Simulating 70,000 Players

The next step in the iGaming analytics project was to create realistic synthetic data to analyze retention and churn — even without access to real players.

Main Goals of the day

Generate 70K synthetic players using numpy and pandas
Split players into control and treatment groups
Simulate sessions, deposits, and feature usage
Add a binary churn indicator (player left or not)

Step by Step

📍 Step 1: Defined statistical distributions for sessions (Poisson) and deposits (Gamma)
📍 Step 2: Introduced treatment boost (+2 sessions, +€20 deposits)
📍 Step 3: Simulated feature adoption (Binomial: 0.4 vs 0.7)
📍 Step 4: Added dynamic churn probability by player group

Insights

Treatment users showed higher average deposits and slightly more sessions.

This mirrors what happens in feature rollout or promo exposure scenarios in real iGaming data.

Code Snippet

import numpy as np
import pandas as pd

n = 70000
groups = np.random.choice(['control', 'treatment'], size=n)
sessions = np.random.poisson(5 + (groups == 'treatment') * 2)
deposits = np.random.gamma(shape=2, scale=50, size=n) + (groups == 'treatment') * 20

Next Step

Run a statistical test (A/B) to check if the retention difference is significant Add retention rate and p-value calculation with Chi-square test