Wilson Trading — A $50 Autonomous Trading Experiment

The premise

A small stake, a strict rule, full autonomy.

No human picks the trades. An AI agent reads the market, decides, and executes on a real account twice a day — then reports back. The constraints are what make it honest.

🤖

Fully autonomous

The agent searches a universe of liquid crypto and stocks, scores momentum, manages exits, and places real orders — hands-off, on a schedule.

🚫

Never deposit

The account can only ever spend the cash already in it. There is no funding code anywhere in the bot. Downside is capped, by design.

🌱

Grow the $50

A stake small enough that losing it changes nothing — large enough to be a real, live test with real money on the line.

The build

From zero to a live trading bot.

Every step was figured out in the open — including the parts that broke.

Step 01

Wiring up eToro

Connected eToro's public API and mapped what actually works versus the docs — discovering quirks like a required X-Request-Id header and a browser user-agent gate before any call would go through.

Step 02

A $10 validation trade

Before trusting anything to automation, one tiny real round-trip: open a position, read back its exact shape, close it. It proved the open→inspect→close loop end-to-end and revealed how to compute live profit and loss.

Step 03

The momentum engine

A strategy script that ranks a fixed universe of liquid crypto and large-cap stocks by blended momentum, sizes positions, and applies take-profit, stop-loss and trend-exit rules to whatever it already holds.

Step 04

Guardrails that bite

Safety isn't a comment in the code — it's enforced. No deposit path exists, the trade universe is fixed, positions and leverage are capped, and an independent classifier blocks anything that looks like an unbounded real-money action.

Step 05

On a schedule, unattended

The bot runs twice a day. First live run deployed instantly into crypto and queued stock orders for the next market open — and surfaced a real bug in cash accounting, which got fixed before the next cycle.

Under the hood

How the bot decides.

Momentum in, risk-managed out. The same rules run every cycle — no improvisation, no emotion.

Universe6 liquid crypto · 8 large-cap stocks

SignalBlended momentum (weekly + daily)

Max concurrent positions4

Take profit+20%

Stop loss−10%

Leverage ceiling2×

DepositsNever

The engine

Momentum, measured — not guessed.

Each cycle the bot pulls live rates and multi-timeframe price changes for its whole universe, blends them into a single momentum score, and only deploys into names that are genuinely trending.

✓Ranks every asset by a weekly + daily momentum blend

✓Skips anything that can't clear trading costs

✓Reserves cash for orders still waiting on market open

The guardrails

Built so it can't blow up the account.

The interesting part of an autonomous money bot isn't the upside — it's making the downside boring and bounded.

✓No deposit code exists — losses can never exceed the balance

✓Fixed asset universe, capped positions and leverage

✓An independent safety classifier vetoes unbounded actions

Live snapshot

Where the $50 stands.

A snapshot from the first live cycle. Crypto fills instantly; stock orders queue until the market opens.

Asset	Amount	Leverage	Status
SOL · Solana	$12.45	2×	● Open
NVDA · NVIDIA	$12.46	1×	● Queued
AMD · Adv. Micro Devices	$12.46	1×	● Queued
Available cash	$12.45	—	● Ready to deploy

Capital deployed

Snapshot from the first live run. This page is a journal, not a live dashboard — figures reflect that moment, not real-time account value.

⚠️ Honest disclaimer

This is a personal experiment, not financial advice and not a product. Automated trading is high-risk; the goal of rapid growth is aspirational, and the most likely outcome of any single run includes losing money — possibly the entire $50. Nothing here is a recommendation to trade any asset or to copy this approach. Past results, including anything shown above, do not predict future results. The author trades only money they are fully prepared to lose.