QuantKernel 6.2 is a self-adaptive, self-healing quantitative research engine — the system detects its own performance degradation and auto-corrects without human intervention. ML health monitoring triggers conditional retraining when win-rate degrades. A supply-chain knowledge graph runs pre-market at 09:05 ET. A degenerate watchdog quarantines underperforming strategy variants automatically.
Six research strategies span equities, ETFs, options, a growth-compounder sleeve, and macro ETF rotation. Each morning at 09:20 ET, subscribers receive a directional watchlist — algorithmic output with reference levels and risk context. You decide if, when, and how to act. Not personalised advice.
Four equity strategies span momentum, intraday breakout, and market-neutral arbitrage, plus a physically isolated growth-compounder sleeve powered by survivorship-free fundamental screening. A fifth options strategy layers systematic premium income analysis. A sixth macro rotation strategy ranks sector and factor ETFs weekly by relative strength. Each strategy publishes its watchlist to a dedicated Discord channel.
Rebuilt on the DB2 Adaptive Donchian Channel engine — lookback N is dynamically calibrated to the σ5/σ30 volatility ratio, so the channel widens in choppy markets and tightens in trending ones. The engine flags potential setups when price approaches adaptive channel boundaries. The same adaptive channel logic covers major index ETFs as well as individual equities — QQQ and SPY macro trends are analysed with identical precision.
FinBERT NLP sentiment scoring screens each candidate: strong bearish conviction removes a ticker from the watchlist regardless of technical alignment. Each publication includes ATR-derived reference risk levels (1.5×) for subscriber context.
Two classic institutional intraday algorithms running in parallel. Dual Thrust computes an asymmetric range (Max(HH−LC, HC−LL)) from the prior N bars and enters on a directional break above or below it. R-Breaker runs a 6-level dual-mode state machine — breakout mode when price pierces the buy/sell break levels, reversal mode when intraday extremes trigger the setup/reverse sequence.
Both engines share the same FinBERT sentiment gate and hard 15:50 ET close — zero overnight exposure, zero gap risk.
Identifies statistically correlated equity pairs (e.g. NVDA/AMD, JPM/BAC) that have temporarily diverged beyond their historical Z-score band. Goes long the underperformer, short the outperformer.
Exits when Z-score reverts to within 0.5. Market-neutral by design — the pair hedges gross market direction.
A physically isolated growth-compounder sleeve powered by survivorship-free Point-In-Time (PIT) data — the same institutional-grade data standard used by systematic hedge funds. The three-gate PITFilterPipeline screens for fundamental acceleration (positive 2nd-derivative of YoY revenue growth + operating-margin inflection), RS percentile top 10%, and a Minervini-style VCP (Volatility Contraction Pattern) with quantified volume dry-up.
Architecture guarantee: the sleeve is completely isolated from all other strategies — no shared capital, no shared risk state, no shared signal path. The only connection to the QuantKernel core is a read-only RegimeGateway. In ATTACK/NEUTRAL regimes, the watchlist scope expands to include new growth candidates. In HEDGE/ICE_POINT regimes, existing high-conviction names remain on the watchlist — the regime filter does not force removal of candidates mid-trend.
Ranks 20 liquid sector and factor ETFs (XLK, XLF, XLE, XLV, XLI, XLY, XLP, XLRE, XLU, XLB, XLC, QQQ, IWM, GLD, TLT and more) by a composite 1-month and 3-month relative strength score. Each week, the top 3 ETFs by RS rank enter the watchlist as long-bias candidates; the bottom quartile is flagged as avoidance. Simple, rule-driven rotation with no discretion.
FinBERT macro sentiment and VIX regime gate run before each weekly publication: risk-off environments (VIX > 28 or macro sentiment negative) shift the entire watchlist to defensive sectors (XLP, XLU, GLD). Published every Sunday evening before the Monday open.
Beyond each strategy's own risk reference level, a portfolio-level risk filter evaluates independently before each publication. In v6.2, the engine also monitors itself: a DegenerateWatchdog detects 3σ signal-quality decay over two consecutive days and quarantines the affected strategy for five trading days — no candidates from a quarantined strategy appear in the watchlist. A strategy state machine (PRODUCING → DEGRADED → QUARANTINE) ensures degrading variants never silently continue publishing. The engine was designed from day one with a beta-neutral philosophy: liquidity-anchored ETFs dynamically monitor sector-level macro conditions, filtering out candidates exposed to market-wide systematic drawdown risk while preserving pure idiosyncratic setups.
