Post-market dummy refinement: 2026-07-23

[github-actions]

Goal

Assign this issue to Copilot as a daily post-market refinement task.

Review today's dummy-trading behavior, identify only evidence-backed improvements, and propose a small safe change if the data justifies it. Do not add strategy parameters randomly.

Latest Paper P&L

• Paper date: 2026-07-23
• Portfolio value: ₹500000.0
• Today P&L: ₹0.0 (0.0%)
• Total P&L: ₹0.0 (0.0%)
• Alpha vs NIFTY: 1.522%
• Positions: 0
• Trades in latest paper run: 0

Incubation Gate

• Decision: incubating
• Days: 4/20
• Blocking checks: 9/11
• Failed checks: min_trading_days, backtest_oos_positive

Live Market Snapshot

• Live snapshot IST: 2026-07-23T15:26:36.526591+05:30
• Market open flag: true
• Live day P&L: ₹0.0 (0.0%)
• Live actions: 0
• Stops: 0

Latest Dummy Transactions

• No EOD paper trades in latest run

Daily Analysis Context

• Trading date: 2026-07-23
• Signal counts: BUY 32, HOLD 18, SELL 0
• Market regime: NIFTY trend filter is risk-off

Backtest Lab Context

• no_regime_filter: return 8.22%, alpha 6.86%, maxDD -15.26%
• momentum_only_weekly: return 6.75%, alpha 5.39%, maxDD -10.01%
• momentum_weekly_churn_control__news_on: return 6.75%, alpha 5.39%, maxDD -10.01%

Copilot Instructions

• Inspect the dashboard/data outputs and current strategy code.
• Look for repeated or material failure patterns: stale data, bad churn, weak ranking, poor risk behavior, oversized drawdown, or confusing UI.
• If there is no strong evidence for a code change, say so and leave the system unchanged.
• If a change is justified, keep it small, testable, and paper-only.
• Preserve the one-month incubation rule. Do not recommend real-money trading.
• Include validation steps and expected effect on the readiness gate.

[claude]

No code change today — evidence matches the already-tracked finding from #23/#26

Reviewed today's snapshot (2026-07-23) against the refinement rules. Conclusion: no evidence-backed change is warranted.

What was checked

1. 0 trades / 0 positions despite BUY 32, HOLD 18, SELL 0. Same root cause as the last two days: USE_MARKET_REGIME_GUARD in src/paper_trader.py sets risk_block_new_buys = True whenever src/market_regime.py reports the NIFTY trend filter as risk-off, regardless of signal count. With 0 existing positions there's also nothing to sell. This exact scenario (risk-off, 0 trades, buy-ranked candidates present) is already locked in by a regression test, test_market_regime_risk_off_blocks_new_buys in tests/test_paper_trader.py, on the still-open PR Test: lock in risk-off regime guard blocking new buys #24 (branch auto-refine/issue-23-2026-07-21). No new behavior surfaced today that isn't already covered there — merging Test: lock in risk-off regime guard blocking new buys #24 remains the right way to close out this thread rather than opening a duplicate PR.

2. Trend across the last 3 days is healthy, not alarming: BUY signals falling (40 → 37 → 32) and HOLD rising (10 → 13 → 18) as the risk-off regime persists, while alpha vs NIFTY has actually improved each day (0.21% → 0.999% → 1.522%). This is the risk guard doing its job (avoiding new exposure in a risk-off tape), not a failure pattern.

3. Incubation gate: 4/20 days, failing min_trading_days and backtest_oos_positive. Both fail by construction this early — min_trading_days (src/incubation_report.py:248-251) is a plain day-count needing 20, and backtest_oos_positive (src/incubation_report.py:317-320) needs more walk-forward history to resolve. No bug indicated.

4. Backtest lab numbers are byte-identical to the last several days (no_regime_filter 8.22%/6.86%/-15.26%maxDD; both momentum_only_weekly and momentum_weekly_churn_control__news_on at 6.75%/5.39%/-10.01%maxDD). Consistent with the prior explanation in Post-market dummy refinement: 2026-07-22 #26: the news archive (added 2026-07-19) still has too few accumulated days (ablation_verdict() in src/news_ablation.py reports insufficient_data when news_days_used <= 0) to move the news-on variant away from the news-off baseline over the historical replay window. Self-resolves as the archive grows.

5. Test suite: python -m unittest discover -s tests -p "test_*.py" → 82 tests, 1 pre-existing unrelated error (test_news_archive fails to import: ModuleNotFoundError: No module named 'requests', a missing sandbox dependency listed in requirements.txt, not a code defect), 12 skipped, rest passing.

Per the daily-refinement rules, "no change" is correct when the data doesn't clearly justify one — that's the case again today. Closing this issue; #24 remains the tracked follow-up for the recurring risk-off observation.

[github-actions]

🤖 Daily refinement — no code change today. No code change: 2026-07-23 snapshot (BUY 32/HOLD 18, risk-off regime, 0 trades, day 4/20) matches the already-diagnosed and test-covered pattern from #23/#26 (open PR #24); commented with evidence and closed issue #27.