# DataQuant

> DataQuant is a stock screener that reads every liquid US stock four ways: technicals (graded bases, relative strength, pivots), fundamentals (the latest earnings call, read by AI), popularity (institutional activity, named 13F holders, social sentiment and news) and a plain-language AI write-up on top. It covers about 3,300 US stocks and ETFs on NASDAQ, NYSE and AMEX above $300M market value, plus Indian NSE stocks, screened every 15 minutes in market hours and again after the close. It reports what the screen saw; it is a research tool, not investment advice.

## The four readings

- **Technicals.** When a close clears the pivot of a base, the base is graded S, A+ or A by its length, its depth (25% or less) and whether it tops out at the 52-week high (blue sky). The breakout is scored 1 to 10 from factors that measured as a profit-factor lift over 97,789 historical breakouts: relative-strength bucket, base grade, institutional activity, how far and on how much volume the close cleared the pivot, an episodic-pivot flag, 13F register turnover, and a penalty past 5% extension. The dashboard shows 7 and above by default. The fail level sits 7% below the pivot. A separate deep-base kind (25-35% deep, blue sky, cleared by 3%+ on 1.5x volume) is graded on its own terms.
- **Relative strength (RS).** Trailing returns (50% 3-month, 30% 1-month, 20% 1-week) ranked 1-99 against the stock's own market. Breakouts with RS under 89 ran a 1.73 profit factor, 89-95 ran 2.04, 95 and above ran 2.45.
- **Fundamentals.** For every US stock on the screen, a language model reads the latest earnings call transcript and returns a tone score from -1 to +1, guidance direction (raised, maintained, lowered), highlights in management's words and risk flags. The Winners and Beat & Raise lists add growth, margins and earnings-surprise history.
- **Popularity.** An institutional activity score from 0 to 10 (heavy-volume up and down days, prints at the high, up/down volume): at 7 or more, 32% of breakouts reached +20%, against 11-13% at the bottom of the scale. The 13F register names the funds that hold each stock and how much the list turned over. StockTwits/X sentiment and the latest news sit beside it.
- **AI.** Each reading carries a short plain-language write-up. "Ask the screen" is a chat that answers from the screen's own rows. AI agents can connect through the MCP server below.
- **Other row kinds.** Handles (tight late-stage consolidations near the pivot), retests (pullbacks to the pivot while the base holds), extensions (past the 5% pivot zone; after 5 days they move to a tracking list), and gap retests (tracked, not emailed).

## Market analysis tools

- **Market Health gauge**: a 0-100 regime score answering "does this tape reward breakouts?" — benchmark trend vs 50/200-day averages (SPY and QQQ combined for the US, taking the worse; NIFTY for India), O'Neil distribution days over the last 25 sessions, and breadth (% of the scanned universe positive over 1 month and 1 week, with a deterioration flag). Three regimes: risk-on (≥70), caution (45-69), risk-off (<45, "avoid new entries").
- **Sector Strength**: every scanned stock's trailing returns rolled up by sector, ranked strongest-first with median 1-week/1-month/3-month returns, top-quintile leader counts, and top names per sector.
- **Base X-ray**: every consolidation base a stock built in the last 2 years — pivot, depth, duration, coil, volume dry-up, failed pokes, and breakout outcome (volume ratio, subsequent run) — drawn as boxes on a full native chart with entry/stop levels, failed-poke and breakout-bar markers, daily/weekly timeframes, and MA50/200 overlays.
- **Market Pulse** (free, no login): live market health, trending tickers, and market news cross-referenced against live signals — at https://dataquant.ai/pulse

## Track record and honesty

The published backtest is deliberately narrow: 22 signals over a 30-day window (May 22 – Jun 20, 2026), each held 20 trading days or stopped at -8%, averaged equal-weight (+6.9% average return, 64% win rate). This is live production data with all signals disclosed, but it is one market regime — a choppier July window ran well below it, which DataQuant reports rather than hides. Results update monthly, wins and losses both; the Market Health gauge exists precisely because breakout systems are regime-dependent.

## Research: exit rules, sizing and regime (September 2026)

Full note: https://dataquant.ai/learn/exit-rules-study — 26 sell rules tested over 97,789 graded (S / A+ / A) blue-sky breakouts across 5,484 NASDAQ and NYSE symbols from the 1970s to September 2026, a 7% hard stop on every rule, then replayed day by day through a slot-limited portfolio (5 or 10 slots, 1% of equity at risk per trade, 12 random within-day orderings). Frictionless fills on today's listings, so returns are relative, not forecasts.

