Forum > Pazaak Probabilities

Mator
Joined: Dec 13
2013-12-19, 06:18
Hello!

I'm a fellow pazaak fan, and I've recently gotten into probability analysis of the game. I currently have a program that allows me to simulate millions of games every few seconds with a wide range of conditions, including player hands. It's still in need of a bit of work, but it's capable of producing some interesting data. Here's what it looks like:

http://puu.sh/5Rjxy.png

The "match" play of 3 games with 1 hand is not yet coded, but soon I'll have written the code to do this as well! :)

Here's some sample data:
[hidden][quote]
Simulating 1000000 side-by-side games
Player 1's hand: +/-6, +/-5, +/-4, +/-3
Player 2's hand: -5, +5, +3, -4
p1wins: 297820
p1busts: 49159
p1full: 9238
p2wins: 58397
p2busts: 275154
p2full: 22666
ties: 643783[/quote][/hidden]


I'll make this program publicly available, including the source code, after I feel it's 100% completed. You can find more information on this endeavor over on LucasForums.


-Mator

DarthRevan
Joined: Nov 09
2013-12-19, 08:51
In short: This is amazing!

I have never heard of a Monte Carlo simulation before, but I am not experienced in probability analysis ;)

The simulation results might help to improve the Pazaak Cantina AI: e.g. the algorithm, that picks the AI's sidedeck (10 cards, from which 4 are chosen) could benefit from such data sets.

Please let me know, if you have any suggestions for the AI :)
Developer of Pazaak Cantina.
Definition: 'Love' is making a shot to the knees of a target 120 kilometers away using an Aratech sniper rifle with a tri-light scope (HK-47)

Mator
Joined: Dec 13
2013-12-19, 10:28
The Monte Carlo method is, in short, where you generate a bunch of random results within a system and then use them as a means to understand how the system behaves. It was invented by a gambler, so it seems only natural to apply it to pazaak. See wikipedia for a more intelligent sounding explanation.

I'll be able to test decks against each other and the effectiveness of various strategies once the program is complete. E.g. How much standing on 19 increases your chances of winning in regards to your deck/hand as well as your opponent's deck/hand. Same with standing on 18, 17, or 16.

Anywho, I'll pop back in here once I've made my next major update to the program and start providing data. If anyone wants to request a specific simulation (or a specific hand/deck face off) please do so.


-Mator