quant-portfolio

Operations — System Architecture & Live Pipeline

Current technical state of the running system. This document reflects the present architecture and is kept current; for the history of how it got here, see CHANGELOG.md. For the project narrative and backtest, see README.md. For the backtest process and rationale behind each sleeve, see RESEARCH.md.

Paper account: DUP447680 (IBKR). Live dashboard: https://isaacnicas.github.io/quant-portfolio/live-dashboard.html

Strategies

Sleeve	Status	Signal	Notes
Trend-following (anchor)	LIVE	TSMOM + CS-Mom, regime filter, fast-exit	Monthly rebalance; ETF universe
Mean-reversion	LIVE	50-day z-score, ±1.5 dead-band	Asymmetric momentum filter, 2-per-sector cap
VRP (VIX contango)	LIVE	SVXY carry, ^VIX/^VIX3M	Gate-governed: suspend / reduce 50% / full size. All three VIX gates currently clear; running at full 10% NAV allocation.
PEAD	BUILT, ORDERS PAUSED	Cross-sectional cohort SUE ranking	Awaiting qualifying 5+ stock cohort (mid-July Q2 season)
FactorTiming	PLANNED (Phase 1B)	—	Pre-registered in governance; no strategy file yet. Dashboard shows “Planned”.
EquityCarry	PLANNED (Phase 2)	—	Pre-registered in governance; no strategy file yet. Dashboard shows “Planned”.

Architecture

data_feed.py — IBKR daily bars (3yr history), port 4002
signal_engine.py — trend signals (TSMOM, CS-Mom, regime filter, fast-exit)
mean_reversion_strategy.py — mean-reversion sleeve
vrp_strategy.py — SVXY-based VRP sleeve
governance_gates.py — VIX risk governor (3-condition gate, 2-day deactivation confirm)
portfolio_risk.py — ERC weighting, 11% vol target (observation mode: computes risk-parity weights daily and logs to portfolio_state.jsonl, but does not size live orders; fixed sleeve fractions are used instead)
order_engine.py — order submission (MKT DAY), port 4002
monitor.py — NAV/position read + dashboard JSON, port 4002
multi_strategy_monitor.py — daily blended report

Execution pipeline

Two-stage split to satisfy order-type timing constraints:

4:15 PM ET — daily_routine_v2.bat: data feed → signals (all sleeves, including trend_following_strategy.py which persists regime state) → governance → orders staged to pending_orders.json → monitor → dashboard push. A consistency check (C1-C8) runs automatically and flags any structural integrity failures.
7:00 AM ET — premarket_submit.bat: submits staged orders as MKT DAY. Health-check email at 7:15 AM.

Automated via Windows Task Scheduler (weekday triggers, AC-power only, 30-min limit, no concurrent instances).

Interactive Brokers connection

Runs headless on IB Gateway, port 4002 (paper). Live port is 4001.
IBC 3.24.0 (IbcAlpha) handles auto-login. Headless launch via C:\IBC\StartGateway.bat; IBC config at C:\IBC\config.ini (TradingMode=paper, ReadOnlyApi=no).
IBC config lives outside the repo; credentials are gitignored and must never be committed.
TWS (port 7497) retained as a manual fallback until Gateway proves stable over several cycles. The old port line is kept commented in each script.

Environment

Python 3.14.6 via the py -3.14 launcher (not bare python / pip).
ib_insync 0.9.86 — requires an asyncio patch at the top of every script for Python 3.14 compatibility.
All scripts at C:\QuantTrading\TrendFollowing\.

Known issues

IBKR subscription Error 10089 → live prices read from local prices.csv rather than requested via API for order sizing.
StartGateway.bat not yet wired into Task Scheduler — Gateway must be launched and logged into paper before the 7:00 AM submit job. Until scheduled, the Gateway path is not fully autonomous.

Roadmap

Wire StartGateway.bat into Task Scheduler ahead of the 7:00 AM submit job.
Mean-reversion universe expansion to S&P 400 MidCap (target ~Jul 10): avg daily volume > $5M, no earnings within 5 trading days, target 80–120 names.
PEAD order activation pending a qualifying cohort (mid-July Q2 earnings).
Factor-timing and equity-carry sleeves (later phases).

Paper trading and research only. Not investment advice.