S&P 500 Outlook

This page aligns US growth, rates, labour, sentiment, and commodity data and shows both one-week-ahead and one-month-ahead outlooks. Return magnitude comes from a regression ensemble, while direction comes from direct classification.

Neutral Current regime: Global Updated: 2026-04-30 19:47
Forecast horizon
Target month 2026-05-01
Feature month 2026-04-24
Expected return +0.68%
Expected band -1.76% - +3.13%
Probability of gains 68.8%
Directional hit rate 65.0%
Probability of downside 1.8%
Downside threshold Worse than -3% over the next 5 trading days
Backtest observations 60
MAE 1.69%
RMSE 2.45%

Equity curve if the 1W signal were traded

Comparison assumptions: the threshold variant skips weeks when the absolute forecast return is below 1.00%, and the probability filter uses 60% / 40% cutoffs.

Long + short Go long for the next forecast horizon when the return forecast is positive, and short when it is negative.
Strategy cumulative return -
Vs. buy & hold -
Buy & hold cumulative return -
Sample span -
Strategy annualized return -
Buy & hold annualized -
Strategy max drawdown -
Hit rate -
The top panel shows cumulative equity curves for each strategy; the lower panel shows weekly returns for the selected strategy against buy & hold. Both curves start at 100.

Performance by strategy

Strategy Cumulative return Annualized Max drawdown Hit rate Active entries Trades

Upside drivers

The indicators that pushed the current return forecast higher.

S&P 500 3M volatility +0.20pt
World Gold +0.17pt
World Brent Spot Price +0.07pt
VVIX to VIX ratio +0.06pt
US real rate proxy +0.06pt

Downside drivers

The indicators that pushed the current return forecast lower.

S&P 500 12M drawdown -0.16pt
World WTI Spot Price -0.16pt
S&P 500 Index -0.08pt
10-Year Treasury Yield -0.04pt

Model setup

This is a statistical read on macro and market conditions, not a direct trading signal.

  • The target is the next one-week S&P 500 return.
  • The weekly view is aligned to Friday closes and the monthly view to month-end closes. Monthly and quarterly indicators are lagged to reflect release timing before they enter the feature set.
  • Each run retrains the regression ensemble for return size and the direct classifiers for direction using 36 engineered features built from 16 input series.
  • The one-week and one-month models use different training windows and feature mixes. The one-month model also layers KMeans-based regime experts on top of the global model. Return magnitude blends tree 70%, linear 20%, and sign-conditioned return 10%; up-probability blends the tree classifier 60% and linear classifier 40%.