The Specific Signal
We track every prediction our system makes against actual outcomes. One specific setup has produced an 85.2% win rate over 16,000+ predictions: bearish pattern signals on stocks in the bottom 40% of their own volatility history.
This isn't cherry-picking. It's a consistent pattern across hundreds of real predictions. Here's how to identify it and why it works.
What 'Bearish Pattern + Low Vol' Means
Two conditions have to line up:
- The pattern match: our similarity search finds a stock's recent chart matches 10 historical charts, and at least 7 of those 10 went DOWN over the next 5 days (we call this a bearish up_count_5d ≤ 3)
- The regime: the stock's own 20-day realized volatility sits in the 0-40th percentile of its trailing 252-day history — i.e., this is a stock that normally doesn't move much
When both are true, historically 85.2% of the time the actual 5-day return was negative. Sample size: 454 setups out of 16,438 total predictions.
Why It Works
Low-volatility stocks are usually large-cap names with established businesses, stable cash flows, and institutional holders. When a stock like that starts forming a pattern that looks like historical breakdowns, it's often because fundamentals or narrative have broken down — and institutional selling tends to be methodical, not noisy.
In high-volatility names, by contrast, bearish chart signals are often followed by short-squeeze reversals, news-driven pops, or retail meme-stock behavior. The signal-to-noise ratio is much worse.
Academic work supports this: Daniel & Moskowitz (2016) showed momentum strategies have their worst performance in panic regimes with high cross-sectional volatility. Our finding is the inverse of that — bearish pattern momentum works best in CALM regimes on CALM names.
What It Doesn't Mean
A few important caveats before you go shorting low-vol stocks:
- Win rate ≠ profit. The 85% win rate is for direction. Magnitude matters — some of those 15% losers were sharp reversals.
- The dataset covers a bearish period (March-April 2026). In a sustained bull market the base rate flips and this edge likely compresses.
- Short-selling has borrow costs, hard-to-borrow lists, and asymmetric risk. Even a 'high probability' short is not a free lunch.
- This is one specific signal from one dataset, not a blanket trading strategy.
How to See This Yourself
The /api/v1/accuracy/by-regime endpoint on Chart Library returns accuracy bucketed by any of 12 context dimensions including trailing_vol_pct. Specific query: dimension=trailing_vol_pct, horizon=5. You'll see the 85.2% bearish win rate in the Q1 bucket.
Or use the regime filter toggle on any search at chartlibrary.io/app — switch to 'Regime-matched only' and the stats recalculate.