A Monte Carlo simulator of the card game's V-format. It estimates the probability of, in the same game, having any given combination of cards (however many copies per card), across any given combination of game timings, while prioritizing any set of cards to keep in hand. It tries to simulate real play for the sake of realistic results.
E.g. "probability of having a 'PG' by the end of turn 2 AND a grade 3 unit to ride at the start of turn 3", or "probability of having Golden Beast Tamer at the start of turn 3, AND, having 2 copies of Alice in the main phase of turn 3, AND, having either 1 Jumping Jill or 1 Purple Trapezist on turn 4's main phase".
The user must provide an input file containing at least two fields:
- a deck,
- and a test (propositional formula) to be evaluated.
The program runs as many simulations as instructed and subsequently outputs the percentage of simulations that passed the test (i.e. the percentage of games where the formula was evaluated as being True). Though it tries to approximate real games, this is a simple simulator. It does not use AI, nor does it simulate calling rearguards, defending, or selecting which cards to remove from the game during G-assist. It does not use unit effects except for the starter (draws when rode upon), Draw Trigger's draw effect, and the gift mechanic, which also draws when riding an 'accel' gifted unit if the 'gift' in the input file is 'accel2'. Therefore, the output probability serves as a baseline that is always lower than that of real games where draw/cycle/tutor/etc effects are used. The purpose of the simulator is, then, to estimate said probabilistic lower limit, but also to provide an idea of the difference of outcome when: varying the number of copies of cards (including triggers), varying the damage received, softly prioritizing different cards, and even aggressively looking for key cards during mulligan by sending back decent ones.
Explore the help menu for detailed explanations of each field of the input file. Have fun!
- Python 3
Open a terminal in the script's directory and type the following command to run the script:
python3 raibot.py
You will then be prompted to type in your input file's name and the number of simulations you would like to run. Results typically deviate < 1% with 100 000 runs (takes < 10 seconds). Type in a command with the following format:
<file name> <number of runs> [seeGame]
For example, using the input file gbt.txt (a sample file with a Golden Beast Tamer deck):
gbt.txt 100000
You may opt to include seeGame as the last argument to visualize games. Select a small number of runs so as to not flood your terminal.
gbt.txt 1 seeGame
You can also type about for tips and guidelines, or help for input file formatting specifications.
If the program crashes due to input file misformatting, it is likely caused by spaces or empty lines being present where they shouldn't be, or missing where they are required.
Looking to implement:
- Plussing - adding draw and soul charging effects
- Soul - currently the only zones accounted for are the player's hand and VG circle
- Literal negation - why not
- P-format support - starter unit behavior, strides, quadruple drive, ...
Joao Silva
This project is licensed under the MIT License - see the LICENSE file for details.