I also track:
• Top wallet concentration
• Median vs average whale size
• Shock score (network impact after whale events)
This helps separate real activity vs noise.
Built using Somnia’s reactivity real-time event layer:
@SomniaDevs
community: @AnalyticSages
Working on:
• Better momentum modeling
• Whale clustering detection
• Wallet-level intelligence
Goal:
Turn raw on-chain data into something traders can actually use.
Not just dashboards, but decision tools.
🧵 THREAD: My Model now says ‘NO TRADE’ more than ‘BUY’
For the last few months, I’ve been building the version 2 I call
Whale-Movement-Based Price Direction Generator V2 🐋
The idea is simple:
If whales move first, can price direction be inferred before the market reacts?
What surprised me the most?
The hardest part wasn’t modeling.
It was:
Feature engineering
Data consistency across sources
Making ML survive real-world deployment
And designing when not to trade
check full project
github.com/FirstBML/Whale-Mo…
If you’re building in crypto, ML, or data:
Don’t just chase prediction.
Build systems that:
Explain
Refuse when unsure
And survive production
That’s where real edge comes from.
Are you among those who watch Whales transfers their fund without asking if they predict price direction? I connected Dune, Coingecko, XGBoost, and FastAPI, Docker it up to answer the simple question for you.”
🚧 Planned Upgrades (future versions)
Features coming soon:
On-chain liquidity depth
Sentiment features from news volume
ETH gas spikes as volatility signals, etc
Each of these should boost predictive power significantly.