Crosshair

Find the shot · take the shot
Snapshot
Loading…

Market

Loading…

Week Map — NQ prop engine

Loading…

Real account — Robinhood mirror

Loading…

Paper agent — Tradier sandbox

Loading…

Activity — Claude check-ins

Loading…

The decision screen: status first, evidence one tap deeper. Full analysis lives in Today; controls live in Desk. Labels: RULE-BASED deterministic code · AI RESEARCH model suggestions · LIVE human-controlled.

Sector heat map

20-day money flow · green = strength

Loading…

Loading tape…

Trade posture

Loading…

Catalysts ahead

Loading…

Autotrader (paper)

Loading…

Everything here updates from live feeds when the tab opens. Tap any icon for a plain-English explanation. Educational only — not financial advice.
Go deeper — the full market data — macro regime · sectors · chart

Sector momentum — 20 days

Loading…

SPDR sector ETFs, 20-day return (Yahoo). The same ranking is injected into every AI play prompt.

Chart

TradingView embedded chart (free widget) — full indicators via the toolbar. Loads when this tab opens.
L1 · Macro regime
What changed—
    —
    —

    Where we are

    GOLDILOCKS growth ↑ · inflation ↓ REFLATION growth ↑ · inflation ↑ SLOWDOWN growth ↓ · inflation ↓ STAGFLATION growth ↓ · inflation ↑ GROWTH → INFLATION →
    Growth axis—
    Inflation axis—
    Trajectory origin: seeded baseline

    Macro inputs — edit to stress-test

    Policy & risk overlay

    Sector tilts — macro view

    L2 · Sector tape

    Layer 2 · Does price confirm the macro?

    Each sector's macro tilt (L1) set against its actual relative-strength momentum. Agreement = high conviction. Conflict = the macro says one thing and the tape says another — the trades worth a second look.

    Snapshot · 17 Jul 2026
    SectorMacroMomentumRead
    L3 · Stocks in favored sectors

    Layer 3 · Names inside the favored sectors

    The most liquid leaders in the sectors that clear both filters (macro + momentum). These are a research shortlist, not signals — precise per-name ranking needs the market-data backend. Focus sectors are driven live by L1 × L2.

    📅 Market rhythms this week — recurring calendar patterns, graded by evidence

    Intraweek playbook — today & tomorrow

    Evidence-graded timing patterns for the trading week. strong = robust & current · moderate = real but conditional · weak = historical/decayed, shown for awareness. Bias, not a guarantee.

    Today

    —

    Tomorrow's hints

    —

    Event radar

    Your agent

    A robot trader that never sleeps: checks the tape every 30 seconds, hunts swing setups every 15 minutes and scalps every 5 (paper money), and emails you when a watched level trips. The AI thinks; the agent watches and executes paper.

    on duty
    Live means approval-gated, not autonomous. The agent may watch and propose continuously. It cannot place a live entry unless deployment enables live execution and you approve that exact, unexpired proposal. Paper mode can trade autonomously.

    Ask the AI for its one best play

    The AI plays a trader running a hypothetical $10,000 options account — aggressive but smart, happy to tell you to sit on your hands, and willing to size up on a heavy hunch. He looks at the live tape, options data, sectors, and catalysts, then gives you one play (not a menu) with exact entry, profit target, and stop — or a straight "not today." Educational only — not financial advice.

    Not run yet

    Bull vs Bear

    Two AIs argue today's market — one all-bull, one all-bear — and a third judges the fight and says what a $10k trader should do about it. Adversarial pressure kills lazy theses.

    Work a ticker together

    Type any symbol. The AI argues both sides of it and names the one price level that matters. Like the setup? Hand it to your agent — he watches that level every 30 seconds and emails you the moment it trips.

    Live market context

    Loading live trend & options…

    Sandbox autotrader paper money

    The AI trades the 0DTE bucket automatically in a Tradier sandbox (fake $100k) — every play is validated against the real-time chain, capped at $600 risk, defined-risk only, then placed as a paper order. Runs each weekday ~10:05am ET; real money is not connected.

    Play journal

    Every generated set of plays is logged server-side (90 days). Load to see how each underlying has moved since the ideas were generated — the honest scoreboard.

    Read this first. These ideas are generated on demand by an AI model for a hypothetical $10,000 account and a specific risk profile. They are not personalized advice, may be wrong or out of date, and do not account for your situation, taxes, or costs. 1–3 DTE options decay fast and can go to zero. Do your own research; consider paper-trading first.

    Legacy live-account mirror — REAL MONEY

    Read-only state published by the prior session-based workflow. Agent execution uses Tradier and is shown in the proposal and audit sections above.

    Account

    Loading…

    🎯 On the watchlist setups & triggers

    Trades the AI is stalking but hasn't taken — each with the trigger that would make it live. Nothing here is an open position.

    Ground rules. Designated trading capital only. Every order is broker-previewed, tagged for deduplication, reconciled after submission, and audited. Filled long options receive a broker-native stop; a live reconciliation failure trips the kill switch and sends a critical alert.

    📌 RSI-2 pullback — daily watch adopted · signal only · SPY + QQQ

    Loading today’s reading…

    How this works — tap to expand. Adopted from the 2026-08-01 strategy tournament; the watch flags setups, and every entry stays a human decision.

    The rule. Buy SPY or QQQ at the close when the 2-day RSI dips under 10 while price is still above its 200-day average — a sharp 2–3 day dip inside a healthy uptrend. Sell the close that finishes back above the 5-day average, usually 3–6 trading days later. Long only, shares (not options — no theta clock on a multi-day hold), no leverage, no stop in the backtest (the 200-day filter is the regime stop).

