FOUR PROPRIETARY MODELS. THREE PUBLIC OVERLAYS. ONE DECISION ENGINE.
Every signal, score, and tier on this platform is the output of four proprietary models, each backtested and validated out-of-sample, and three public technical overlays re-implemented from source with full per-bar state persistence.
This page documents what each system does, how it is validated, and where the disclosure boundary sits. Proprietary models disclose outputs, validation windows, and realized track records; feature definitions and parameters are withheld under the standard quant-disclosure model. Public overlays are documented in full.
Audience: active traders, portfolio managers, allocators, and institutional research teams evaluating signal quality. All performance figures are from historical backtests and are not indicative of future results.
Options context (not options flow): screener IV ATM / IV-%ile / IV RNK and the chart ATM IV layer are archive-derived ATM call implied vol for CBOE weekly-options names — liquidity context, not a scored signal. See also the markdown methodology note on ATM IV.
FOUR PROPRIETARY MODELS
Intellectual property. Outputs and track records are disclosed. Feature definitions are not.
DIRECTIONAL SIGNAL
5-model composite plus Best Combo V5 stable evidence. Fires on confirmed momentum, not prediction.
The Directional Signal is a daily BUY/SELL classification computed independently for every ticker in the universe. It is not a single indicator. It is a composite of 5 independent sub-signals, each capturing a distinct edge. Its headline PF is a full-history closed-trade diagnostic. Best Combo V5 is a separate pre-registered stable-evidence state machine. It normalizes five signal families, smooths their weighted vote, and publishes only point-in-time out-of-fold decisions that pass the evidence and risk gates.
PIPELINE
5 independent sub-signals. trend, flip detection, volatility compression, momentum persistence, and microstructure. Each has its own parameter set and produces a binary BUY/SELL per bar.
Stable-evidence Best Combo V5. White Arrow, ATR, HiLo, Coil, and regime each contribute a bearish, neutral, or bullish vote. A short trailing EMA, broad-support checks, and two confirming closes drive entries and exits at the following session's open. Evidence determines the holding period; CRISIS is an immediate risk override, and evidence must reset before re-entry.
Evidence gate. Proven-combo screens and alerts require at least 20 out-of-fold trades, positive net expectancy and confidence, limited single-trade concentration, and daily drawdown ≤35% or ≤65% with at least 20% improvement over buy-and-hold. No validated edge, no published action.
Regime filter. The same signal is attenuated in CRISIS, tightened in BEAR. Regime context (METHOD·REG) modulates signal confidence before output.
VALIDATION
Every arrow carries a full-history per-stock profit-factor diagnostic. Best Combo V5 publishes stitched point-in-time out-of-fold evidence after an eight-year boundary, net of 20 bps round-trip. The former per-ticker K-of-N search is internal only and never substituted. Arrows are recomputed daily across 12,000+ symbols. Decayed edges are automatically down-weighted.
9,147
Universe
8 yr
Backtest
5
Sub-signals
≥ 1.5
Min PF gate
Proprietary: sub-signal feature definitions, inter-signal weights, support thresholds, smoothing parameters, and state-transition hysteresis.
THREE PUBLIC OVERLAYS
Well-known frameworks, faithfully re-implemented. Additive context for the proprietary models.
Triple Exhaustion detects overbought/oversold conditions using a three-indicator confluence gate combining MACD-Z, Stochastic, and DMI. All three indicators must hit extreme simultaneously for a bar to enter an exhaustion zone, with a 4-tier severity ladder quantifying how sustained the condition is. Per-bar z-scores are persisted and feed back into the directional signal pipeline.
CONFLUENCE GATE
All three conditions must hold simultaneously for bar t to enter an exhaustion zone. Seller exhaustion = oversold (watch for a down-move release). Buyer exhaustion = overbought (watch for an up-move release).
Once the confluence gate fires, the severity tier is determined by |MACD-Z| at peak. Tier colors match the chart overlay palette.
●
REGULAR|z| ≥ 2
First zone entry. Caution.
◉
SUPER|z| ≥ 3
Standard exhaustion. Counter-trend watch.
⊛
ULTRA|z| ≥ 5
Rare intensity. High-probability reversal zone.
✦
MEGA|z| ≥ 7
Extreme. Historically marks major inflection points.
RELEASE MARKERS
When the exhaustion zone ends. The confluence gate breaks. a one-shot release starburst fires on the chart:
✦BUY RELEASE
Fires when a SUPER seller-exhaustion zone (oversold) releases.
Interpretation: fade the down-move.
✦SELL RELEASE
Fires when a SUPER buyer-exhaustion zone (overbought) releases.
Interpretation: fade the up-move.
The arrow carries the peak |z| from the just-ended extreme run. not the trigger bar's z-on-release. This captures true severity. The public record measures each release over the following 21 sessions; it does not persist BUY or SELL until an opposite release. The EXH BUY tab also shows a descriptive stock-path table at bull-spread DTEs (leftover vs SPY ~0); that is context, not a forecast.
LOPJLB implementation. Indicator definitions (MACD, Stochastic, DMI) are public; the confluence gate, 4-tier severity ladder, z-threshold calibration, and integration with the directional signal pipeline are proprietary. Full per-bar z-score timeseries is persisted and available in the chart overlay.
OPERATING PRINCIPLES
Three constraints the entire stack respects.
EVIDENCE BEFORE NARRATIVE
No call is published on thematic conviction alone. Best Combo V5 uses point-in-time out-of-fold PF, expectancy, confidence, concentration, and relative drawdown gates. Unqualified evidence remains visible as research context, never as an action.
SIGNAL-LEVEL DIVERSIFICATION
Single-indicator strategies collapse at regime transitions. The composite blends independent sub-signals so aggregate Sharpe exceeds any individual sleeve. Diversification is applied at the signal level before it reaches the portfolio level.
TRANSPARENT OUTPUTS, PROPRIETARY INPUTS
Every published strategy, signal, and profit factor is accompanied by its full statistical record. The feature definitions and parameters that produce them are not disclosed. This mirrors the quant-ETF disclosure model.
Disclaimer. Exact formulas, parameter values, feature definitions, and signal weights for the four proprietary models are not disclosed. The three public overlays are documented in full above. All performance figures are from historical backtests and are not indicative of future results. LOPJLB publishes independent investment research and does not provide personalized investment advice.