MarketSightAI Workbench & Bar Replay Guide
Master the differences between Full Inspection and Blind Replay (?replay=1), 4 core UI sections, and 5-step causal logic.
Full-Session Static Forensics
Displays 100% of session bars and markers with unrestricted navigation. Best for reviewing whole-day macro structure and algorithm accuracy.
Bar-by-Bar Blind Replay Challenge
Future bars strictly hidden to eliminate lookahead bias. Step forward bar-by-bar, auto-play, and test your instincts via Bar N+1 predictions.
1. Core Modes Comparison: Inspection vs Blind Replay
| Dimension | Full Inspection (/workbench/) | Blind Replay (?replay=1) |
|---|---|---|
| Candle Visibility | 100% complete view. All session candles visible. | Future bars physically hidden to eliminate lookahead bias. |
| Topbar Controls | Displays [Bar Replay] entry button; step dock hidden. | Active dock: ◀ Prev, Next ▶, -10/+10, Auto-play, and Exit. |
| Primary User Flow | Click any candle freely to inspect its 5-step causal verdict. | Submit Bull/Bear/Chop decision before advancing. |
| Scorecard & Ranking | Static session statistics and scorecard export. | Real-time grading: Accuracy %, streak, and Trader Rank badges. |
| Best Use Case | Post-market macro review and full-session signal audits. | Mastering Al Brooks BPA instincts, disciplining trader psychology. |
2. In-Depth Breakdown of 4 Core UI Modules
Top Navigation & Control Bar
Provides global orchestration and mode switching across desktop and mobile:
- Symbol & Trade Date: Clearly displays current symbol (e.g. NASDAQ.QQQ, CFFEX.IM) and historical closed trading date.
- Preset Session Selector: Dropdown with 15+ curated causal sessions covering US equities, futures, options, and index ETFs.
- Replay Dock Controls: In replay mode, expands dock controls: ◀ Prev, Next ▶, -10/+10, Auto-play, and Exit.
- Dual HUD Mode: Toggles floating sub-window and dual matrix columns for multi-timeframe pattern inspection.
- Import Local JSON: Supports drag-and-drop import of offline WorkbenchSession JSON files for zero-leak interactive audit.
Main Chart Stage
Powered by TradingView Lightweight Charts engine with causal bar-by-bar recalculations:
- Multi-Period MAs: Dynamically renders EMA20, EMA50, EMA200, and institutional VWAP baselines.
- Key Price Magnets: Plots Opening Range extremes (ORH/ORL) and Prior Day magnets (PDH/PDL/PDC) for liquidity sweeps.
- Pattern & Signal Badges: Pins verified Price Action badges (Double Bottom, Breakout Bar, H2/L2) directly above/below signal candles.
- Zero-Lookahead Updates: Under blind replay, every step cleanly recalculates state without future leakage.
Cloud Lake & Bar N+1 Dock
The primary interactive control hub at the top of the sidebar, integrating lake navigation and decision testing:
- 4-Level Cascade Filter: Instant jumps across Strategy ➔ Symbol ➔ Date ➔ Setup directly to target setups.
- 🎲 Random Blind Replay: Picks a random signal bar slice across the cloud data lake for unannounced blind replay practice.
- ⏮ Setup Jump: Skips sideways chop to jump straight between high-probability setup signal bars.
- 🎯 Bar N+1 Quiz: Predict 🟢 Long, 🔴 Short, or ⚪ Wait; receive instant scoring against the causal price action model.
- Trader Rank Badges: Dynamic rank progression (e.g. Al Brooks PA Specialist) based on cumulative accuracy and win streaks.
5-Step Causal Audit Panels
Provides comprehensive mathematical dissection and institutional-grade algorithmic scrutiny for each bar:
60m macro trend bias, 15m structure, and 5m execution state.
Candle body %, close percentile, bar overlap %, and follow-through.
Maps geometric anchors and explains why low-probability signals were filtered.
Outputs primary directive, mathematical EV, entry, stop-loss, and target.
Tracks trade lifecycle states and computes exact price watershed threshold.
3. Local Python Backtest JSON Import & Data Privacy
If you run MarketSightAI’s Python engine locally, you can export session packages via `systems.workbench.exporter`. Simply click [Import JSON] in the top navigation. Your charts and 5-step forensic records render entirely inside your browser (zero data leaves your device).