It makes sense to me that, in a game with multiple computer players, the tendency of the computers should be to play very near the expected value for each painting. If the computer played tight, the human player could always win by playing slightly looser (buying everything for one dollar more than the CPU players are willing to pay).
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Ylaine Gerardin
United States
Boston
Massachusetts
flag msg tools
Avatar
mb
Re:Modern Art computer game w/ AI available for download
Mike Mayer (#75741),

I'll work on fixing that dual auction problem and will probably have it fixed by the end of the day.

As for the high-paying computers, there's a discussion of this in one of the threads here with regard to "tight" groups (that try to limit the amount the seller makes) versus "loose" groups (that bid against each other to prevent any potential buyer from making much of a profit thus hugely benefitting the seller).

There are two problems with a "tight" group:

(1) They require cooperation among the players
(2) If a single player plays slightly "looser" than all the other players, he will win every auction, make a large profit, and win the game.

It would be too difficult (and I believe would remove a lot of the fun from the game) to invent some system where the human player negotiates with the computer players in order to achieve cooperation, and especially if this system were made enforceable on the human player, which is the only way it could realistically work.

This leaves a few other solutions:

(1) All the computer players could play "tight" all the time and try to restrict how much profit the seller makes, in which case the human player simply outbids the computers by 1, buys virtually any painting they want for a healthy profit, and wins the game every time
(2) All the computer players could play "tight" and cooperate with themselves anytime the human is selling but be willing to play "loose" anytime the human is buying in order to prevent the human from winning so easily as described in (1). In this case, the human not only makes much profit from buying (the same as they do currently), but makes way less of a profit every time they sell compared to the other players. The human loses every time.

So, I think you'll agree that in the absence of any realistic cooperation scheme, the best idea is just for all the players to play "loose", make big money by selling to keep them in the game, and expect to make the marginal difference that wins the game through buying. Of course, sometimes you'll be able to win through selling alone, but I wouldn't count on it.

A computer program will very rarely replicate the full experience of negotiation, cooperation, coercion, threats, and bluffs that occur in any board game where the players aren't completely isolated from each other. Just think of it as an incentive play the real game every once in a while.

Of course, in the meantime, the computer game is a pretty good way to spend 10 minutes, should improve your real game, and does replicate the experience of playing with a group of people who utterly refuse to cooperate with each other. You know the type (:
3 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Mike Mayer
United States
Arcadia
California
flag msg tools
designer
Avatar
mbmbmbmbmb
Re:Modern Art computer game w/ AI available for download
Snapper (#75759),


Thanks for the explanation about the AI!

Just to be clear, I wasn't putting the program down. I still think its a nifty program, and you're right, I'll probably start losing again real soon. laugh

-Mike M
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Ylaine Gerardin
United States
Boston
Massachusetts
flag msg tools
Avatar
mb
Re:Modern Art computer game w/ AI available for download
Mike Mayer (#75770),

I know (: Thanks for the support!
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Ylaine Gerardin
United States
Boston
Massachusetts
flag msg tools
Avatar
mb
Re:Modern Art computer game w/ AI available for download
I believe I have fixed the playing a dual on top of another dual bug. The updated files are available at the same old links.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Rod Spade
United States
Elizabethtown
Pennsylvania
flag msg tools
badge
Avatar
mbmbmbmbmb
Re:Modern Art computer game w/ AI available for download
Snapper (#74358),

Great program! Thanks!

I think that one time I sold a closed-auction painting, and someone bought it for the exactly the same amount as what I had bid. I believe the auctioneer is supposed to win all ties.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Sean McCarthy
United States
Seattle
Washington
flag msg tools
designer
badge
Avatar
mbmbmbmbmb
Re:Modern Art computer game w/ AI available for download
Played it and got creamed.
But they're going down next time... well maybe I'll score higher than at least one...
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Ylaine Gerardin
United States
Boston
Massachusetts
flag msg tools
Avatar
mb
Re:Modern Art computer game w/ AI available for download
rodspade (#77222),

Thanks! You're right about the rules, but I probably will not be able to fix the program for a while, possibly a week. Hopefully this wont affect play too much.

Glad you're enjoying it!

Gabriel
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Burke Glover
United States
Unspecified
Delaware
flag msg tools
badge
Avatar
Great job!
Snapper (#74358),

This program is wonderful! Great job and thanks!

-Burke
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Rod Spade
United States
Elizabethtown
Pennsylvania
flag msg tools
badge
Avatar
mbmbmbmbmb
Re:Modern Art computer game w/ AI available for download
Snapper (#74358),

I found another bug:

The player on my right played a lone "double" card as the fourth from an artist. I played a matching card to end the round. I was then first to take a turn the following round - It should have been the player on my left.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Jeff Barrett
United States
Irvine
California
flag msg tools
mbmbmbmbmb
Re:Modern Art computer game w/ AI available for download
Snapper (#74358),

Nicely done. Thanks!
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Ylaine Gerardin
United States
Boston
Massachusetts
flag msg tools
Avatar
mb
Re:Modern Art computer game w/ AI available for download
rodspade (#91110),

Thanks for the catch. I've been really busy with school lately and haven't had time for any programming, but I intend to put out a new version toward the end of March. Here are the known issues I am going to fix:

-A computer is occasionally allowed to play a double auction on top of another double auction
-Fixed price auctions don't appear correctly in the history window (and possibly some other history issues)
-Some buttons are occasionally enabled when they shouldn't be
-Turns don't pass correctly when the round ends by completing someone elses double auction
-Auctioneer doesn't win a tie in the case of hidden auction

Thanks again!
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
1 , 2 , 3  Next »   | 
Front Page | Welcome | Contact | Privacy Policy | Terms of Service | Advertise | Support BGG | Feeds RSS
Geekdo, BoardGameGeek, the Geekdo logo, and the BoardGameGeek logo are trademarks of BoardGameGeek, LLC.