Edge Engine

Quantitative Trading Pipeline Dashboard

Detail

Edge Engine

๐Ÿ—๏ธ Platform Health Building
Alpha (Notebooks)29/29 โœ“
Beta (Pipeline Scripts)29/29 โœ“
Hotpath (Rust)Not Started
Active Issuesโ€”
๐Ÿ”ฌ Pipeline Status Ready
Config Cache29/29 RPCs
Feature Cache1,763 trades
Offline ReadyYes โœ“
Last RunNever
๐Ÿ“ˆ Strategy Status Pre-Production
Active StrategyCSP via ML Cascade
Current Versionfactor_timing_v2.0.0
Prior Versionv1.0-binary (FAILED)
C-Findings Open3 of 8
โฑ๏ธ Daily Operations Not Active
Shadow TradingNot Started
Existing Shadow Trades67
Daily ScansNot Scheduled
Drift DetectionNot Built
โšก Recent Issues
Loading issues...
๐Ÿ—๏ธ Platform Health โ€” Alpha / Beta / Hotpath
๐Ÿ—„๏ธ Infrastructure
๐Ÿ”ฌ Pipeline Flow โ€” 7 Phases, 29 Steps

Model: factor_timing_v2.0.0 | Session: No active session

Click any step to view details in the side panel โ†’

๐Ÿ“ˆ Strategy Lineage โ€” CSP via ML Cascade

Thesis: Cash-secured put selection via 3-stage ML cascade (win predictor โ†’ catastrophe filter โ†’ EV calculator)

โœ—
v1.0-binary FAILED
Expert review panel: unanimous non-certification. 8 critical findings (C-1 through C-8).
๐Ÿ“… Jul 2025โŒ Sharpe -0.89๐Ÿ”ด AUC gap 0.256๐Ÿ”ด 3 models were 1
Click to expand details โ–พ
v1.0-binary โ€” Expert Review Summary

Review Date: July 2025
Outcome: Unanimous non-certification (3/3 reviewers)
Sharpe Ratio: -0.89 (target: >0.5)
AUC Train/Test Gap: 0.256 (severe overfitting)

8 Critical Findings:
C-1: Three models are actually one (identical feature weights)
C-2: Severe overfitting โ€” AUC gap of 0.256
C-3: Paper trading broken โ€” 0 trades executed
C-4: Catastrophe filter only 20% recall
C-5: No position sizing at all
C-6: Only 3 walk-forward folds (need 10+)
C-7: Feature redundancy โ€” 62 features, many correlated
C-8: Class imbalance not addressed

Decision: Full pipeline rebuild required โ†’ factor_timing_v2.0.0
โ†’
P0 Fixes Done
fillna(0) โ†’ fail-fast, MC contamination fix, architecture plan, shadow orchestrator, Prometheus metrics.
๐Ÿ“… Jan 2026โœ… C-2 contamination fixed
Click to expand details โ–พ
P0 Fixes โ€” Emergency Remediation

Timeline: December 2025 โ€” January 2026
Scope: Immediate fixes before full rebuild

Changes Made:
โœ… Removed all fillna(0) calls โ†’ replaced with fail-fast assertions
โœ… Fixed Monte Carlo contamination (train data leaking into test)
โœ… Designed 7-phase pipeline architecture (ADR-1)
โœ… Built shadow trading orchestrator (499 lines)
โœ… Added Prometheus metrics schema

Impact: Unblocked v2.0.0 pipeline development
โ—
factor_timing_v2.0.0 In Pipeline
Full pipeline rebuild: 7 phases, 29 steps. Addressing all 8 C-findings. New bucket architecture, HPO, walk-forward.
๐Ÿ“… Feb 2026๐ŸŸข C-1 โœ“๐ŸŸข C-2 โœ“๐ŸŸก C-4 partial๐ŸŸข C-5 โœ“๐ŸŸก C-6 built๐ŸŸข C-7 โœ“๐ŸŸข C-8 โœ“
Click to expand details โ–พ
factor_timing_v2.0.0 โ€” Current Build

Architecture: 7 phases, 29 steps, StepRunner framework
Total Code: 12,941 lines (scripts) + 6,522 lines (shared)
Offline Test: 0 DB violations, all 29 steps pass structural check

