Documentation Index
Fetch the complete documentation index at: https://www.tradenet.org/docs/llms.txt
Use this file to discover all available pages before exploring further.
Why We Built TradeNet Terminal
I want to explain why this thing exists. Not the marketing pitch. The actual reason.
The Frustration
I have been trading crypto derivatives for years. And every single tool available to me had the same fundamental problem: static models.
Coinglass, Kingfisher, Exocharts. They all predict where liquidations sit. That is not the issue. The issue is their models do not adapt. They estimate leverage distributions using fixed assumptions. When market conditions shift, when leverage behavior changes, when a new regime starts, the model stays the same. You are trading on projections that were accurate yesterday but may not be today.
I got tired of trusting models that did not keep up with the market.
The Question That Started Everything
What if the model adapted to the market instead of the other way around?
If you can model open interest distribution across exchanges, track leverage ratios, and project where margin calls will trigger, you can build a map of future liquidation events. But the hard part is not the initial model. The hard part is keeping it accurate. We built a calibrator that uses real forceOrder events as feedback, adjusting leverage weight distributions in real time based on actual liquidation data. Not a static formula. A self-calibrating system.
That is the idea. Simple to state. Hard to build. But that is what we set out to do.
Why It Had to Be a Desktop App
The first decision was the tech stack. And the answer was obvious to me: this could not be a web app.
Browser-based charting tools hit a ceiling fast. Canvas rendering is slow. WebSocket connections are fragile. Memory management is a constant fight. Every existing terminal runs in a browser or wraps one in Electron. And they all feel like it. Sluggish charts. Laggy data. Frozen tabs when the market spikes.
We built TradeNet in Rust with the Iced GUI framework. The rendering pipeline uses wgpu (the rendering backend for Iced’s canvas) for GPU acceleration. Charts render on your graphics card, not your CPU. Data processing runs on dedicated Python/FastAPI backend servers on Hetzner. Your machine handles display. The servers handle the math.
The result is a terminal that stays smooth when every browser-based tool locks up. During a volatility spike, when data throughput increases tenfold, the terminal does not skip a frame. That is not a nice-to-have. When the market is moving fast, that is exactly when you need your tools to work.
Rust was the right choice for another reason: reliability. No garbage collector pauses. No memory leaks over time. The terminal runs the same at hour one as it does at hour twelve. If you run a browser terminal for a full trading session, you know what I mean. Eventually it starts crawling.
Multi-Exchange Aggregation
Crypto liquidity is fragmented. A liquidation cascade on Binance affects Bybit pricing. OI changes on OKX ripple through Hyperliquid. If you are only watching one exchange, you are seeing a fraction of the picture.
TradeNet aggregates data from Binance, Bybit, OKX, and Hyperliquid. The liquidation prediction model ingests data from three exchanges (Binance, Bybit, OKX) via WebSocket. Hyperliquid is connected on the frontend for trades and orderflow. The heatmap reflects the combined picture from the three backend exchanges.
This matters more than most people realize. A liquidation cluster that looks small on Binance alone might be massive when you combine all three exchanges. The aggregated view shows you the real magnitude.
The Self-Calibrating Algorithm
The prediction model is not a static formula. Markets change. Leverage behavior shifts. New products launch. Funding rate dynamics evolve.
A fixed model would decay over time. It would be accurate for a month, then drift as the market adapts. We built the algorithm to recalibrate continuously. The calibrator uses forceOrder events from all three backend exchanges as feedback, with soft attribution in percent-space and volatility scaling. It adjusts leverage weight distributions based on actual liquidation data. The model today is different from the model a week ago, because the market is different.
This is the piece that took the longest to get right. Building a model that predicts liquidation clusters is hard. Building one that stays accurate as conditions change is significantly harder. We spent more time on the calibration system than on any other single component.
Where This Goes
The liquidation prediction heatmap is the starting point. Not the endpoint.
The vision for TradeNet is a terminal that gives you a structural edge across every dimension of crypto derivatives trading. Liquidation prediction was first because it is the highest-impact feature. But we are building toward a full suite: screener, volume profile, advanced OI analytics, cross-exchange correlation tools.
Every feature follows the same principle. Adapt to the market in real time, not rely on static assumptions.
We are in closed beta now. The terminal is real, it works, and traders are using it daily. If you trade crypto derivatives and you are tired of static models that do not keep up, this is what we built for you.
Martin
