game balancing sf5

12
Balancing opportunity in SF5 with linear optimisation Author : Jean-Philippe Mallette

Upload: jean-philippe-mallette

Post on 22-Jan-2017

123 views

Category:

Documents


0 download

TRANSCRIPT

Balancing opportunity in SF5 with linear optimisation

Author : Jean-Philippe Mallette

Goal of Balancing in SF5

The goal is to Identify which character to make more or less powerful.

Step to Complete the Balancing Analysis

1. Get the battle data (win / loss) for every character against all other character2. Turn the raw data into a matchup chart data3. Turn the matchup data into a payoff chart data4. Perform the linear optimisation5. Plot the result and make recommendation

Get the raw dataWeb scraping on a fan database page http://www.eventhubs.com/tiers/sf5/

For every character web page, I have extracted the match avg. score against all the other character.

Prepare the data

Raw Data into a Matchup Chart

divid the result by 10 to make the scale to 1

Example with Alex and ken

Matchup chart into a Payoff Chart

Perform the Linear optimisation

Assumption Of the Linear Optimisation

Assumption

The assumption of our model is that at time t, a player community Ct consists of t-1 players, and a new player joins the game, choosing a character who maximizes her win probability in a game against a randomly chosen player from C.

We can make game balancing adjustment by knowing which character should play if player want to maximizes theirwin probability

The Linear Optimization Model

We also add additional constraint related to frequency of play rate. We are going to solve this linear optimisation for every possibility of xi play rate frequency probability. We will then plot the minimum probability and the maximum probability that a player play character i

we only consider probability from 0 to 1 by 0.1 (0, 0,10. 0,20 … to 1) we could have considered more

xi = play rate frequency of xivij = matchup data of fighter i vs fighter j

Here the graph shows the minimum and maximum probability the play rate of eachcharacter base on the assumption of our model.

Recommandation - Player should basically play chun_li in

30% to 100% of their match. Let’s lower her power.

- Player should never play zangief. Let’s increase his power.

Result & Recommandation

Github code

Link : github.com/tascor/Game-Balancing

THANKS FOR LISTENING !