RARE-SIGNAL SHORT-TERM TIMING
Two extremes that history says actually work.
Most technical "sell" signals fail on the S&P 500 because the market rises 58–70% of any given day. Most "buy" signals only produce a marginal edge. But we tested dozens of them against 36 years of data and found two rare setups — one at bottoms, one at tops — that have historically produced returns meaningfully different from a random day. This page shows them, honestly.
Where we stand right now
Two independent signals. Either can be active; usually neither is.
Panic Buy Loading…
– 5-day returnFires when the S&P has dropped 7% or more over the last 5 trading days. The intuition: that's a proper flush. Fear is elevated, forced selling is done, and buyers with cash are stepping in. Historically these episodes were followed by +3.8% avg over the next 20 days and +8.6% avg over the next 60 days. Fires ~2× per year on average.
Peak Warning Loading…
– 14-day RSIFires when the 14-day RSI drops back below 80 after being above 80. The RSI has to actually get to 80+ first (which means euphoric momentum), and then start to fade. Historically these were followed by −0.9% avg over the next 10 days (only 35% positive), and −0.4% avg over 20 days (41% positive) — the only "sell" signal we tested that produced genuinely negative expected returns. Very rare: ~once every 2 years.
S&P 500 with every historical signal
Green dots mark Panic Buy triggers, red dots mark Peak Warning triggers. Notice how sparse both are — that's the feature, not a bug. If a signal fired every week, it wouldn't mean anything.
Most recent Panic Buy signals
Most recent Peak Warnings
How they performed — historical backtest since 1990
For every day one of these signals fired, we recorded the S&P's forward returns 5, 10, 20, and 60 trading days later. Compared against baseline (the return distribution of any random trading day) so the edge is visible.
Three things to understand before you use these
Rarity is the whole point.
Panic Buys fire roughly 2× a year. Peak Warnings fire roughly once every two years. If you want a "what should I do this week" signal, this page won't give you one most of the time — it'll say "nothing." That is the correct answer 95%+ of the time.
The Panic Buy is the stronger of the two.
+3.8% average forward 20-day return (vs +0.77% baseline). That's about 5× the baseline expected return, at 73% positive. When the market drops 7% in a week, the odds historically shift very sharply in your favor. This is the classic "buy when there's blood in the streets" pattern — quantified.
The Peak Warning is fast-fading.
At 5-20 days the signal produces negative expected returns. At 60 days it's positive again (+2.6%) — the upward drift reasserts itself. This is not "sell everything and go to cash." It's a 2-4 week caution flag saying momentum has peaked and short-term returns are likely below average.
How to actually read this page
- Default state: neither signal is on. That's most of the time. It means the market isn't in an extreme setup — use the main BATS gauge and the Momentum page instead.
- When Panic Buy is on: the setup is a "buy the flush" signal with a 4-8 week horizon. History says the odds favor buyers strongly. But every fire is different — the 2008 GFC and the COVID crash both fired multiple times as the market kept falling before ultimately bottoming.
- When Peak Warning is on: think "trim, not exit." The signal warns of below-average forward returns for 2-4 weeks, not a crash. Historically the market often drifts sideways-to-down after this signal before resuming its rise.
- These two signals do not conflict. They measure different extremes. Both can be off (most of the time). Both can be on in the same year at different moments. They cannot both be on the same day.
Reminder: None of this is investment advice. Historical edge is not a guarantee. Even the Panic Buy signal fired repeatedly during the 2008 crisis while the market continued lower — the eventual rally didn't come until March 2009. Always do your own research or talk to a financial professional.
Methodology
Data: S&P 500 daily closes since 1990, from our
own data/spx.csv (refreshed daily via GitHub Actions).
All math is done live in the browser — no server-side pre-computation.
Panic Buy trigger: today's close is more than 7% below
the close 5 trading days ago. Formula:
(close_today / close_5_days_ago) - 1 < -0.07.
Fires on the day the threshold is crossed; can fire on consecutive days
if the market keeps falling.
Peak Warning trigger: yesterday's 14-day RSI (Wilder smoothing) was above 80; today's RSI is at or below 80. That means the index had to reach a genuinely extreme overbought state first, then start to lose steam.
Backtest baseline: we compare each signal's forward returns to the return distribution of every single trading day in the dataset. Baseline: +0.19% at 5d, +0.77% at 20d, +2.31% at 60d; 57.8% / 63.6% / 70.2% positive. The market rises far more often than it falls, which is why "sell" signals are so hard to build.
Why not more signals? We tested 11 popular technical setups (RSI crossovers at 20/80, Bollinger band violations, stochastic overbought/oversold crossovers, VIX regime shifts, blowoff-top rules, MACD zero-line crossovers, and more). All the others produced returns statistically indistinguishable from the baseline. These are the two that survived.