An analog is a prior market window that resembles the current setup. Its value is the comparison set, not a single dramatic chart.
A useful analog starts with an explicit lookback window. Price shape may be compared through normalized returns, turning points, volatility, drawdown, and trend. Macro or cross-asset context can add a second layer, but it should never conceal a weak price match. The result is a ranked set of prior windows, each with a visible similarity score and exact dates.
The median path is only a summary. Two analogs can rise sharply while two fall and still produce a smooth line that no historical instance actually followed. Read the individual paths, median outcome, dispersion, worst drawdown, agreement, and sample size together. A wide range is information: it says the historical record did not resolve the setup consistently.
Adjacent dates from the same episode are not independent evidence. A sound engine settles overlapping candidates into distinct occurrences and fixes every forward window after the match date. Thresholds and ranking rules must be defined before outcomes are inspected; otherwise an attractive result may be the product of repeated searching.
Choose a lookback that matches the question, inspect the actual dates, compare the forward result with the ticker's unconditional history, and test nearby lookbacks. If the conclusion disappears after a small window change, treat it as fragile context rather than a durable read.