Constructing predictive filters with RSI and walk-forward tests

Fix the observation stream first

Indicator construction is sequenced after market observations are characterized and digital-signal-processing tools are reviewed. Predictive filters are constructed to produce a forecast from ordered market observations at a defined sampling interval.

In this workflow a Kalman filter is a forecast model that updates an estimate from ordered price, volume or breadth observations at a defined sampling interval and lookback. Strategy construction uses both daily bars and intraday bars as sampling choices.

Build the forecast-style indicator layer

The indicator layer under construction includes zero-lag filters, predictive filters, Chebyshev filters and a Rocket RSI variant.

The relative strength index is an oscillator built from ordered price observations over a fixed lookback and used here as a forecast-style transform, not as a standalone trade rule.

Assemble and test one procedure

Optimization is framed as a search for constructions that remain usable across timeframes and across more than one market. Walk-forward analysis is presented as a distinct technique for testing the assembled strategy as one procedure.

Walk-forward analysis is a procedure that turns entry, exit and abstention rules into one testable system by rolling the holding-period evaluation through successive out-of-sample windows.

Ordinary school math is enough

The construction path is described as reachable from high-school algebra and trigonometry rather than from specialized prior coursework.