Methodology Changelog
Every scoring revision from R1 through R29, covering signal additions, removals, weight changes, and threshold adjustments. Naming fixes and pipeline plumbing are excluded. See how the scoring system works and current results.
- First, foreign-listed companies that file annual reports under international accounting standards now receive full fundamental scoring. Previously these tickers returned no financial data and were capped well below 🏆 SPOTLIGHT or ✅ OPPORTUNITY thresholds regardless of business quality.
- Second, parallel backtests on a cold cache no longer silently miss fundamental data: a per-ticker lock prevents duplicate EDGAR fetches across workers, and per-worker rate pacing keeps total request volume within SEC limits. A same-session re-fetch bug that caused certain tickers to be re-queried on every backtest date was also fixed.
- Third, a fresh genetic algorithm run was performed with both core fundamental quality weights released from their previous fixed values. The GA redistributed weight between them, adjusted other weights, and tweaked the 🏆 SPOTLIGHT and ✅ OPPORTUNITY criteria.
Added a staleness guard to the TTM (Trailing Twelve Months) net income calculation. Companies with multi-year SEC EDGAR filing gaps could previously and incorrectly trigger our strong ROE (Return on Equity) signal using stale quarterly data (up to 35 months). The guard rejects any TTM window whose most recent quarter is more than 548 days old, relative to the screening date. A small number of historical signals were corrected as a result. No weights, signals, or thresholds changed.
Alpha vs SPY: +130.6%Full backtest rerun confirmed identical results to R26: +163.2% alpha vs SPY, same weights and thresholds throughout. VWAP signals renamed to VWMA across pipeline and documentation.
Alpha vs SPY: +163.2%First fully automated weight search: a genetic algorithm optimised all signal weights simultaneously across 14 years of backtested data. The system diversified away from reliance on a single dominant entry signal, redistributing toward a broader set of smaller signals. Quality scores for trend-confirmation signals increased; two low-value signals zeroed. Tier thresholds also adjusted by the GA: 🏆 SPOTLIGHT combined and entry minimums loosened slightly; ✅ OPPORTUNITY quality minimum tightened by one point.
Alpha vs SPY: +163.2%Tested adding relative-strength momentum as a scored quality signal. Daily alpha dropped significantly; revision rejected and reverted to the R24 baseline. Published here because not every round improves things.
Alpha vs SPY: +46.2%A previously zeroed confluence signal removed from scoring entirely; its freed points redistributed to earnings growth. Minor refinement building on R23's signal removals.
Alpha vs SPY: +93.8%Two signals with unanimous zero-predictive-value findings across three independent validation methods removed from scoring (detection kept for use in confluence calculations). A previously tested confluence signal confirmed at zero after producing no consistent return improvement across test rounds.
Alpha vs SPY: +87.2%Six rounds of threshold refinement on top of the R16 signal configuration. ✅ OPPORTUNITY quality minimum adjusted upward; 🏆 SPOTLIGHT quality minimum raised; 📌 MONITOR floor reduced. Signal weights unchanged across all six rounds.
Two entry signals eliminated after statistical validation showed near-zero predictive value and high redundancy. Freed scoring weight redistributed to volume breakout detection and a multi-indicator confluence. A quality signal also zeroed; its points absorbed into a smooth-momentum composite.
Alpha vs SPY: +78.7%The SMA20/50 crossover signal elevated to primary entry anchor after walk-forward validation showed 100% positive periods. A formal feature-selection framework introduced, separating signal relevance analysis from weight optimisation. Two signals with near-zero predictive value reduced.
✅ OPPORTUNITY tier minimum entry threshold raised after observing that some signals reaching the tier had insufficient entry timing confirmation. Thresholds only; no signal weight changes.
Corrected a scoring engine bug where certain entry signals were not accumulating correctly in backtests (evaluated as zero across all prior rounds R1-R11). First round with reliable historical results. The SMA20/50 crossover and a multi-indicator confluence emerged as the strongest entry signals once scores were computed correctly.
ROE strength and debt/equity ratio designated as permanent, load-bearing quality anchors. These two signals account for the majority of the quality score ceiling and are never zeroed in subsequent rounds regardless of statistical test results.
Alpha vs SPY: +73%Initial 5-stage pipeline built and validated: price and volume filters (Stage 1), trend filters using Minervini and Weinstein criteria (Stage 2), 100-point scoring system (Stage 3), sentiment analysis (Stage 4), paper trading (Stage 5). Baseline alpha approximately +41% before correct backtest verification at R12.
This changelog covers scoring methodology only: signals, weights, and signal tier thresholds. Infrastructure, naming, and web changes are not listed.
Get Daily Advanced Signals Free
Actionable and freshly screened US stocks in your inbox every trading day.