The Anatomy of an Over-Constrained Volatility Screener
Quantitative screeners frequently fall prey to condition compounding, where an analyst seeks pristine chart setups by appending increasingly tight volatility thresholds. In our test scan executed inside TC2000 EasyScan, a baseline of 3,842 common stocks trading above five dollars was subjected to four overlapping compression filters. The original intent was noble: isolate equities coiling into explosive volatility breakouts before momentum occurs.
The filtration sequence began with a classic squeeze requirement where 20-period standard deviation bands narrow inside 1.5 ATR channels. This single rule instantly eliminated 3,598 instruments, leaving 244 candidates. Next, an NR7 condition was introduced, requiring the latest daily range to sit smaller than any of the previous six sessions. This pruned the list down to 21 symbols. However, the designer subsequently added a multi-day True Range threshold under 1.5% and a Donchian channel width limit in the bottom decile. The final candidate output dropped to zero.
Why Coincidental Confluence Destroys Practical Candidate Lists
When four distinct formulas measure essentially the same underlying phenomenon—price range compression—they do not provide four independent confirmations. Instead, they act as an accidental multiplicative filter that penalizes even microscopic intraday wicks, discarding quality consolidation setups for trivial deviations:
- Overlapping statistical variables penalize single-session wick spikes while completely ignoring constructive basing structures that developed over weeks.
- Market-wide volatility expansions naturally shift benchmark percentiles higher, meaning static compression bounds guarantee zero candidates on active macro days.
- The surviving tickers in such hyper-narrow scans are frequently defunct special-purpose entities, halting acquisitions, or stalled micro-caps with artificial price quotes.
Refactoring Framework: The Two-Tiered Volatility Funnel
To fix an over-optimized squeeze pipeline, analysts must decouple collinear indicators. Rather than demanding absolute Bollinger contraction AND Donchian constriction AND NR7 AND True Range minimization on the exact same closing tick, construct a two-tiered evaluation. Use the broad Bollinger-Keltner relationship as a baseline gate to surface 150 to 200 candidates, then score or sort candidates by average true range percentile instead of hard-filtering them out. This preserves an actionable universe for manual chart review while completely eliminating empty screener runs.