- **Best exit per trade and for the book: trail the stop 20% under the peak high.** 3.73% mean, profit factor 1.80, 95-bar average hold; in the 5-slot book 7.0% CAGR with a 31% max drawdown, the best Sharpe of any rule, a third of the trades.
- **Best fast-turnover exit: close below the 50-day moving average.** 1.10% mean, profit factor 1.35, cuts the hard-stop rate from 45% to 25%. Turnover buys nothing in a slot-limited book (5.4% CAGR, 38% drawdown), and combining MA50 with a trail is worse than either alone.
- **Selling into +20% (whole or half) adds no edge.** Half at +20% then MA50: 0.99% vs 1.10% for plain MA50. Last or near last in every portfolio configuration. Only ~1 in 7 trades reaches +20% under MA50 management.
- **The 20-day average is too tight (PF 1.16); O'Neil's +20%/8-week/MA50 composite adds nothing over MA50 alone (0.99%).**
- **Peter Brandt's 3-day-low trailing stop is harmful from entry (PF 1.03) and a middle path armed after +20% (PF 1.50); stacked on MA50 it clips winners.**
- **Grade dominates the exit.** S-grade beats A-grade by 2-4x in profit factor on every rule (S with the 20% trail: 7.42% mean, PF 2.94).
- **Sizing: 10 slots at 1% risk ≈ 9% CAGR unlevered, best Sharpe; five slots at 2% risk earns the same with twice the concentration.** Fully invested drawdown is 37-39% regardless of risk per trade; only holding cash lowers it (5 slots at 1% = 71% invested, 31% drawdown).
- **Biggest lever in the study: fill free slots strongest relative strength first.** 9.0% → 14.3% CAGR with the same drawdown. An RS floor alone does little; the ordering is what matters.
- **Market regime: the S&P 500 200-day average helps only if it also exits.** Entries-only: 8.0%. Entries and flatten below: 10.9% with the worst drawdown cut from 39% to 21%. Combined with RS ordering: ~14% CAGR, 26% max drawdown, Sharpe 0.92 (12.7% / 29% from 2010 on).
- **Market-health score as a switch: caution (45-69) is not a sell; risk-off (<45) is the line, but health-based rules degrade after 2010 while the 200-day rule holds.**
- **Equity-curve circuit breakers destroy the strategy** (a 15% drawdown halt: 4.4% CAGR, idle 77% of days). Only scaling risk down with drawdown does no harm.
- **Rotation (fixed 42-day holds, strongest-first refills) raises raw return to ~19% at 10 bp costs but with a 55% drawdown; stall signals carry no information beyond a plain clock.**
- **A 90-day time cap is an exit, not a backstop** (costs the trail 2.5 points of CAGR); a one-year backstop is used instead.
- **Base depth: under 10% deep halves the 20-day stop-touch rate but mean return and +20% frequency rise with depth**, so shallow is a tight-tier label, not a gate.

## Guides (plain-English, grounded in DataQuant's live data)

- [Exit Rules Study: What Should Close a Breakout Trade?](https://dataquant.ai/learn/exit-rules-study): Research note — 26 sell rules over 97,789 graded breakouts since the 1970s, replayed through a slot-limited book, then sized and filtered by regime and relative strength. Full tables, methodology and FAQ.

- [What Is a Volatility Contraction Pattern (VCP)?](https://dataquant.ai/learn/volatility-contraction-pattern): The quiet base before a stock's biggest move — DataQuant's exact machine definition of the VCP, and an honest example of a pattern that passed every gate but wasn't tradeable.
- [Relative Strength (RS), Explained](https://dataquant.ai/learn/relative-strength-explained): The 1-99 leadership rank — how DataQuant computes it per market, why leaders lead, and why RS is a filter, not a trigger.
- [Distribution Days, Explained](https://dataquant.ai/learn/distribution-days-explained): Institutional selling into strength — the definition, the 4+ warning threshold, and the August 2026 case where a perfect-looking index hid a failing tape.
- [Reading a Base X-ray](https://dataquant.ai/learn/reading-a-base-xray): Every metric on DataQuant's base cards — pivot, depth, coil, volume dry-up, accumulation, failed pokes, blue sky — with real worked examples.
- [The Honest Backtest](https://dataquant.ai/learn/honest-backtest): How DataQuant measures itself — methodology, the good month and the bad month, and why the numbers are never annualized.
- [The Market Health Gauge](https://dataquant.ai/learn/market-health-gauge): The "M" in market direction — trend, distribution days, and breadth rolled into one 0-100 regime score.

## Full text for language models

The complete text of every guide and research note, in one plain-text file: https://dataquant.ai/llms-full.txt

## MCP server (for AI agents)

DataQuant exposes its free data layer as a Model Context Protocol server at `https://dataquant.ai/mcp` (streamable HTTP, stateless, no auth). Tools: `get_market_health` (the 0-100 regime gauge, US or India), `get_sector_strength` (ranked sector rollups), `get_base_xray` (every base a stock built in 2 years, with metrics and breakout outcomes), and `search_learn` (full-text access to the guides above, with canonical URLs for citation). Signals themselves are subscription-only and not exposed via MCP.

## Key pages

- [Home](https://dataquant.ai/): the four readings, today's screen (live market health and the bases that cleared their pivot), research, pricing and FAQ
- [Market Pulse](https://dataquant.ai/pulse): free live market health, trending tickers, and news
- [Dashboard](https://dataquant.ai/dashboard): the live screen with graded breakouts, a per-stock drawer (overview, chart, technicals, fundamentals, popularity, history), base X-ray charts, sectors and filters (subscription; 30-day free trial, then $49/month or $449/year)

## Disclaimer

DataQuant publishes screen output for research and educational purposes only. It is not a registered investment advisor and nothing on the site is a recommendation to buy or sell any security.