Data Science
Model that allows you to choose the optimal probability of falling out of skins from loot boxes
Lootbox
virtual container in computer games, which contains items of different value and purpose.
Skin
virtual change in the appearance of the player, weapons or other elements in the game.
Client:
shop of skins and loot boxes
Main information
How it works
There is a chest on the customer's website
User:
buys
a chest
a chest
opens it
wins a prize - skin from the game
Features
The mechanics of opening loot boxes are similar to the lottery - the player never knows what exactly he will receive and whether it will be worth the money spent.
The odds ratios are now being matched using an algorithm that is slow.
Our task
Develop a system that would automatically select the odds ratios
Conditional scheme
virtual box with gifts
Lootbox:
Opening cost:
100 rubles
For each box,
the following is set:
set of items that can fall out of it
the probability of these items being dropped
Skins drop rate:
Gift 1
cost 50 rubles
Probability:
Gift 2
cost 200 rubles
Probability:
The user has a chance
to win a prize more expensive than the loot box itself
to win a prize more expensive than the loot box itself
Our task:
adjust the coefficients so that the user is interested and the customer is profitable:
expensive gifts -
not too small probability of falling out
not too small probability of falling out
cheap gifts -
the maximum probability of falling out
the maximum probability of falling out
Eventually:
Player
keeps the excitement
Customer
works in a plus
Initial data:
What have we done
Histogram of the distribution of the chances of dropping items in a loot box
Project team
Project manager
Konstantin Kubrak
Data analyst
Oleg Durygin
Interesting fact
This is an international project, we gained experience of working with a company from Scotland
Difficulties of the project
domain specificity
a lot of math
Contact us