Personal AI Trading

A rule-based crypto bot with a hard risk layer — and an honest account of whether it works.

What it does

Three indicators vote on every bar. Before any entry is allowed, the bot searches 1.5 million labelled historical situations for the ones most similar to right now, and requires enough of them to have worked out. Every order passes a risk manager that sets a stop and a target at entry.

A supervisor pauses trading on losing streaks, bad weeks, stale data, exchange errors and violent markets. Position size never grows after a win, and profit is locked away in steps where the bot cannot reach it.

Does it make money?

Barely, and not enough to run. Measured on real Binance data, the validated configuration returns about +0.24% a year. On a 400 USDT account that is 96 cents, against roughly $50–70 a year to host it.

An 80% win rate was also built and tested, by moving the target closer to the entry. It won 80.1% of 286 trades and lost 7.9% over the year. A win rate is not an edge; expectancy is.

The risk controls held throughout: worst single trade −2.25%, worst drawdown −2.52% against a −20% ceiling.

Latest validation

loading…