The scanner is a second author on the same desk: closed Binance USDT-M candles, Wilder RSI and MACD, confirmed pivots, then a row tagged Scanner — or Both when the desk already had that pair/side/timeframe.

Scanner pipeline from klines to desk row
Cron on the :07 / :22 / :37 / :52 minutes. M15 hits are stored but not pushed as outbound alerts.

What it actually computes

On each pass it pulls top USDT-M volume, loads klines, then:

  • RSI 14 with Wilder’s smoothing (the same construction as the original 1978 relative strength index).
  • MACD 12 / 26 exponential averages with a 9-period signal line — Appel’s classic histogram family.
  • Pivots with a left/right window so a “high” is only a high after later candles have printed. That is how we avoid open-bar repaint.
  • Regular and hidden from whether price and oscillator extremes move together or against each other.

That is the same family of ideas you get from a TradingView RSI/MACD pane, except the print is frozen on close and dumped into the public feed with an entry snapshot.

Hit rate by public sourceDesk 51.5%; Scanner 47%; Desk + scanner 48.5%51.538.625.712.90DeskScannerDesk + scannerHit rate by public source
Desk is the bulk of history. Scanner is newer and still a smaller scored sample. “Both” means the same setup hit both pipes inside the dedup window.
Scored scanner-style rows by oscillatorRSI 2,194; MACD 5082,1941,645.51,097548.50RSIMACDScored scanner-style rows by oscillator
RSI prints more often than MACD on this book. Oscillator is empty on older desk rows that never carried taxonomy.

How it shares the feed

Two-hour dedup on pair + side + timeframe. If the desk already has the setup, the scanner merges the row to Both instead of cloning it. Public labels never name internal origins — only Desk, Scanner, or Both.

See a pair’s M15 / H1 / H4 stack on any pair page, or filter scanner only.