About

One engine, run where anyone can watch

Aegium Labs is a one-person project from Germany. It ranks crypto perpetual futures against each other and runs a market-neutral paper engine on that ranking, with every result published daily.

It started as a private research tool, and most of what was tried did not survive. Every attempt to time entries and exits failed out of sample. Picking the best strategy mix on training data did worse than picking at random. One promising edge turned out to be a lookahead bug in the backtester. The Journey page has the full list.

What survived is simple. Rank every liquid coin against the others, hold the strongest long and the weakest short in equal dollar amounts, and rebalance on a schedule. A free account opens the same four engines the public board ranks — marked hourly, inside their backtest cone, with alerts.

The engine runs on paper, at real exchange prices, in the configurations shown on the track record. Every day is frozen, hashed and timestamped in Bitcoin, so the record cannot be improved afterwards.

Principles

Four rules the project keeps

Rank, don't predict

Aegium does not forecast prices. It measures which coins are strong or weak against the rest of the market right now, and ranks them.

Losses stay on the page

Every drawdown is published with its date. The record is never reset to make a chart look better.

Paper is called paper

No real money trades these engines. The prices and the fees are real, the book is not, and every page says so.

Results public, method in-house

What the engines made is public and can be verified by anyone. Which coins they hold and how the score is built stays with us.

Methodology

How the engine works, in four steps

The idea fits in four lines. The work is in the testing, and in leaving the results alone.

01
Rank the field
Every liquid perpetual future is scored against all the others at the same moment, on the same inputs, so a coin is only ever strong or weak relative to the rest.
02
Long the top, short the bottom
Equal dollars on both sides, so the result comes from which coins were picked, not from where Bitcoin went.
03
Test it on data it has not seen
An idea has to hold up out of sample before it goes live. Most did not, and the Journey page lists them.
04
Run it forward, in the open
Paper-traded at real exchange prices. Results are published every day and timestamped in Bitcoin, so they cannot be edited later.
Free while the record builds

Watch the work in public.

Every engine result is published daily, drawdowns included. A free account opens the engine up close.