Mimai

You know that feeling when you're analyzing blockchain data and you realize... game developers are flying blind? 😳 Millions of players. Billions in transactions. Zero tools to predict who's about to leave, etc.. That hit me hard while diving into @solanagaming ecosystem. Because here's the truth: 👉 By the time a player churns, it's already too late. So instead of just tracking what already happened alone... I built something that predicts what's about to happen. =============================== 🎮 Solana Game Analytics, Player Behavior Modeling & Predictive Forecasting (ML-Powered Churn Prediction) =============================== A production-grade platform that analyzes 60M+ transactions from 12 @Solana games ( @AuroryProject, @staratlas, @genopets, @Stepnofficial, @PlayHoneyland, @maxxmxmcoin, @riftapp, @play_evio, @MiniNations, @_portals_, @nyanheroes, @MixMobRacer1), predicts which players will churn 14 days in advance with >85% accuracy, and auto-retrains its ML models whenever fresh blockchain data arrives. .... I love the sound of that BTW 😊 --- 1️⃣ What It Does ↘︎ Aggregates 60M+ on-chain transactions from 12 major Solana games ↘︎ Predicts player churn 14 days before it happens (>85% ROC-AUC accuracy) ↘︎ Auto-retrains 5 ML models and selects the champion automatically ↘︎ Tracks 11 behavioral metrics: activation, retention, cross-game behavior, etc. ↘︎ Delivers insights through 21 REST API endpoints + a gamified real-time dashboard. ..... Tbh. .. I love this UI so much 🥰 ..... Everything updates automatically. Everything is backed by verified Solana blockchain data...with clickable links. --- 2️⃣ How I Built It ↘︎ Data: 11 custom @Dune Analytics queries + custom classifier Python Script for on-chain analysis ↘︎ ML Pipeline: 5 algorithms (Random Forest, XGBoost, LightGBM, Gradient Boosting and Logistic Regression) trained in parallel with SMOTE balancing + auto-champion selection ↘︎ Backend: @FastAPI (21 endpoints, 72-hour caching, sub-100ms responses) deployed on @Railway ↘︎ Frontend (My favourite 😍): @nextjs + React 19 (100% TypeScript, Solana-themed UI, 30s auto-refresh) on @vercel ↘︎ Features Engineered: 10 predictive signals per user — activity patterns, momentum, consistency, recency --- 3️⃣ Why This Matters Solana gaming is exploding... but lack retention tools. This dApp (has Wallet Connect feature) gives: 📍 Early warning signals — catch at-risk players 2 weeks before they leave 📍 Cross-game insights — understand behavior across the entire ecosystem 📍 Zero-maintenance ML — models auto-improve with new data 📍 Actionable intelligence — trigger campaigns, interventions, VIP monitoring For Solana: 📍 First comprehensive gaming analytics platform 📍 Open-source models + all 11 Dune queries for community use --- 🔗 Explore the System • Full Demo: drive.google.com/file/d/1Pr1… Built by: @defi__josh Inspired by: @indiedotfun @oladeeayo @DairusOkoh @apostleoffin @AnalyticSages @alphabot @superteam @KttyLabs @Purrvananew --- Fr... This one fused blockchain analytics, Machine Learning, full-stack engineering, and DevOps into a single autonomous system. What excites me most? 👉 Game studios can now predict churn before it happens. 👉 The system retrains itself — zero manual work. 👉 The UI is bada_ss 😊😇 Because if Web3 gaming is going to scale, we need tools that help developers keep the players they earn. 💯 --- BTW.. 👉 What's the #1 reason you think players churn from blockchain games? --- Gm Gm 😉