S&P 500 analogs: the historical similarity result needs its register

A dated correction preserves a source-scoped closing-index record while withdrawing unsupported analog selection, forward-return and forecast claims.

What changed in this article

Editorial correction · 9 September 2026. The original article was published on 14 February 2026 and described a 60-trading-day S&P 500 pattern ending on 13 February, a search of historical windows since 1950, the top 20 analogs and forward return distributions. It did not preserve the complete historical series, data vintage, candidate-window list, similarity output, non-overlap implementation, out-of-sample separation or calculations. The stated +4.53% versus +2.52%, 75% versus 66.2%, episode table, drawdown distribution and “more likely whipsaw” interpretation are withdrawn. The original URL and publication date remain unchanged.

Historical similarity can describe a conditional sample. It cannot become a forecast certainty or a substitute for a defined out-of-sample test.

What the dated index record establishes

Source date and record	Dated fact	Boundary
13 February 2026 · S&P 500 series at FRED	The S&P 500 series is sourced to S&P Dow Jones Indices through FRED and is defined as a daily closing price index, not a total-return series. The retrieved observation for 13 February 2026 is 6,836.17.	This identifies an index, date and price convention. It does not reproduce the article’s 60-day normalized path, historical analog selection or forward outcomes.
31 December 2025 and 13 February 2026 · same series	The retrieved daily closing observations are 6,845.50 and 6,836.17, respectively.	The series excludes dividends. Any comparison with a total-return benchmark or investable product would require a separate definition.

The FRED record supplies a reproducible index boundary for a future exercise. It does not establish which historical episodes are similar or what will happen next.

What is withdrawn

The 60-observation pattern score, since-1950 candidate universe, top-20 selection, similarity scores, forward 30/60/90-day returns, baseline distribution, win rates, drawdown statistics and the whipsaw-versus-collapse interpretation are not verified findings. No replacement forecast, scenario level or trading signal is published.

The method prompt—freeze the data vintage and selection rule before observing forward outcomes—is retained as an audit requirement, not as a prediction.

Minimum register for a future analog study

Record the exact index or instrument, source and vintage; target-window dates and observations; historical candidate universe; normalization and similarity formula; overlap and separation rules; selection cutoff; forward horizons; out-of-sample boundary; missing-data and holiday handling; return type and dividend treatment; and the complete selected-window and calculation files.

Original publication: 14 February 2026. Correction: 9 September 2026. AI-assisted source checking and writing; no independent human expert review. This page is educational market research, not personalized investment advice.