How it started
March 2021. COVID lockdown, alone in a small apartment for months with nothing but a Raspberry Pi 4 with Debian Linux and too much free time. I started writing shell scripts to backtest trading strategies I found online. Golden crosses, moving average crossovers, the usual stuff everyone recommends.
None of them worked. Not a single one survived proper backtesting.
But I kept going.
The process
- 2021 Started with Bash shell scripts on Linux, testing across multiple pairs. Each backtest took 10 minutes to run. Switched to ksh after discovering it was noticeably faster, then parallelized the runs and optimized the code. Eventually got a single test down to 1–2 minutes. Slow by any standard, but it built the foundation.
- 2022 Narrowed the focus to BNB/USDT - the most promising pair across all early tests, and the one showing the strongest potential for a clean, consistent edge.
- 2023 Upgraded to a more powerful machine dedicated solely to running backtests - 24/7, without interruption.
- 2024 Rewrote the backtester in Python and added a web interface. Multithreading brought a single test down to ~1 second, but the web layer introduced new bugs and slowed down development. Scrapped it.
- 2025 Rebuilt the entire engine in C. A single backtest across 8 years of data now runs in 0.005 seconds. Over 30 base strategies and 20+ filters tested, but never stacked blindly. Adding too many filters at once either caused overfitting or simply stopped working. The final configuration uses only 2-3 filters, each earning its place independently.
- 2026 One strategy survived. Profitable in backtesting and out-of-sample validation across every test period. By this point, tens of millions of parameter combinations had been run across those five years - each one a data point, not a guess. That's what powers the signals on this site.
The backtester running 1.25M parameter combinations at 512 tests/sec
Why BNBUSDT?
BNB/USDT was tested from the beginning - and it consistently outperformed every other pair in the testing pipeline. Rather than spreading thin across a dozen markets, I chose to focus on the one pair where the edge is strongest and most stable.
The numbers back it up: 85.2% win rate, 1.92 Profit Factor, and a perfect 10/10 out-of-sample score - tested across 10 different time periods with 50/50 to 95/5 train/test splits. All 10 out-of-sample periods were profitable. No overfitting to historical data.
Additional pairs are currently in development: BTC/USDT targeting Q2 2026, followed by ETH/USDT in Q3 2026.
Why share it?
My goal is simple: build something reliable enough that trading income gives me more time with my family. No lambos, no "financial freedom" promises. Just a methodical, data-driven approach to markets, shared transparently.
Every backtest result on this site is real. Every live trade is logged publicly. If the system stops working, you'll see it in the numbers before I say a word.