| Factor | Indicator | Value | Z-Score | Pctl | Dir |
|---|
I can see all your economic indicators. Ask me anything — "what does this mean?", "should I be worried?", "explain it like I'm not a finance person" — I'll walk you through it.
Get push notifications via ntfy.sh when new stocks match your criteria. No account needed.
The Economy tab gives you regime, factors, and divergences. This tab is the plan for turning those signals into a disciplined portfolio overlay — without ever writing a fully-autonomous trading bot. The hard work (classification, z-scores, divergence detection) is already done. What's left is translation, execution, and honest backtesting.
Map each regime to a target portfolio, not just a "winners/losers" tag. Add hysteresis: require 5 consecutive days above 60% confidence before flipping — avoids whipsaws on a 52→48% flicker.
Alpaca or IBKR — both free, both have TS/Python SDKs. Don't go autonomous on day one.
Current regime + target allocation from step 01.
Compute drift per asset class.
Generate trade tickets to a local file.
Executes via broker API.
Only flip to auto-execute after 3+ months where you agreed with its tickets ≥90% of the time.
Your divergence detector is the most under-priced thing in this system. No one else aggregates "cardboard production vs retail sales." Treat divergences as tilts on top of regime allocation, not regime calls themselves.
Log every divergence-triggered tilt with entry, exit, and PnL. After a year of data you'll know which divergences actually predict anything — and which are noise.
FRED data goes back decades. Before risking real capital:
% of all ~60 metrics currently in "extreme" territory (|z| > 2). Above 30% = regime stress — shrink position sizes regardless of what the regime says. Extremes cluster right before regime changes.
Require 2 of 5 factors to agree before flipping the regime call. Single-factor regime shifts are usually noise. Composite agreement is signal.