The Illusion of Upward Momentum in Raw Relational Inequalities
Novice quantitative screeners routinely begin with a simple expression: Close greater than the 200-day simple moving average. While the intention is to restrict the universe strictly to equities in secular uptrends, the practical outcome is an unmanageable list of several thousand stocks. The condition lacks boundary limits, treating a stock 0.1% above an exhausted baseline identically to an overextended stock 85% above it. Without upper caps or proximity criteria, the scan functions merely as an arbitrary partition rather than an analytical filter.
Core Structural Takeaway
An open-ended inequality filters direction but ignores magnitude and maturity. Effective condition design always pairs directional positioning with a bounded dispersion envelope or an active rate-of-change constraint to avoid sweeping in late-stage blow-off tops alongside dead-cat bounces.
Dispersion Without Closeness: Why Raw Inequalities Overwhelm Worklists
Screening pipelines are designed to reduce large instrument universes into homogeneous subsets that can be systematically audited. When an analyst runs an unbounded comparison like C > AVGC200 on a broad equity database of 6,000 instruments during an ordinary expansion cycle, between 70% and 82% of all tickers pass effortlessly. The resulting worklist retains virtually the entire market breadth, forcing secondary filters to shoulder the entire computational burden.
More dangerously, the distribution of surviving candidates contains diametrically opposed price structures. At one extreme lie parabolic leaders that have stretched three standard deviations above their benchmark curve and face immediate mean-reversion risk. At the opposite extreme sit deteriorating cyclical names that momentarily ticked two cents above a declining moving average before resuming a broader downtrend. By treating all positive differences as equivalent, the scan introduces profound structural heterogeneity into downstream review stages.
Overextended Outliers and Decaying Baselines
Because the single inequality registers true across any positive spread, it cannot differentiate between healthy trend consolidations, explosive climax runs, and downward-sloping averages experiencing brief price punctures.
Filter Metric Profile
- Condition Tested
- C > AVGC200
- US Equities Baseline
- 6,420 Symbols
- Post-Scan Output
- 4,890 Symbols (76%)
- Diagnostic Verdict
- Functionally Unbounded
Constructing Bounded Bands and Relative Distance Metrics
To transform an unhelpful binary check into an informative condition, you must bind the price distance within a defined percentage channel or normalize it through normalized volatility metrics. For instance, requiring price to remain between 1.01 and 1.12 times the 200-day moving average ensures that candidates occupy a verifiable continuation phase rather than an extended blow-off state.
Furthermore, the slope of the reference line itself must be incorporated into the logic stack. Testing that the moving average today exceeds its level twenty bars ago (AVGC200 > AVGC200.20) guarantees that the structural baseline is actively rising, discarding declining curves that produce misleading bullish crosses during counter-trend rallies.
Bounded Stacking Rule
Never introduce a moving average filter without defining both an upper proximity boundary and a slope verification statement. Combining (C > AVGC200) with (C < AVGC200 * 1.15) and (AVGC200> AVGC200.10) systematically eliminates over 65% of spurious bull-market noise before secondary indicators run.