The AR(2) chip on the Fear & Greed forecast is a second-order autoregression: it predicts the next reading from the last two days instead of one. That extra lag lets the forecast carry short-term momentum — a falling index keeps falling for a few days before the pull back toward equilibrium takes over.
The OU baseline reverts smoothly and immediately toward its long-run mean. But real sentiment sell-offs tend to overshoot: fear feeds on fear for a stretch before it exhausts itself. AR(2) matches that by remembering the last two moves, so a sharp drop has measured downward drift built in for the next few sessions rather than snapping straight back. It is still a mean-reverting model — just one that acknowledges momentum on the way there.
AR(2) is a small, disciplined step up from OU: two lags instead of one, no exotic machinery. Whether that step is worth it on current data is a question you can answer directly — the "Model performance" table under the Fear & Greed forecast re-runs the walk-forward backtest and shows AR(2)'s coverage, MAPE, RMSE, Brier and directional hit-rate right beside the OU baseline and every other model.
OU reverts smoothly toward the mean from one day's reading. AR(2) looks at the last two days, so it can carry short-term momentum — a falling index keeps falling briefly before the pull toward equilibrium dominates.
Because sentiment sell-offs tend to overshoot rather than reverse instantly — fear feeds on fear for a stretch. The second lag lets the forecast reflect that persistence instead of snapping straight back to normal.
That depends on current data, and you don't have to guess: the 'Model performance' table under the Fear & Greed forecast shows AR(2) and OU side by side on coverage, MAPE, RMSE, Brier and hit-rate so you can compare them directly.
Browse all S&P 500 tickers to see this metric applied to individual companies.
Educational research only — not investment advice.