Macro Market Environment

Loading macro indicators from St. Louis Fed (FRED)...

Market Environment Backtest

Checks whether the Market Environment risk score has actually been useful historically, instead of just looking plausible. This reconstructs the score at every month in the selected date range using real FRED macro data and SPY price history, then compares each month's score against SPY's subsequent 3/6/12-month performance. A well-calibrated score should show negative correlations (higher risk score → lower future return) and "Critical"/"Elevated" months averaging worse forward performance than "Low" months.

"Since 1999" is the recommended default, not the earliest technically possible date: it's the point by which every factor already has real FRED data, since Fed Funds Pressure (EFFR) only starts in July 2000 — an earlier custom start date means that one factor falls back to a fixed baseline constant instead of a real value for the months before its data begins. Custom start dates are still accepted back to 1990-01-01 if you want a longer (but less complete) history anyway. Earlier start dates also fetch more history and take longer the first time (up to ~30 seconds); results for a given period are then cached for a day. Start dates must be at least 13 months before today. Leaving the end date blank scores every month through the most recent one available; setting it only limits which months get a score — the forward return/drawdown for the last few scored months can still look at real price data after that date, since that's necessary to see what actually happened next.

How to read these results
Forward horizon (the 3 panels below)
Each panel repeats the exact same analysis, just changing how far into the future it looks before checking what actually happened. For example, if a month's risk score was reconstructed for January 2020, the "3-Month" panel records SPY's return from January 2020 to April 2020; the "6-Month" panel records it through July 2020; the "12-Month" panel through January 2021. Every scored month is reused across all three panels — only the length of the forward-looking window changes. Comparing all three shows whether the score's predictive value holds up, strengthens, or fades the further out you look.
Risk tranche (Low / Elevated / Critical)
The same Low (score < 50), Elevated (score ≥ 50), and Critical (score ≥ 75) bands used by the live Market Environment gauge, applied retroactively to each historical month's reconstructed score.
Avg forward return / max drawdown
For every month that fell into a tranche, this averages SPY's subsequent total return (and worst peak-to-trough decline) over that panel's horizon. If the score is useful, "Critical" and "Elevated" months should show worse (lower/more negative) averages than "Low" months.
Correlation with forward return / drawdown
Ranges from -1 (perfectly predicts lower returns as the score rises) to +1 (perfectly predicts the opposite). Values near 0 mean no real relationship. A genuinely useful risk score should be negative and ideally get more negative at longer horizons.
Sample size
The number of historical months with enough data to include. Small tranche counts (especially for "Critical", which is rare) mean that tranche's average is less statistically reliable.

Stock-analysis model audit

Model Insight Report

This report explains whether the system's Bullish/Bearish Score behaves sensibly for a ticker: are scores compressed or extreme, which regimes appear most often, which factors dominate, and how complete the underlying data is.

What it shows Score distribution, regime frequency, factor influence, and data coverage.
Why it matters It makes the model auditable instead of asking users to trust a black-box score.
What it is not It is not a prediction, probability, trading signal, or forward-return backtest.

Enter a ticker and run the report.