17
Simulations Evaluating game system behavior Petri Lankoski Södertörn Univeristy

Simulations: Evaluating game system behavior

Embed Size (px)

DESCRIPTION

Global Game Jam Stockholm Presentation some additional slides & bullets (that I removed to fit the presentation in 20 minutes)

Citation preview

Page 1: Simulations: Evaluating game system behavior

Petri Lankoski Södertörn Univeristy

SimulationsEvaluating game system behavior

Page 2: Simulations: Evaluating game system behavior

Petri Lankoski Södertörn Univeristy

SimulationsGame systems with random component are

complex

Simulations can help to understand how a part of the system behaves One does not need ready game for simulation

Does not replace playtesting But simulation can show the features work in the

long run

Balancing weapons & troops non-symmetrical things are hard to balance

Page 3: Simulations: Evaluating game system behavior

Petri Lankoski Södertörn Univeristy

Simulating a game systemModel

sum of two six sided dice -> sum of two random numbers between 1 to 6

Weapon: change to hit, damage dealt & fire rate

Simulating system Run model many times to learn how the system

behaves Run 50000 times and calculate distribution or

averages, average damage per minute, etc.

Page 4: Simulations: Evaluating game system behavior

Settlers of Catan

Page 5: Simulations: Evaluating game system behavior

Petri Lankoski Södertörn Univeristy

Simulation How the players gain resources

Simplified Robber vs no robber discard Only resource amount simulated, not types

Assumptions Four player game 0-3 resources at hand when ones turn ends

Model for using resources One specific board set-up

The results does not vary much board to board The results can vary with not optimal settlement placements

50 000 iterations used

Page 6: Simulations: Evaluating game system behavior

Petri Lankoski Södertörn Univeristy

Simulation set-up• 4 victory point set-up

• Settlements -> cities• 6 victory point sim• 1&2) 8 victory point sim

Page 7: Simulations: Evaluating game system behavior

Petri Lankoski Södertörn Univeristy

Model#!/usr/bin/python import randomfrom collections import Counter

# board model (2 victory points) field1 = { 2: {'white': 0, 'blue':0, 'red': 0, 'orange': 0}, 3: {'white': 0, 'blue':0, 'red': 1, 'orange': 1}, 4: {'white': 1, 'blue':1, 'red': 0, 'orange': 0}, 5: {'white': 0, 'blue':2, 'red': 1, 'orange': 0}, 6: {'white': 1, 'blue':1, 'red': 1, 'orange': 1}, 8: {'white': 1, 'blue':1, 'red': 1, 'orange': 1}, 9: {'white': 1, 'blue':0, 'red': 0, 'orange': 1}, 10: {'white': 1, 'blue':0, 'red': 1, 'orange': 1}, 11: {'white': 0, 'blue':0, 'red': 1, 'orange': 1}, 12: {'white': 0, 'blue':0, 'red': 0, 'orange': 0}}

The above model does not contain handling for robber The code for simulating this model is bit more complicated

Page 8: Simulations: Evaluating game system behavior

Petri Lankoski Södertörn Univeristy

Resource gain

Page 9: Simulations: Evaluating game system behavior

Petri Lankoski Södertörn Univeristy

Robber Effect

Page 10: Simulations: Evaluating game system behavior

Petri Lankoski Södertörn Univeristy

Balance of set-up1 2 3 4

White 2.0553 2.6120 3.1700 3.7267

Blue 2.0761 2.6593 3.2396 3.8224

Red 2.0808 2.6661 3.2496 3.8348

Orange 2.0892 2.6745 3.2605 3.8454

• Resource gain for each color is very similar• White might have small disadvantage

Page 11: Simulations: Evaluating game system behavior

Petri Lankoski Södertörn Univeristy

What can one learn?

Easy to run what if scenarios Robber -> discard all Discard if more than four resources

Estimating the costs for building

Balance of the the initial set-up

Page 12: Simulations: Evaluating game system behavior

Petri Lankoski Södertörn Univeristy

Monopoly

Page 13: Simulations: Evaluating game system behavior

Board & Movement

Petri Lankoski Södertörn Univeristy

Chance to end Up In a square1/40 = 2,50%?

3 doublesin a row

1/16 Card takesto Jail

A player can increaseprobability to land to These squares(out with doubles)

Page 14: Simulations: Evaluating game system behavior

Petri Lankoski Södertörn Univeristy

Chance to Land at a Square

Page 15: Simulations: Evaluating game system behavior

Petri Lankoski Södertörn Univeristy

Break Even Times

Page 16: Simulations: Evaluating game system behavior

Petri Lankoski Södertörn Univeristy

What we learnedStaying in prison strategy alters changes to

land other squares Long prison stay good at the end game

Break even time downward trend is good

Breakeven times are long Slow start Note that one cannot build before owning all

squares with that color

Page 17: Simulations: Evaluating game system behavior

Petri Lankoski Södertörn Univeristy

Thank You!

http://petrilankoski.wordpress.com