From a single idea to live trading on a real account — through a graveyard of failed strategies, honest backtests, a 24/7 server, a dashboard, and my first month live.
My hypothesis was simple: humans trade inconsistently and emotionally, while the market never sleeps. A disciplined algorithm — even on a small deposit — should beat manual trading, as long as it has a real edge.
Automated passive income on Bybit crypto futures, built with copy-trading in mind.
No guessing. Only rules, validated on historical data with real trading costs included.
Prove the edge on data first — then risk money. Never the other way around.
I didn't fall in love with ideas — I killed them with data. Each was tested on history with taker fees and slippage.
| Concept | What it was | Verdict |
|---|---|---|
| v1 — MACD + EMA, 5m | high-frequency scalping | Fees ate the edge |
| v3 — Pullback, 15m | pullback to EMA in a trend | No edge in the signal |
| v4 — Mean Reversion | Bollinger + RSI reversals | Unprofitable · 60 combos |
| v5 — Confluence, 5m | EMA + Stoch + Volume scalp | Unprofitable |
| v2 — the one that worked | a systematic trend approach | Edge survived out-of-sample |
After ~220 tests: on 1–15 minute timeframes the signal is essentially a coin flip, and fees push it into the red. The market is too efficient there. That cut off the dead ends — and left what actually works.
The smaller the timeframe, the more of the profit fees consume. The edge only survives on higher timeframes, where moves dwarf the costs. Scalping is mathematically unprofitable — proven on the numbers.
The real advantage isn't the signal — it's trading the right instruments. I systematically rank the market and hold only its current leaders, refreshed on a fixed cadence. Strong stays strong.
Smaller risk per trade and a cap on open positions produce better long-run results. Greed — 2%+ per trade — destroys the account through volatility.
Trend-following wins ~37% of trades, but winners run 2–2.5× larger than losers. Profit comes from rare long trends, not from hit rate.
The signal itself stays proprietary. What actually keeps the account alive is a strict, multi-layered risk framework applied to every single trade — the part that matters most, and the part I'm glad to show.
No position ever puts more than 1% of capital at stake. Size is derived from the protective stop, not from conviction.
After a set daily loss the bot stands down until the next day — no revenge trading, no spiralling.
At most three positions open at once, spread across different pairs, so no single move can sink the account.
Every trade carries a stop from the moment it opens; a trailing stop then locks in profit as the move runs.
It's easy to curve-fit parameters to the past and admire beautiful numbers. I deliberately guarded against it — validating on data the optimizer had never seen, and pooling every trade into an honest portfolio model.
Parameters were tuned on one window, then profitability was tested on a separate, unseen one. The edge stayed positive — so it isn't overfitting.
Every backtest bakes in a 0.055% taker fee plus slippage. Without that, the numbers lie.
The edge is real but thin: results depend on the period, and historical drawdowns reached ~40%. I know this, and I don't hide it.
The bot runs on a dedicated server in Germany, around the clock. I can close my laptop and it keeps trading.
systemd brings it back within ten seconds of any crash, and starts it automatically after a reboot.
Startup, every trade, errors, a daily report and a weekly digest — delivered straight to my phone.
After a restart it picks up open positions from its database and keeps managing them without a hitch.
From a bare page to a full web trade journal: balance, open positions, history grouped by week, an equity curve and drawdown — all live, behind secure HTTPS.
June 2 – July 3, 2026 · real money, not a backtest · 56 closed trades.
The strategy behaved exactly by design: a 37% win rate, but the average winner ran 2.5× the size of the average loser. Risk management kept the drawdown shallow.
Honestly: one month and 56 trades is a small sample, and a tough market stretch hasn't arrived yet. This is an encouraging start, not final proof. Past performance does not guarantee future results; futures trading carries the risk of losing capital.
A bot that trades on its own, 24/7, with managed risk and full transparency. My results are live and open in real time — and now you can copy them.
Copy my trades on Bybit →This material is for informational purposes only and is not investment advice. Crypto futures trading carries a high risk of losing capital. Past performance does not guarantee future results.