Bid Trade
Mode:

Notifications

Options Learning

Aggregate analysis of closed options trades with actionable recommendations

to
Total Trades
--
Win Rate
--
Best Structure
--
Theta Contribution
--
By Directional Bias

Trade-level bias inferred from structure + premium sign. Neutral = non-directional structures (iron fly/condor, butterfly, calendar, straddle).

Bias Trades Win Rate Avg P&L Total P&L
Loading...
By Premium Direction

Credit = short premium (collected at entry). Debit = long premium (paid at entry). Derived from entry_credit sign.

Premium Trades Win Rate Avg P&L Total P&L
Loading...
Structure Performance
Structure Trades Win Rate Avg P&L Total P&L
Loading...
IV Rank Impact
Exit Tier Analysis
Delta Exposure
Theta Decay vs Actual P&L
Theta Capture Efficiency (short premium only)

What % of theoretical theta decay actually lands as P&L. Low capture = gamma/vega is bleeding.

Loading…

Implied vs Realized Move per structure

Did IV predict the move? Ratio < 1 = IV overpriced (short-premium friendly). Ratio > 1 = realized range exceeded implied (structure got run over or long premium caught it).

Loading…

Implied Skew same-expiration multi-leg only

IV spread across legs at entry (max − min). Flat <0.5pp, mild 0.5–2pp, steep >2pp. Does skew steepness predict outcome for your verticals / condors / flies?

Loading…

Ex-Dividend Exposure short premium on American-style ETFs

Trades holding across a quarterly ex-dividend date face elevated early-assignment risk. Calendar is hardcoded for SPY/QQQ/IWM — extend in app/services/options/ex_dividend_calendar.py.

Loading…

Pin Risk by underlying distance to strike at exit

Short-premium tents benefit from pinning; long premium and breached shorts lose. If at_pin underperforms safe, exit earlier.

Loading…

Entry Timing (Eastern)
GEX Anchor Level

Win rate by which GEX level was nearest to the strike

Vol Trigger Position

Win rate by price position relative to zero gamma at entry

GEX Level Proximity

Win rate by distance from nearest GEX wall/level

Global Insights cross-strategy

Patterns that span your entire options book — "iron flys on Mondays", "calendars at 10:00 entry", etc. Each insight is bucket WR vs the rest of the book, tested at the configured confidence threshold.

Confidence 90%

No global insights yet. Click "Mine Patterns" to run across your book.

Pre-Entry Advisories alert-only

Real-time warnings when a prospective entry matches a known weak bucket (block / warning) or affirms a strong one (info). Driven by the cross-strategy Global Insights above.

No advisories yet. They accumulate as live options signals fire; you can also use Preview above to test.

Recommendations
Confidence 90%

No recommendations yet. Select a strategy and click Generate.

Recommendations controls — how they work

Three knobs control how the options learning engine proposes and scores parameter changes to your strategies. Here's what each one does.

Confidence slider

Minimum statistical confidence (= 1 − p-value) a comparison must reach before the engine emits a recommendation. Default is 90% (i.e., p < 0.10).

  • For each analysis, best vs worst bucket is tested using a two-proportion z-test (n ≥ 30 per group with healthy win/loss counts) or Fisher's exact test at smaller n.
  • Each bucket needs at least 3 trades to qualify for comparison.
  • Lowering the slider surfaces marginal recs that narrowly missed the bar — useful for tuning. Raising it requires stronger evidence.

Generate

Runs the recommendation engine on the selected strategy (or every options strategy you own, via All options strategies). It analyzes ~7 dimensions — IV rank, delta, entry hour, exit tier, GEX anchor, vol trigger, GEX proximity — and emits a rec whenever the best-vs-worst win-rate difference clears the confidence threshold.

  • Each rec shows the actual bucket comparison, the statistical test used, and a narrative of why it's being suggested.
  • Pending recs that a re-run no longer supports are marked stale — the data or threshold has changed since they were first generated.
  • The Why no recommendations? panel lists every analysis block that was skipped (low coverage, not significant, no change needed).

Evaluate verdicts

For each applied recommendation, compares the trades made under the new configuration against the prior version's baseline window and writes a verdict: ✓ helping ⚠ hurting — neutral pending

  • Uses the same statistical tests as Generate; classification requires ≥ 5 post-apply trades (or 14+ days since apply) plus significance above the default threshold.
  • A nightly job runs this automatically at 4:45 PM ET on weekdays. This button forces it now — useful right after you apply a rec and want to stamp its starting state.
  • The verdict badge appears next to each applied rec, with reason text showing the p-value and post/baseline win rates.

Note: Each Generate run also marks pending recs on the same strategy as stale if the current data no longer re-emits them. That way you'll never accidentally Apply a rec the engine has already walked back.