Each strategy bot can be told to skip an entry when market conditions look unfavourable (e.g. don't go long when SPY is falling). Right now those filters are log-only — bots still trade, but they record which trades would have been blocked if the filter were enforced. After enough trades, this table tells us whether the filter actually predicts losers.

Flagged = the filter would have blocked this entry  |  Unflagged = filter said OK

Verdict: enforce ✓ = flagged trades did measurably worse than unflagged (flip the filter ON); drop ✗ = flagged trades did better than unflagged (filter is wrong-signed — drop it); neutral = no clear difference; need data = fewer than 5 trades in either bucket.

The flags currently being tracked:

• spy_against_long — long entry while SPY is below its intraday VWAP (against the tape).
• spy_against_short — short entry while SPY is above its intraday VWAP.
• vix_low — selling premium when VIX < 14 (premium too thin to compensate for tail risk).
• rs_underperformer — wheel CSP on a name underperforming SPY by > 10% over 30 days (future INTC/QCOM-style holders).
• earnings_close — opening a CSP within 5 days of earnings (assignment risk at the post-earnings price).
• stopped_out — strategy already lost > $300 today; further entries are throwing good money after bad.