forecasting-ch01-marketslisted

# Chapter 1: The Market Discovers the Future ## Scope and intake Use when market prices or trading signals need a persistence benchmark and information-timing audit. Do not use for investment allocation advice based only on chart patterns. Ask only for unresolved material inputs: Is the target spot, futures or an event-contract price? What is the quote frequency? When could the signal have been known? Are transaction costs relevant? ## Input contract and additional evidence A regular OHLC price series of at least 30 rows with consistent bounds (low at or below open and close, high at or above them). Declare `rules` (the shipped config declares persistence, momentum and mean reversion) before running; `k` is the lookback for momentum and mean reversion and `train_fraction` the share of history reserved before scoring begins. Minimal **format illustration**, not sufficient training data: ```csv timestamp,open,high,low,close 2010-01-01,100.0,101.7,99.0,100.7 2010-02-01,100.7,102.2,99.7,101.2 ``` ## Executable interface Exact CLI columns: `timestamp,open,high,low,close`. Supported method controls: `as_of, benford_column, frequency, horizon, k, rules, season, seed, train_fraction`. The tool scores predeclared one-step rules on closing prices at every origin after the training share: persistence (last close), momentum (last close plus the average change over `k` periods) and mean reversion (mean of the last `k` closes). It reports MAE per rule, the directional hit rate of