Real-time prices, fundamentals, news sentiment, and SEC filings — ingested continuously and ranked across healthcare, technology, and oil & gas. Built for clarity, not noise.
Your account has been created. An administrator must approve access before the terminal becomes available. You will be admitted as soon as approval is granted.
MARKET DASHBOARDcomposite scores · live
UNIVERSE
—
tickers tracked
TOP COMPOSITE
—
—
BOTTOM COMPOSITE
—
—
UP TODAY
—
—
DOWN TODAY
—
—
LAST UPDATE
—
UTC
SECTOR ›
TICKERSfull ranked list
BACKTESTtop-N equal-weight hold · 1y
SECTOR ›TOP-N ›
PICKS
NOTE: This simulates equal-weighted hold of the CURRENT top-N over the past year — not periodic re-ranking with historical scores (which would require a longer score history we accumulate over time). Benchmark = S&P 500.
ML LABexperimental · LightGBM rankers · per-sector
▮ FLAGSHIP MODELS
Two flagship models:
V6 — within-sector LightGBM ranker;
V8 — universe ranker trained on (NYSE $500M+) ∪ (S&P 500), 2,172 tickers × 44 features including price/vol/risk-adjusted measures, macro signals, SEC filing activity (8-K, Form 4 insider transactions, 10-Q/K), and eight fundamental valuation features (earnings yield, sales yield, book yield, profit margin, revenue/EPS growth). V8 has fundamental visibility into overpriced names — expensive stocks with weak growth get systematically downweighted.
V7 is a LightGBM ranker trained on top-20%-by-universe-relative-return labels — it predicts which names will outperform the S&P 500 over the next 3 months. After audit: the IC signal is real (t = +2.09 ★ post-corrections), but the live $1M basket's return is dominated by factor exposure (β_MKT ≈ +1.10, β_SMB ≈ +0.96, β_CMA ≈ +1.14) rather than stock-selection alpha. Functions as a leveraged small-cap + low-investment factor tilt. β_MOM ≈ 0 — V7 is not momentum mimicry. For the full corrections cascade and FF regression details see F8 AUDIT.
V6 — SECTOR-RELATIVE RANKERflagship within-sector alpha model · two live $1M paper books
▮ §1 THE STATISTICAL CASE — V6'S HONEST EDGE
▮ §2 PER-SECTOR PERFORMANCE — WHERE V6 ACTUALLY WORKS
▮ §3 V6'S LIVE PICKS — TOP-3 BY ml_score_v6_pit PER SECTOR
▮ §4 TWO LIVE IMPLEMENTATIONS — LONG-ONLY vs SECTOR-NEUTRAL
▮ §5 LIVE FACTOR EXPOSURE — WHAT V6 IS ACTUALLY HOLDING
▮ §6 LONG-ONLY POSITIONS · LIVE P&L
▮ §7 SECTOR-NEUTRAL POSITIONS — THE IMPLEMENTATION THAT CAPTURES V6'S EDGE
V8 — PRODUCTION ALPHA MODEL.
Universe ranker trained on the (NYSE $500M+) ∪ (S&P 500) panel — 2,172 tickers, 44 features spanning price/vol metrics, macro signals, SEC filing activity, and fundamental valuation ratios.
Live as a $1M paper basket, vol-weighted (score / sqrt(vol_60), 8% cap, 2% floor) across 20 names with the full HF risk-control stack: portfolio vol target, HMM regime overlay, pairwise correlation gate, mean-reversion guard, earnings blackout, soft sector cap, and 12-month rolling drawdown killswitch.
ANALYTICSvisual deep-dive into every model
Visualizations of the flagship models running on the full S&P 500. HMM 5-state regime overlay (structural sizing/caps/exits) · V6 sector-relative picker (within-sector ranker) · V8 factor-orthogonal alpha (universe ranker). Monte Carlo simulations for both V6 and V8 portfolios shown here.
▮ MONTE CARLO — 10,000 SIMULATED 1-YEAR FUTURES FOR THE $1M PORTFOLIO
Fan chart of portfolio value over the next year. The dark center line is the median trajectory across 10,000 simulations; the bands are the 25/75th and 5/95th percentiles. Wider = more uncertainty. Generated via block-bootstrap from 5 years of daily returns — preserves real-world fat tails and vol clustering.
1-YEAR OUTCOME DISTRIBUTION
MAXIMUM DRAWDOWN DISTRIBUTION
Left: where each of the 10,000 simulations ended after one year. Right: the worst peak-to-trough drop along each path, regardless of where it finished. Drawdowns are what hurt emotionally — over half the simulations hit a 20%+ drawdown at some point, even with positive expected returns.
▮ V8 MONTE CARLO — 10,000 SIMULATED 1-YEAR FUTURES FOR THE NEW V8 PAPER BASKET
▮ V6 ROLLING 12-MONTH IC PER SECTOR — IS THE EDGE STILL ALIVE?
Trailing 12-month mean IC per sector — answers "what has the model been doing for me lately?" Lines above zero mean V6 is currently picking winners in that sector. Big swings up or down show genuine regime changes (not just averaging artifacts).