    Why it's credible. This is the Connors RSI-2 family — short-term mean reversion in index ETFs — with decades of published out-of-sample persistence, not a curve-fit of ours. Our own 10-year backtest (SPY+QQQ daily, $10k notional per trade): 160 trades, 72% win, profit factor 2.56, +$9,755, max drawdown −$1,086, expectancy ~$61/trade. Same test, Week Map benchmark on its 130-day tape: PF 1.32. Full chart in the journal/session record.

    Why it fits this account. Low frequency (~1–2 setups a month), needs only end-of-day checks — no screen time, no intraday automation. It buys fear inside uptrends, so its edge shows up in exactly the choppy pullback weeks that starve the trend-following prop bot — an uncorrelated second engine, not a duplicate.

    Honest caveats. Holds through overnight/gap risk with no hard stop — the worst single trade was −$458 per $10k, but a 2020-style crash entry would be worse. Backtest ignores dividends (helps, slightly). Adopted 2026-08-01 as a watch rule only: the Worker recomputes this daily, the card above shows the reading, and any actual entry is Zac's call at the close.

    📖 Playbook — every strategy we've tested reference · as of 2026-08-01

    The whole research record in one place: what each strategy is, its profit factor left alone (recent window vs all history), when it wants to trade, and its honest status. Evidence lives in docs/prop-bot.md; the forward lanes keep score and this card gets updated when a status changes.

    🟢 RSI-2 pullback (shares) — swing, 2–6 day holds · VALIDATED · adopted watch
    PF left alone: 2.56 over the last decade · 2.55 over the prior 23 years (positive in every era incl. dot-com and 2008). A 30-day stretch usually holds only 1–3 trades — judge it by quarters, not months.
    When it trades: SPY/QQQ above the 200-day only; fires on 2–3-day fear dips (~1–2/month). Uncorrelated with the intraday book — it works the daily clock.
    The strongest evidence in the house: full gauntlet pass, 33 years, 450+ trades. All-weather instrument.

    🟢 RSI2-LEAP-60d — long-term, ~3-month holds · VALIDATED (historical) · paper-first
    Same trigger, instrument = 10%-ITM 365-DTE call held ~60 sessions. PF left alone: ~7 in the recent decade · 2.61 over 23 untouched years (modeled premiums). Timing beats the no-timing control everywhere, most in 2008 (+$2,154 vs −$1,984).
    When it trades: same signal as shares. Known weak spot: lags shares in deep bears (00–02) — vol tax on long calls. Fair-weather amplifier, not a replacement.
    Forward paper record runs from the next armed signal; shares stay the adopted instrument until it earns the switch.

    🟡 Week Map (NQ 8/14 EMA) — day trade, futures · REGIME-LOCAL · dry-run bot
    PF left alone: 1.32 on the current-regime 130d tape (~30 trades/30d) · 0.88–0.93 on 2022–25 history — the edge exists only in the 2026 tape character so far. Discovery Reality Check p=0.37.
    When it trades: NQ RTH only (edge does NOT travel — ES/GC/IWM all negative). Wants energy: bottom-tercile momentum days are its worst tape in every era (the WM_STANDDOWN lane skips them). Conviction = the regime dial above (rolling forward PF20), the only regime detector that survived testing.
    House verdict on props: recorded recommendation was against funding on this evidence; proceeding anyway = a deliberate bet the current regime persists — minimum size, dial confirming, stand down on the tripwire (rolling 20-trade PF < 1.0 or 3 straight red weeks).

    🟡 Overnight drift (NQ) — overnight holds · PARKED · personal account only
    Buy the 16:00 close, sell the 9:30 open. PF 1.39 / +$20k on the 130d window at 3 MNQ — but never gauntleted on long history, and structurally incompatible with prop rules (holds through liquidation). Worst night −$5,708.
    If ever revived: personal futures account only, and it owes the full gauntlet first.

    ⚫ The graveyard — tested and killed, with cause of death
    IDX-1050 (ES/RTY 20/50 cross): looked great in-sample (PF 1.37–1.50), survived fill audits and real-futures checks — then 2.2y untouched history came back negative on both and the Reality Check hit p=0.62. The reference lesson: same-window replications can share one regime artifact.
    Level-rejection scalps (NQ 1m, 3 structures, gated + ungated): ~500 trades, zero profitable configs. Costs own the 1–10min horizon.
    VWAP 2σ reversion (PF 0.91) · ORB standalone (PF 0.96) · Mag7 pop via options (PF 0.14–0.49) · TTM squeeze entries/exits (drag both sides) · scaled/BE exit variants (all lost to the plain bracket).
    Stand-down days: nothing works there — all 14 entry families negative across 3 eras. Flat is the trade.
    House rules distilled: nothing gets believed before the gauntlet (untouched multi-year history + selection-corrected stats), the forward record outranks every backtest, and PF numbers here are simulations — real fills decide.

    Agent

    Loading…

    Risk & limits display-only · enforced server-side

    Loading…

    Connections

    Loading…

    Learn

    The 16 lessons in force and every sourced write-up — the evidence behind the rules.

    Device

    Forgets the stored control key on this device. Theme & accent live behind the ⚙ up top.

    Knowledge bank — what we trade on, and why

    Every rule the system trades with, distilled from our own backtests, the forward record, and sourced research. Governance: the forward record outranks everything, and origin of an idea is not evidence. Full sourced write-ups live in the repo (links open GitHub).

    Lessons in force

    Loading…

    What this is. A disciplined, transparent read of the cycle and the tape — decision-support, not investment advice or a predictor. Tilts are relative, the frameworks are probabilistic, and a transitional regime gives a weaker signal than a clean one. Everything here runs on live feeds: macro from FRED, tape and sectors from Yahoo, options metrics from CBOE, headlines from the wires.
    Crosshair · a private trading cockpit
    macro regime → sector momentum → the shot