The system continuously calibrates position sizing based on the VIX volatility matrix — replacing rigid on/off switches with fluid exposure scaling. When market tail-risk spikes, the system dynamically scales down exposure or halts new entries entirely. Preserving capital during chaos is what enables faster compounding during normalcy. This is the core reason historical max drawdown stays within −8.3% despite volatile market conditions.
Pre-market news sentiment scoring monitors macro events, Fed announcements, and earnings volatility using advanced linguistic processing. If sentiment aggregates below a rigid negative threshold on a watchlist candidate, that ticker is flagged and a risk note is included in that day's publication.
Position sizing is computed automatically so that the maximum risk on any single transaction never exceeds 2% of total portfolio capital. This mathematical allocation rule applies uniformly across all four trading strategies with no subjective exceptions.
Each strategy operates within its own isolated capital sleeve. An adverse drawdown in the Intraday Breakout Engine does not affect Trend Tracker AI sizing. Losses are strictly contained within pre-defined strategy boundaries before they can affect the broader portfolio capital.
Before any candidate reaches the morning watchlist, the engine classifies current market regime via SPY EMA9/EMA50 trend alignment and ADX momentum confirmation — outputting one of four states: ATTACK, NEUTRAL, HEDGE, or ICE_POINT. In HEDGE and ICE_POINT regimes, new growth candidates are automatically suppressed and the publication scope contracts to only the highest-conviction existing names. No candidate from a regime-misaligned strategy appears in the watchlist, regardless of how strong its individual signal scores.
Upside and downside reference levels in each watchlist publication are derived from the Micro Order Penetration Grid — not generic ATR multiples. The engine calculates 10 dynamic levels from Standard Pivot Points (P, R1–R3, S1–S3, prior day open/high/low/close). Upside targets are anchored just below the nearest resistance cluster; downside references just above the nearest support cluster — placing reference levels where institutional depth naturally concentrates.
Every publication is permanently archived in Discord history — fully transparent and open to subscriber audit at any time. Reference levels are algorithmic outputs; actual results depend on each subscriber's own execution.
The entire onboarding flow is automated. From payment to your first morning publication, no manual steps required on our end.
Click the subscribe button and complete payment through Whop's secure checkout. Supports global credit cards and Apple Pay — fully encrypted end to end.
Whop automatically links your Discord account and grants you the VIP Subscriber role within seconds — no email, no invite link, no manual steps.
Each morning at 09:20 ET your daily watchlist is published to your channel — #VIP-Alpha for equity research, #VIP-Theta for options analysis. Review the directional candidates, reference levels, and risk context — then decide independently if, when, and how to act.
Each publication is clearly labeled by strategy — you always know which system flagged the setup, what the thesis is, and the reference risk levels. Directional bias only. You decide how to act.
No analyst opinions. No hindsight calls. Every Discovery watchlist and Radar risk flag is generated by QuantKernel's research engine — pure, objective, rule-driven data, published uniformly to all subscribers at 09:20 ET each morning before the open.
System updates, market structure observations, and signal logic breakdowns — written by the team building QuantKernel.
⚠ Risk Disclosure & Legal Disclaimer: All content, signals, data, and charts provided on this website are generated by automated Python-based quantitative research systems and are intended solely for educational and technical research purposes. Nothing on this site constitutes securities investment advice, a financial recommendation, or any form of managed account service. We accept no liability for any trading results incurred by subscribers. Stock trading involves significant risk; you may lose your entire principal. Past performance — whether from backtesting or live trading — does not guarantee future results. Any trades placed by subscribers based on signals distributed through this service represent the subscriber's own independent investment decisions. All execution risk, slippage, losses, and legal consequences arising from such decisions are borne solely by the subscriber. This service does not involve discretionary management or custody of client funds.
(I) All signals provided by QuantKernel are the objective output of a fixed quantitative rule engine operating on probabilistic strategy distributions. They constitute strategy probability-distribution indicators only and do NOT constitute investment advice of any kind (Not Financial Advice / NFA). (II) Options and equity trading carry extreme risk. Due to network latency, broker friction costs, and slippage caused by extreme market volatility, actual fill prices may differ materially from system simulation prices. Users bear absolute and sole responsibility for all position sizing, risk management, and account exposure. QuantKernel accepts no liability whatsoever for losses arising from execution deviation between signal price and market fill price.