Walk-Forward Neural Networks
An adaptive network that retrains on a rolling window of recent bars and predicts ahead — genuinely out-of-sample at every bar, in real time.
A walk-forward network is the adaptive alternative to training once and reusing the result. As the indicator walks along the chart it repeatedly trains a fresh, short-lived network on the bars immediately before each bar and predicts a chosen number of bars ahead. Because each prediction only ever sees past data, the output is genuinely out-of-sample at every bar — a realistic picture of how the model would have behaved in real time.
Adapts as the market drifts
A standard network learns the market once and assumes that lesson keeps holding. A walk-forward network assumes the market drifts and keeps relearning from the most recent bars, so it adapts to a changing regime on its own. You set the lookback — how many recent bars it retrains on at each step — to trade off how fast it reacts against how much data it sees.
Built for real-time use
All the walk-forward calculations are cached in the plugin's internal memory and only computed as far as needed, so you can stop and resume training and use it live — the network trains and predicts on the latest bar as new data arrives. It is heavier to compute than a pre-trained indicator and runs on a single symbol with the feed-forward model, which is the trade-off for its adaptiveness. It is driven from one AFL call.