Net Sharpe
0.83
Gross Sharpe
0.88
Net CAGR
35.49%
Net Total Return
185.03%
Net Max DD
-57.92%
OOS Sharpe
3.16
Code Review Links
- Repo: when2buy/alpha_vault
- Branch:
review/baseline_momentum_live_v1-20260603 - PR: https://github.com/when2buy/alpha_vault/pull/7
- Commit:
98a5d7763bf8bcf4a21564924a15a17bede36b9f
Eval Artifacts
- strategy-eval dashboard
- internal performance/cost/capacity report
- bias report
- summary.md · summary.json
Eval window: 2022-12-13 to 2026-06-02, 869 daily rows, OOS start 2026-01-01.
P0/P1/P2 Summary
- P0 core metrics computed; lookahead shift test passes in the eval report, but lead tests show much higher Sharpe if future data is allowed, so timestamp discipline remains a hard review point.
- P1 OOS since 2026-01-01: Sharpe 3.16, CAGR 370.74%, max drawdown -21.89%.
- Portfolio interaction: max abs correlation 0.46 vs candidate_candidate_quality_momentum_v1; passes <0.8 matrix threshold.
- Capacity: current $20k sleeve is small; eval capacity stress still flags large-AUM review because participation/capacity changes with symbol mix and lower-liquidity names.
Live Fill Status
latest live rebalance complete Last live submit: 2026-05-29T15:30:24.215181Z; open baseline orders: 0; latest 2026-05-29 run filled 7/7 orders.
| Submitted | Symbol | Side | Qty | Notional | Status | Filled Qty | Avg Price |
|---|---|---|---|---|---|---|---|
| 2026-05-29T15:30:09.913443+00:00 | AMD | OrderSide.BUY | 583.91 | OrderStatus.FILLED | 1.138425844 | 512.91 | |
| 2026-05-29T15:30:10.298454+00:00 | INTC | OrderSide.SELL | 16.891977456 | OrderStatus.FILLED | 16.891977456 | 119.02 | |
| 2026-05-29T15:30:10.612675+00:00 | LLY | OrderSide.SELL | 1.672919679 | OrderStatus.FILLED | 1.672919679 | 1093.56 | |
| 2026-05-29T15:30:10.920974+00:00 | LRCX | OrderSide.BUY | 1992.79 | OrderStatus.FILLED | 6.244305899 | 319.137152 | |
| 2026-05-29T15:30:11.055588+00:00 | MU | OrderSide.BUY | 242.99 | OrderStatus.FILLED | 0.25091075 | 968.432 | |
| 2026-05-29T15:30:11.394449+00:00 | QCOM | OrderSide.SELL | 7.437664439 | OrderStatus.FILLED | 7.437664439 | 250.91 | |
| 2026-05-29T15:30:11.525874+00:00 | SOXX | OrderSide.BUY | 512.72 | OrderStatus.FILLED | 0.899095503 | 570.262 |
Current Managed Positions Snapshot
| Symbol | Qty | Market Value | Unrealized P&L | Unrealized % |
|---|---|---|---|---|
| AMD | 7.603972844 | 3916.806412 | 450.264614 | 0.12989 |
| INTC | 19.678094294 | 2168.525991 | -217.247794 | -0.09106 |
| LRCX | 6.244305899 | 2098.149225 | 105.359224 | 0.05287 |
| MU | 7.22833266 | 7716.967948 | 2319.574871 | 0.42976 |
| QCOM | 12.399335549 | 2962.201263 | 243.564513 | 0.08959 |
| SOXX | 3.495181503 | 2139.365646 | 284.37832 | 0.1533 |
| UNH | 0.000000861 | 0.000325 | -0.000016 | -0.04692 |
Review Entrypoints
AGENTS.md: operational handoff, live state, cron and duplicate-submit guard.strategy_manifest.json: PAW/AlphaVault strategy metadata.tools/ops_baseline_deploy.py: StrategyRegistry registration, signal generation, target sizing, ExecutionGateway order creation, metrics/state writes.scripts/run_ops_baseline_live.sh: live cron launcher, lock and duplicate-submit guard.core/alpaca_execution.py,tools/backtest_report.py,tools/rolling_sharpe_report.py,tools/generate_baseline_page.py: data/backtest/report path.
Known Risks / Required Review
- Lookahead/timestamp: live runner drops in-progress daily bars; reviewers should verify all signal bars are closed and execution happens after signal timestamp.
- Schedule guard: cron is UTC
30 19 * * 5, targeting 15:30 New York only during DST. Needs a winter-time rule or scheduler that understands America/New_York. - Duplicate-submit guard: currently guards by UTC date and
last_live_submit_at; it prevents same-day double submit but is coarse for manual repair runs. - Order ids: strategy attribution depends on client_order_id prefix
av_basemomelivev1. - Sizing/cap: buys use notional, sells use 9-decimal qty; $20k sleeve is fine, but large AUM needs symbol-level participation caps.
- Broker/fill assumptions: eval uses cost/slippage assumptions from metadata; live market orders can differ during volatile windows.
- Shared account isolation: live account is shared at account level; strategy isolates via baseline subtraction, StrategyRegistry, client_order_id attribution, and per-strategy lock. Account-level lock remains a TODO.
- Formal promotion: PBO/DSR remain needs_input without a candidate-family trial panel; live readiness should not rely on this report alone.