C-Finding Resolution:
โœ… C-1: Unique model fingerprints enforced (Step 2.1 Uniqueness Agent)
โœ… C-2: Contamination fixed, expanding window walk-forward
๐ŸŸก C-3: Script built (499L), needs real data run
๐ŸŸก C-4: Improved approach, 50%+ recall target (was 20%)
โœ… C-5: Half-Kelly sizing with $0.65/contract friction
๐ŸŸก C-6: 10+ fold expanding window built (1,051L), needs data
โœ… C-7: GATE step with SHAP + permutation importance
โœ… C-8: Configurable SMOTE/weighting in training step

Remaining Work: Run with real data, resolve issues, certification
โ—‹
Shadow Trading Next
30-day observation period. Daily scan, paper execution, performance tracking. Minimum 50 trades.
๐Ÿ“… Target: Feb-Mar 2026โณ Pending certification
Click to expand details โ–พ
Shadow Trading โ€” Validation Phase

Duration: 30 calendar days minimum
Trade Minimum: 50 paper trades
Success Criteria:
โ€ข Sharpe ratio > 0.5
โ€ข Max drawdown < 15%
โ€ข Win rate > 55%
โ€ข No catastrophe filter misses
โ€ข Drift detection: < 2ฯƒ feature drift

Prerequisites:
โณ Pipeline certification (Phase 6)
โณ All issues resolved
โณ SLACK_BOT_TOKEN configured

Status: Blocked on pipeline certification
โ—‹
Production Future
Rust hotpath integration. Real-time inference via Polygon.io WebSocket. Live trading with circuit breakers.
๐Ÿ“… Target: Q2 2026โณ Pending shadow validation
Click to expand details โ–พ
Production โ€” Live Trading

Target: Q2 2026
Architecture: Rust hotpath with ONNX model inference
Data Feed: Polygon.io WebSocket (real-time)

Components Needed:
โ€ข ONNX model export from certified Python model
โ€ข Rust inference engine (sub-millisecond)
โ€ข Broker API integration (order routing)
โ€ข Circuit breakers (max loss, max positions, volatility halt)
โ€ข Prometheus + Grafana monitoring

Prerequisites:
โณ Shadow trading validation (30 days, 50+ trades)
โณ Risk committee approval
โณ Broker account setup
๐Ÿ“‹ C-Finding Resolution Tracker
โœ…C-1: Three models are actually one โ†’ Unique model fingerprints enforced (Step 2.1 Uniqueness Agent)
โœ…C-2: Severe overfitting (0.256 AUC gap) โ†’ Contamination fixed, expanding window walk-forward
โš ๏ธC-3: Paper trading broken (0 trades) โ†’ Beta script built (499L), needs real data run
โš ๏ธC-4: Catastrophe filter 20% recall โ†’ Improved approach, 80% acknowledged aspirational, target 50%+
โœ…C-5: No position sizing โ†’ Half-Kelly sizing with $0.65/contract friction
โš ๏ธC-6: 3 walk-forward folds โ†’ 10+ fold expanding window built (1,051L), needs real data
โœ…C-7: Feature redundancy (62 features) โ†’ GATE step with SHAP + permutation importance
โœ…C-8: Class imbalance not addressed โ†’ Configurable SMOTE/weighting in training step
โฑ๏ธ Daily Operations Lifecycle

Recurring daily events for shadow trading and system health. Not yet active.

๐Ÿ“… Today's ScheduleNot Active
08:00Morning Scan โ€” Evaluate universe against cascade pipelineโฌœ
08:30Trade Execution โ€” Paper trades for approved signalsโฌœ
12:00Midday Check โ€” Position monitoring, stop loss reviewโฌœ
15:00EOD Settlement โ€” Mark final PnL, update snapshotsโฌœ
16:00Daily Slack Report โ€” Trades, PnL, running metricsโฌœ
20:00Drift Detection โ€” Feature distribution monitoringโฌœ
21:00Health Check โ€” DB connectivity, cache freshness, model integrityโฌœ
22:00Backfill โ€” Sync any missing trade/feature dataโฌœ
Shadow TradingNot Started
Observation Periodโ€”
Trades (Shadow)67 (historical)
Win Rateโ€”
Sharpeโ€”
DateTickerTypePremiumOutcome
Alerts & Escalation
Unacked Alerts0
Open Investigations0
Kill SwitchReady
๐Ÿ› Issues & Error Tracking

Live from pipeline_errors table. Click any issue to expand details.

Total
โ€”
Critical
โ€”
Open
โ€”
Resolved
โ€”
Loading issues...