second_derivative_balanced_paper_v1 Review Package
Generated 2026-06-03T08:38:53.364335Z from local strategy-eval outputs and AlphaVault paper state.
paper onlyshared Alpaca paper accountweekly Friday 19:30 UTC
Gross Sharpe
1.778
Net Sharpe
1.677
Net CAGR
34.65%
Net MaxDD
-25.96%
OOS Sharpe
3.266
Max Corr
0.632 vs candidate_ai_infra_hedged_v1
TC Drag
1.8860% NAV/yr
Paper Probe
MRK 100.0 status=filled
Decision Context
| Area | Status | Notes |
|---|---|---|
| P0 eval | PASS | Bias shift/autocorr passed; core net Sharpe 1.677; drawdown -25.96%. |
| P1 promotion | Review | OOS Sharpe 3.266; portfolio max corr 0.632; paper trading return stream still needs input. |
| P2 live readiness | Not live-ready | Fill assumptions are needs_input; live promotion requires queue, latency, spread crossing, and more paper fills. |
| Current paper status | filled | One-shot paper probe av_secderivbalpaperv_20260602T181724Z_MRK_b_001; symbol MRK, notional 100.0. |
Review Entrypoints
strategies/second_derivative_balanced_paper_v1/strategy/runner.pystrategies/second_derivative_balanced_paper_v1/strategy/signal_model.pystrategies/second_derivative_balanced_paper_v1/tests/test_signal_model.pystrategies/second_derivative_balanced_paper_v1/README.mdscripts/run_second_derivative_balanced_paper.sh
Known Risks
- Paper account is shared with math_three_signal_blend_paper_v1 and huang_infra_bottleneck_barbell_paper_v1; sleeve isolation depends on ignored_baseline and must not sell peer holdings during ad-hoc probes.
- Normal submit path has a Friday 19:25-19:45 UTC guard; approved one-shot probes must not permanently disable this gate.
- Duplicate guard uses state.json last_paper_submit_at UTC date; manual probes can block same-day repeats but not later Friday cron if on a different UTC date.
- Execution uses fractional notional buys and broker-position qty sells through ExecutionGateway; review dust handling and sell attribution before full rebalance.
- Client order id short prefix is secderivbalpaperv; confirm no collision with other strategies.
- Research eval is based on daily bars and supplied positions/returns; PBO/DSR and fill/latency assumptions are needs_input.
- OOS Sharpe is strong but max drawdown is about -26%; formal live promotion needs hidden holdout, fill model, and paper return stream.
Artifacts
/workspace/steve/alpha_vault_worktrees/review_second_derivative_balanced_20260603/strategies/second_derivative_balanced_paper_v1/reports/review_20260603/eval-report/summary.md/workspace/steve/alpha_vault_worktrees/review_second_derivative_balanced_20260603/strategies/second_derivative_balanced_paper_v1/reports/review_20260603/eval-report/summary.json/workspace/steve/alpha_vault_worktrees/review_second_derivative_balanced_20260603/strategies/second_derivative_balanced_paper_v1/reports/review_20260603/eval-report/eval_dashboard.html/workspace/steve/alpha_vault_worktrees/review_second_derivative_balanced_20260603/strategies/second_derivative_balanced_paper_v1/reports/review_20260603/eval-report/bias_report.html/workspace/steve/alpha_vault_worktrees/review_second_derivative_balanced_20260603/strategies/second_derivative_balanced_paper_v1/reports/review_20260603/eval-report/internal_report.html
Notes For Reviewers
AlphaVaultOps is treated as control-plane only: runtime config, schedule registry, state, event logs, and broker gateway. Strategy code lives in AlphaVault under the strategy folder; no secrets are included in this package.