Market signals, computed on a computer I can point at.
A single Go binary polls end-of-day prices, stores them in SQLite, computes its indicators from scratch, and backtests a moving-average crossover — all on a Raspberry Pi 3B in my room, reachable only through a Cloudflare tunnel. No open ports, no cloud, no framework. The numbers above are the machine reporting on itself.
Tracked symbols
Your watchlist
Your watchlist is saved in this browser and nowhere else. The Pi fetches these on request, holds them in memory for 15 minutes, and never writes them to its disk — restart the server and every trace is gone.
Price and moving averages
Relative strength index, 14 periods, Wilder smoothing. Bands mark 30 and 70.
Strategy vs buy and hold
The signal is read on each day's close and filled at the next day's open, never the same close it was computed from. Entry and exit are both charged the cost above. Long only, no leverage, daily bars, no survivorship correction — a sanity check on the logic, not a claim about future returns.